Ville Pulkki, Matti Karjalainen — Communication Acoustics- An Introduction to Speech, Audio and Psychoacoustics

1060 published claims from this book.

• A 10-15 percent bandwidth window serves spectral analysis for audio tech
• A 30-40 ms delay separates fusion from echo in the precedence effect
• A competing tone at another frequency suppresses a cochlea channel’s response
• A complete mathematical model of the entire auditory system will never exist
• A concert hall has 24 billion acoustic degrees of freedom
• A concert hall’s acoustic field has billions of vibrational degrees of freedom.
• A dedicated cone-of-confusion coordinate system describes hard-to-discriminate sound directions.
• A full speech communication chain spans brain, vocal organs, channel, and hearing.
• A good microphone needs flat response from 20 Hz to 20 kHz.
• A guitar body’s effect can be simulated by convolution with its measured impulse response
• A healthy cochlea emits new frequencies when stimulated by two tones
• A holistic mathematical theory of the full auditory system is unattainable
• A human listener is needed to connect objective sound pressure with perceived loudness
• A kettledrum’s tone is perceived with a missing fundamental
• A mass on a spring forms the simplest vibrating system
• A microphone’s sensing principle determines whether its pickup pattern is omnidirectional, dipolar, or cardioid
• A plus or minus 5 dB loudspeaker deviation goes unnoticed without a direct reference
• A set of harmonically related partials shifted together still changes perceived pitch.
• A single amplitude envelope cannot capture how separate spectral parts evolve
• A single amplitude envelope misses that harmonics decay and modulate at different rates
• A single auditory nerve fibre’s firing is unpredictable and noisy
• A single feedback process yields complementary onset-emphasis and loudness-integration outputs.
• A single note can sound rough only below roughly 100 Hz fundamental
• A sound signal is the registered time-varying vibration of a medium
• A sound’s onset transients largely determine instrument identity
• A stationary tone’s frequency alone can shift perceived elevation in the median plane.
• A system’s output depends on its remembered history
• A temporary threshold shift signals elevated risk of permanent hearing loss
• A unified source-filter model describes human and synthetic voice alike
• A vibrating string has infinitely many resonant frequencies
• A-weighted decibels are a rough practical approximation of perceived loudness
• A-weighted sound level approximates perceived loudness only crudely, yet remains the noise-measurement standard.
• Above a critical frequency, room acoustics become statistically tractable as diffuse fields.
• Absolute pitch is roughly ten times more precise than trained relative pitch.
• Abstracting away details and concretizing them frame scientific research as one cycle
• Accuracy and function-shape information trade off against speed in threshold measurement
• Acoustic engineering mostly needs cochlear knowledge, but spatial audio relies on SOC function
• Acoustic engineering techniques should be built from perceptual mechanisms, not physics alone.
• Acoustic engineering usually needs only cochlear models, not higher nuclei
• Acoustic reflections arise at any change in tube cross-sectional area
• Acoustics, perception, and signal processing underpin sound-technology R&D
• Adaptation, temporal integration, and post-masking can be simulated by one combined feedback element.
• Adaptation, temporal integration, and post-masking can be simulated in a single auditory element.
• Adaptive filtering changes filter coefficients over time in response to signals.
• Adaptive filters cancel hearing aid whistling caused by earphone-microphone feedback
• Adaptive filters enable prediction, noise cancellation, and echo removal
• Adaptive staircase procedures converge to thresholds by adjusting signal level per answer
• Adaptive staircase procedures trade efficiency and bias avoidance against convergence risk
• Adaptive threshold methods trade fast converging results against convergence risk
• Adjacent cochlear nucleus axons fire at matched phase despite differing absolute start times
• Adjacent-band aliasing cancels by opposite phase when bands are unmodified
• Adjustment steps sized near the JND prevent subjects overestimating imperceptible changes
• Adult cochlear implant success depends on prior experience with sound
• Aerophone design tunes resonance modes so the instrument becomes musically usable
• Aesthetic sound-quality reactions depend strongly on sociocultural background
• Air absorption and room acoustics can be modeled as low-pass and arbitrary filters
• Aliasing creates signal components that never existed in the original
• Aligned harmonic phases make sounds sharper and buzzy
• All sound quality judgments derive from properties of the human auditory system.
• All vowel distinctions reduce to tongue position, openness, and lip rounding
• All-pass filter colorlessness only holds when delay is under the ear’s integration time
• AM tones change perceived frequency spread, giving extra detection cues that noise lacks.
• Ambiguous notes are assigned to one competing stream, altering perceived rhythm.
• Amplitude modulation of noise creates pitch with a flat spectrum
• Amplitude panning creates auditory objects with no real physical source
• Amplitude panning produces equivalent spatial perception across headphones and loudspeakers.
• Amplitude-modulated loudness directly increases perceived fluctuation strength
• Amplitude-modulated tones gain an extra detection cue that noise and FM tones lack
• Amplitude-modulated white noise sounds pitched at the modulation rate.
• Amplitude-modulation detection becomes easier once modulation spreads the spectrum beyond one critical band
• Amplitude-modulation detection improves with tone level but is level-independent for noise.
• An open car window creates a living Helmholtz resonator with strong low-frequency pressure fluctuations
• Analogue, discrete-time, and digital signals form a processing chain.
• Analysis of variance tests whether any independent variable affects a dependent variable.
• Anechoic chambers misrepresent how loudspeakers sound in normal rooms
• Annoyance and disturbance are distinct components of perceived noise quality
• Annoyance and disturbance refer to distinct degrees of noise harm
• ANOVA tests whether independent variables affect listening-test outcomes
• Anti-phase signals vs diotic noise yield 18 dB binaural detection advantage around 200-300 Hz
• Any band-limited linear time-invariant system can be implemented as a digital filter
• Any continuous signal decomposes into sinusoidal partials and convolution simplifies in frequency domain
• Any discontinuity in a medium reflects an incident wavefront
• Any phone of a language can be synthesized with 4-5 parallel formant filters
• Any sound can function as noise depending on the listener
• Any vibration decomposes into weighted sums of a system’s modal oscillations
• Any wave can be decomposed into two counter-travelling component waves.
• Arbitrary-sound roughness is estimable by summing per-critical-band contributions
• Articulation scores require logarithmic modification to attain additivity
• Articulation shapes the voice by filtering glottal sound through the vocal tract
• Articulatory synthesis models speech dynamics but is too complex for practical synthesizers
• Ascending and descending stimulus runs are averaged to derive perceptual thresholds
• Assessors can use their own vocabulary instead of agreeing on descriptors
• Audio analysis bands can follow perceptual scales while processing stays on full-resolution QMF bands.
• Audio and speech technology design should be grounded in perceptual mechanisms
• Audio codec testing requires methods that reveal unknown perceptual attributes
• Audio coders prefer critically sampled filter banks for minimal data transmission.
• Audio content is defined by meaning to a listener, not technical properties
• Audio decorrelation creates perceptually transparent extra independent channels from few signals
• Audio engineering is the production of meaningful sound signals across media.
• Audio production divides into recording, mixing, mastering, and live sound stages
• Audio quality improvements mask video defects, but video quality does not mask audio defects.
• Audio watermarking embeds inaudible, copy-surviving data into signals
• Auditory adaptation acts like high-pass filtering that emphasizes stimulus onsets.
• Auditory bands from the same source are summed to compute total loudness
• Auditory cortex adaptation decreases response to frequent sounds, shaping perception
• Auditory cortex specifically attenuates responses to frequently repeated sounds
• Auditory displays deliver computer information to humans through sonification and earcons
• Auditory distance is underestimated beyond 3 m and overestimated within 1 m
• Auditory distance localization error varies dramatically across studies.
• Auditory elevation calibration requires visual confirmation of sound sources
• Auditory experimentation resembles measuring an unreliable device against an unpredictable system
• Auditory loudness doubles with every 10 dB increase in sinusoid level
• Auditory masking spread can be modelled by convolving with a spreading function
• Auditory modelling accuracy trades off sharply against computational cost
• Auditory models link acoustic stimulus features to perceived attributes computationally.
• Auditory models that extract distance cues from ear signals remain an open research direction
• Auditory nerve adaptation operates like automatic gain control.
• Auditory nerve firing is stochastic, requiring many fibres pooled.
• Auditory nerve phase locking degrades above 1500 Hz and is weak above 4 kHz
• Auditory nerve responses can be experimentally recorded by microelectrodes.
• Auditory perception acts as an unreliable measurement device on a nonlinear system
• Auditory perception must be studied indirectly through measurable external reactions.
• Auditory perception runs parallel bottom-up and top-down processes
• Auditory scene analysis organizes sound events into perceptual streams
• Auditory spatial information alone allows orientation in one’s surroundings.
• Auditory spatial perception emerges from weighted combination of localization cues
• Auditory spectra smooth harmonic detail while preserving formant structure
• Auditory spectrum computation begins with FFT then emulates hearing sensitivity via filtering and weighting.
• Auditory-model quality metrics fail on acoustically captured signals
• Auditory-tactile systems improve music reproduction quality and enable tactile feedback.
• Authentic sound reproduction is judged by perceptual identity, not signal fidelity
• Autocorrelation is the dominant method for linear prediction analysis
• Autocorrelation models explain pitch and can separate concurrent vowel sounds differing in fundamental frequency
• Autocorrelation reveals a signal’s fundamental periodicity through periodic peaks.
• Automatic gain control compresses sound to fit impaired hearing’s reduced dynamic range.
• Average speaking fundamental frequency is 120 Hz for men and 200 Hz for women
• Backward masking is a weak and unreliable effect appearing mainly in novice listeners.
• Bandwidth extension and error concealment are speech-enhancement techniques for missing or narrow data.
• Basilar membrane models range from transmission-line equivalent circuits to finite element methods
• Beamforming combines multiple microphone signals into directive beams
• Békésy audiometry derives hearing thresholds from button-press tracking of a swept tone
• Bending stiffness makes bar vibrations dispersive and inharmonic.
• Best auditory frequency resolution is about one-sixth octave at moderate levels
• Beyond intelligibility, telephony must preserve speaker recognizability
• Beyond the reverberation distance, loudness stays constant while reverberance increases.
• Binaural auditory models cannot yet explain human perception of complex spatial audio
• Binaural hearing aid processing preserves spatial cues that monaural processing distorts
• Binaural hearing yields up to 18 dB detection advantage for low-frequency signals
• Binaural intelligibility level difference measures speech unmasking benefit
• Binaural listening gives 0 to 8 dB advantage in multi-talker conditions
• Binaural listening yields a 5-7 dB intelligibility advantage over monaural
• Binaural listening yields only 5–7 dB advantage over monaural spatially-separated sources
• Binaural models differ on whether the brain uses narrow or wide directional receptive fields
• Binaural recordings capture natural direction cues only when ears match the listener
• Binaural unmasking can be modelled as equalizing then subtracting two ear signals
• Blind source separation becomes very difficult when sources exceed microphones
• Blind-test design and task wording eliminate bias in hearing studies
• Bone conduction through the skull shapes how one hears one’s own voice
• Bone-anchored hearing aids bypass outer and middle ear entirely
• Bottom-up neuron-by-neuron auditory modelling is computationally and knowledge-wise infeasible.
• Bowed string motion is a stick-slip effect producing a sawtooth wave
• Bowed strings alternate between stick and slip motion
• Brief silences are perceptually longer than their physical duration.
• Brief unprotected noise exposure during otherwise protected periods raises hearing loss risk
• Broad-spectrum sounds read as louder because auditory bands integrate across the basilar membrane
• Careful listening tests validly reveal how acoustic attributes become auditory attributes.
• Carrier and envelope time shifts produce lateralization over complementary frequency ranges
• Cars can synthetically reproduce engine sound through the audio system.
• CASA aims to extract all perceivable information from ear canal signals.
• Cascade AGC loops compress steady signals logarithmically while amplifying onsets.
• Cascaded audio compressors in broadcast chains degrade sound quality cumulatively
• Causality forces real systems to lack knowledge of future input
• Chance-level guessing forces threshold criteria above 50% to reduce false positives
• Changing a tone’s frequency alone can make listeners perceive motion without source movement
• Cinema 5.1 surround sound originated in 1976 and reached homes in the late 1990s.
• Cinema has historically pioneered new audio reproduction formats before home use.
• Clean auditory sensation emerges only from pooling many noisy nerve fibres
• Close-miked mono reproduction stays clear because the room barely affects the signal
• Closed headphones gain isolation but become sensitive to leakage effects.
• Cochlear hydrodynamics can be simulated by analog electrical equivalent circuits
• Cochlear implants bypass the conductive auditory path to stimulate nerve fibers directly
• Cochlear liquid mass determines where each frequency penetrates the basilar membrane.
• Cochlear modelling faces a tradeoff between spatial accuracy and computational cost
• Cochlear outer hair cells enable speech amid noise and reverberation
• Cochlear pitch mapping exploits the mass of vibrating liquid varying with route length
• Cognitive context determines whether a sound feature is good or bad
• Coherent amplitude modulation across frequency bands may drive auditory source grouping.
• Coherent sound addition yields double the amplitude gain of incoherent addition
• Coherent sounds interfere by amplitude while incoherent sounds sum by power.
• Coincidence detection models were extended to also encode ILD sensitivity
• Comb-filtering undulations from room reflections move to bass and become inaudible
• Combining compression with a limiter reduces distortion better than a limiter alone.
• Combining DRNL paths produces a double impulse that loses temporal fine structure
• Commercial recordings aim for plausibility, not authentic reproduction
• Commercial telephone bandwidth standards derive from interwar human listening tests
• Communication acoustics joins basic research and engineering toward shared goals
• Comparative MOS ratings allow a test sound to score above the reference.
• Comparing bone- and air-conduction thresholds distinguishes conductive from sensorineural hearing loss
• Complex QMF resolution must be supplemented to match auditory frequency selectivity
• Complex real-world maskers are less effective than simple energetic ones predict
• Compressible air in a volume forms a mechanical spring for resonance
• Compression, limiting, and expansion attenuate or boost audio differently by amplitude.
• Compressors, limiters, and expanders each apply different gain rules to signal level.
• Computational auditory modelling began only in the 1960s
• Computational auditory models serve knowledge, audio-quality measurement, and diagnostics
• Computational modelling is erasing the boundary between theory and experiment
• Computational modelling is merging experimentation and theory into one method
• Computational models can synthesize natural ambient sounds like footsteps and thunder
• Computational models extract quantitative metrics for the ear’s perceptual dimensions
• Computational simulation can generate testable hypotheses about auditory physiology
• Computational simulation is the best method to model the auditory system’s function
• Computed room impulse responses often sound unnatural in the late reverberant tail
• Computer simulation dramatically expanded modeling’s power over physical models
• Computer-displayed continuous-looking curves are actually stored sample sequences
• Computerized speech handling required simplified phonetic notations tailored per language
• Concert hall acoustics shape the music composed for them
• Concert hall quality is codified into standardized subjective attribute vocabularies
• Condenser microphones dominate modern audio because they convert sound to electrical signals efficiently.
• Conductive and cochlear losses produce different loudness growth patterns
• Cone of confusion makes time and level cues ambiguous for elevation
• Consensus vocabulary methods have a panel jointly build shared sensory descriptors
• Consensus-based sensory panels risk leader bias but training can counter it
• Consonant classes are produced by distinct vocal tract closure configurations
• Consonants are less intense and more context-dependent than vowels.
• Consonants differ from vowels in intensity, context-dependence, and articulation constriction
• Constant stimuli remains gold standard but is too slow for routine use
• Constant-stims method maps full psychophysical functions rather than single thresholds
• Consumer devices are engineered to sound masculine and powerful, not merely quiet.
• Consumers historically equated vacuum cleaner loudness with power.
• Context-dependent compensation shapes sound perception much as with color vision
• Continuous-excitation instruments exhibit slow attacks and steady sustain levels
• Continuous-excitation instruments like flutes attack slowly while impulse-excited ones do not
• Converting continuous sound to digital requires time sampling then numeric quantization
• Convolution mathematically links LTI system inputs to outputs.
• Count-comparison and coincidence-detection models imply different brain receptive-field architectures for direction coding.
• Coupled masses vibrate either along or perpendicular to spring direction
• Coupled oscillators produce complex resonances shared by instruments, voices, and rooms
• Cross-correlation sidelobes near 1 kHz stem from interaural phase ambiguities
• Crosstalk-cancelled playback yields externalized sound only in the frontal hemisphere
• Cubic difference tones reveal nonlinear cochlear mechanics even at low sound levels
• Data-based speech synthesis splits into sample-reassembly and parameter-model approaches.
• Deaf, hard-of-hearing, and deafened denote distinct functional categories
• Decorrelation enables creating independent channels while preserving perceptual quality.
• Decorrelators degrade audio quality on transient-rich signals like applause.
• Delayed identical sounds fuse into one source until a 30-40 ms echo threshold
• Delayed identical sounds merge spatially until delays exceed 30-40 ms
• Descriptive sensory analysis maps perceived sound differences onto an N-dimensional attribute space
• DFT-based psychoacoustic models derive an auditory excitation spectrum on the ERB scale
• Different auditory features require different subsets of model elements and time constants.
• Different hearing tasks exploit different model time constants.
• Different musical genres are defined by how much they use dissonant chords
• Digital audio coding made perceptual modeling central to audio engineering
• Digital audio growth made understanding human audition essential
• Digital audio’s rise made auditory perception knowledge essential to engineers
• Digital signal processing is inherently band-limited by reconstruction filters
• DirAC divides captured sound into diffuse and non-diffuse streams
• DirAC reproduction quality depends on one dominant source per frequency band
• Directional coincident microphones suppress room reverberation, killing sense of space
• Directional hearing resolution is worst near 1.7 kHz where head size equals one wavelength
• Directional hearing rests on interaural time and intensity differences.
• Directional hearing-aid microphones improve speech-in-noise SNR by over 7 dB.
• Dispersive waves in stiff bars produce inharmonic modal frequencies
• DIY in-ear binaural recording yields convincing spatial reproduction
• Doppler shift in VR is modeled by dynamically adjusting sample rate or propagation delay
• Doubling perceived pitch of a 2-kHz tone requires about a 15-kHz frequency jump
• Doubling sound level from 30 to 90 dB only doubles perceived sharpness
• Downsampled filter banks enable efficient DSP implementations like FFT for DFT.
• DRNL filter outputs smear temporal fine structure into double impulses
• Dynamic auditory cues can override even strong conflicting visual evidence
• Dynamic binaural head-movement cues override conflicting visual evidence of source location
• Dynamic head tracking restores externalized sound localization in binaural reproduction.
• Each added quantization bit yields 6dB SNR improvement
• Each inner hair cell responds best to one frequency plus a widening critical band
• Each new cochlear structure discovered was cast as the resonant sound receiver
• Each person’s directional hearing is calibrated to their own ear shape
• Each person’s unique pinna and head shape undermines binaural recordings for other listeners
• Each STFT sample represents a circularly continuous tone
• Ear offset from the head’s center makes arrival time shift with source azimuth
• Eardrum and middle ear status can be measured objectively via acoustic impedance
• Echo delay magnitude in telecommunication maps to distinct perceptual degradations.
• Echo delay thresholds produce distinct perceptual and speaking effects
• Echoes and reverberation convey room geometry and surface properties to listeners
• Echolocating bats build image-like environmental representations from reflected chirps
• Electret microphones trap charge in material so no external voltage is needed
• Electrostatic transduction, the condenser microphone’s inverse, delivers the highest headphone and speaker quality.
• Element-based acoustics methods grow computationally costly at higher frequencies
• Enclosures prevent acoustic short-circuiting that silences low frequencies
• Engineering perspectives increasingly serve computational modelling of auditory perception.
• Entering a car generates designed auditory-tactile impressions of finishing quality
• EQ cannot fix directivity problems that a better loudspeaker requires
• Equalization-cancellation subtracts masker signals to reveal hidden targets
• Equalizers invert measured system impulse responses to flatten magnitude response
• Equalizing a loudspeaker response can worsen off-axis sound quality.
• EQUEST injects temporal masking knowledge into echo annoyance estimation
• Estimated ITDs match theoretical values especially at low frequencies
• Even ideal equalization degrades off-axis response and can amplify distortion and noise
• Even moderate 70 dB sound causes temporary hearing threshold shift with hours of exposure
• Every sounding product is continuously evaluated for sound quality
• Evolution produced language and engineering, but machines still lag in pattern recognition
• Evolution shaped localization to favor broadband transient sounds like movement
• Exaggerated interaural level differences signal that a sound source is near
• Excess noise damages cochlear hair cells through several distinct mechanisms
• Excitation in an elastic medium generates waves from the source
• Excitation spread across frequency bands is modelled as convolution with a level-dependent spreading function.
• Fast amplitude modulations within a critical band produce roughness perception
• Feedback (IIR) filters can become unstable, while feedforward (FIR) filters cannot
• FFT block convolution trades speed for double-block latency in reverbs
• Filter bank models outperform DFT in mimicking auditory temporal resolution
• Filter banks drive equalization, audio coding, and spatial reproduction
• First-order Ambisonics reproduction is restricted to three loudspeakers
• Fixed-width filter bands mismatch hearing’s finer low-frequency resolution
• Flat-gain amplification fails for frequency-dependent hearing losses
• Fluctuation strength is standardized against a 4 Hz amplitude-modulated reference tone
• FM synthesis generates harmonic spectra when the modulating frequency equals the carrier frequency
• fMRI identifies active brain regions well but is noisy and time-slow for auditory study.
• fMRI measures slow metabolic changes and is noisy, complicating auditory research
• Forced-choice guessing floors push threshold criterion above 50 percent detection
• Formant tracking to harmonics boosts loudness, not intelligibility, in soprano singing
• Formants are tract resonances conveying speech information
• Forward masking decays within roughly 200 ms after sound offset
• Forward masking persists up to 200 ms after the masker ends
• Forward masking suppresses sounds for 150-200 ms after masker offset even at high levels.
• Free sorting adds post-hoc verbal labels that enable interpreting perceptual dimensions.
• Free sorting lets subjects define their own perceptual similarity criteria
• Frequency discrimination is finest around 0.3% near 500 Hz.
• Frequency JND is best around 500 Hz at about 0.3 percent
• Frequency masking is modelled as convolution with a spreading function
• Frequency modulation detection depends on both carrier and modulation frequency.
• Frequency spread in the cochlea is asymmetric with steeper low-frequency slopes
• Frequency-domain analysis mirrors how the human ear processes sound
• Frequency-domain differences track perceived distortion better than time-domain ones
• Frequency-domain processing enables efficient audio coding for monophonic and multichannel signals
• Frequency-specific hearing loss drives multi-band hearing aid processing
• Fretted instruments set pitch by vibrating string length
• Functional and physiological modeling approaches divide along intended use.
• Gain ratios alone cause loudness shifts; sum-of-squares normalization prevents them
• Game-like testing yields less biased hearing-aid results in children than formal tasks.
• Gammatone filter banks model human auditory filtering using measured nerve responses
• Gammatone filters fail on level-dependence and fast-onset sounds
• Generalized procrustes analysis aligns different subjects’ attribute ratings into common dimensions.
• Gestalt grouping principles apply to hearing as well as vision
• Glottal closing generates most vocal-tract excitatory energy.
• Glottal function is modeled as mass-spring or distributed transmission-line systems
• Grey’s timbre findings confirm that note onset transients largely determine perceived instrument identity
• Grounded theory builds sound-quality explanations from coded interview notes
• Group delay affects perceived quality only when it varies with frequency
• Group delay describes the time-domain behavior of a transfer function
• Group delay of modulation is the phase representation most relevant to hearing
• Group delay perception thresholds depend strongly on signal content and context
• Hair cells actively amplify cochlear vibration rather than only sensing it
• Hair-cell adaptation to sound arises from neurotransmitter pool depletion and recycling dynamics
• Haptic-audio timing errors are as noticeable as audio-visual ones
• Hard-walled rooms behave as three-dimensional resonators with many audible modal frequencies
• Head and pinna scattering imprint direction-dependent spectral notches onto the ear signal.
• Head movements resolve front-back ambiguity by creating direction-specific binaural cue changes
• Head scattering creates level differences between the two ears’ received signals.
• Head separation between ears produces arrival-time and intensity cues for direction
• Head size alone predicts interaural time differences well across azimuth angles
• Head symmetry creates a cone of directions where sound localization is ambiguous
• Head-tracked HRTF filtering restores realistic external sound localization
• Head, torso, and shoulder acoustics shape the localization cues reaching each ear.
• Headphone and loudspeaker playback of stereo differ yet yield similar spatial order.
• Headphone audio internalizes because static binaural cues signal an inside-the-head source
• Headphone binaural reproduction fails without dynamic head-movement cues
• Headphone cushion leakage degrades low-frequency response but spares mid and high frequencies.
• Headphone cushion leaks alter only the low-frequency transfer to the eardrum
• Headphone group delay alters audibility of phase-spectrum effects
• Headphones come in intra-aural, supraaural, and circumaural form factors.
• Headphones originated as noise isolation for radio operators
• Headphones sound ‘in the head’ because they fail to model head movement cues
• Healthy ears continuously emit faint spontaneous sounds linked to frequency sensitivity
• Hearing acts as a spectral analyser extracting quantifiable perceptual metrics from sound
• Hearing aids are miniature signal-processing sound systems that must secure adequate signal-to-noise ratio
• Hearing damage begins after long exposure above 85 dB
• Hearing evolved as an omnidirectional alarm system complementing directional vision
• Hearing forms loudness by integrating sound energy over about 100-200 ms
• Hearing groups frequency bands by shared envelope modulation, reducing masking
• Hearing impairment degrades sensitivity, discrimination, and can cause hyperacusis or tinnitus
• Hearing impairment degrades signal-to-noise tolerance more than quiet speech perception.
• Hearing integrates sound intensity over a fixed 200 ms window
• Hearing is a staged chain from air capture through mechanical conduction to neural processing
• Hearing loss varies independently across different frequency ranges
• Hearing perceives synchronized neural firing across critical bands as buzzy timbre
• Hearing repackages one-dimensional sound signals into multidimensional time-frequency objects
• Hearing segregates sounds by frequency unless they fall within a critical band
• Hearing sensitivity degrades with age and is lower in early childhood
• Hearing sensitivity to frequency can be emulated by filtering and weighting short-time power spectra
• Hearing serves both environmental orientation and communication since early animal evolution
• Hearing thresholds are defined for steady tones lasting over 100 ms
• Hearing waveform beats proves the ear detects time-domain shape
• Hearing-aid engineers’ normal hearing blocks firsthand testing of devices
• Hearing-aid users and small children require loudspeaker-based audiometry for accurate coupling
• Hearing-impaired listeners need higher signal-to-noise ratios than normal hearers do
• Hearing-loss risk dominates noise research because it is predictable from sound properties
• Hearing-loss risk threshold should drop from 85 to 75 dB for full protection
• Hearing-modeling spectrograms use an ERB frequency axis from modeled auditory filters.
• Hearing’s lower limit is the minimum audible field; upper is pain
• Hearing’s temporal resolution varies from 1-2 ms depending on frequency
• Helmholtz resonance shapes violin bodies, speaker designs, and car cabin noise
• Helmholtz resonators serve as transducers, amplifiers, and sound controllers.
• Heterarchical, distributed control can outperform strict hierarchy under changing conditions.
• Heterarchies distribute control for versatility while data become information through observation
• High-frequency auditory analysis preserves only the temporal envelope of signals
• High-pitched soprano singing is unintelligible because harmonics are too sparse for formants
• High-pitched soprano singing sacrifices intelligibility for audibility
• High-quality synthesized speech requires tens of hours of manually-segmented recorded speech.
• Higher-level hearing models rely on assumptions tested only against psychoacoustic data
• Highest common factor of partials encodes perceived pitch
• HRTF definition normalizes ear responses by the head’s absence
• HRTFs can virtualize multi-channel loudspeaker layouts on headphones.
• Human audio-visual synchronicity tolerance is asymmetric because sound naturally lags light.
• Human auditory adaptation neutralizes room acoustics but not device imperfections
• Human auditory memory cannot compare sounds separated by more than a few seconds
• Human hearing adapts to different room acoustics to preserve timbre perception
• Human hearing can in principle distinguish astronomically many timbres.
• Human hearing diverged from other mammals mainly via speech analysis
• Human hearing excels at comparative judgments but fails at absolute ones.
• Human hearing is highly nonlinear and cannot be LTI-modelled beyond peripheral parts
• Human hearing is most sensitive to sounds around 3-4 kHz.
• Human hearing resists linear time-invariant modeling except at the periphery
• Human hearing sensitivity peaks near 4 kHz and falls off at both extremes
• Human hearing’s 120-dB dynamic range exceeds man-made device capabilities
• Human loudness-modulation detection is fastest around 4 Hz
• Human sound localization is remarkably accurate given it processes only two ear signals.
• Human spatial resolution limits make stereo-to-multichannel upmixing perceptually plausible.
• Human spectral analysis resolution runs about 10–15 percent of center frequency
• Human speech evolved as a robust channel tolerant to high background noise
• Humans adapt to a room’s spatial cues through learning over hundreds of trials
• Humans adapt to room acoustics through training over hundreds of trials
• Humans are perceptually tuned to amplitude and frequency modulations across specific rate bands
• Humans can detect interaural timing differences as small as ten to twenty microseconds
• Humans can isolate a single speaker amid competing voices and noise
• Humans can match durations of silence to sounds of similar length.
• Humans cannot reliably report auditory attributes on absolute internal scales.
• Humans detect level changes of about 1 dB at moderate volumes
• Humans localize single sources well but fail at source distributions
• Humans localize sound accurately using only two real-time ear signals
• Humans perceive a fundamental frequency even when it is absent from the signal
• Humans perceive interaural level differences at frequencies where plane waves produce none
• Ideal audio reproduction permits gain change and pure delay without degrading quality
• Idiophone and most membranophone pitch cannot be controlled by the player
• IIR filter feedback can cause instability unless poles stay inside the unit circle
• Image-source room acoustics modelling becomes computationally costly in interactive VR
• Impulsive sounds grab attention because they signal physical state changes
• Impulsive transients grab attention by signaling state changes in physical objects
• Impulsiveness is estimated from level statistics rather than directly measured
• In steady state, reverberant sound level is uniform throughout a room.
• In tonal languages like Chinese, pitch changes alter the phoneme itself
• Individual HRTF measurements give the most accurate external ear emulation
• Individual notes are assigned to exactly one auditory stream, altering perceived rhythm.
• Industry relies on objective methods because exhaustive listening tests are impractical
• Informational masking arises when listeners can segregate signal from masker at higher processing levels
• Inharmonic sounds like bells can produce multiple simultaneous pitches
• Inharmonic tone complexes can yield multiple or virtual perceived pitches
• Instantaneous interaural level differences improve intelligibility beyond better-ear listening alone
• Instrument bodies act as coupled resonators that shape and radiate sound
• Instrument directionality shapes the captured spectrum depending on mic placement
• Instrument sound envelopes are described by attack, decay, sustain, and release phases
• Instrument sounds are endlessly varied yet fall into families sharing generation principles
• Instruments spanning multiple sound-generation classes force contested classification
• Instruments spanning multiple vibration families resist clean classification
• Inter-loudspeaker delay shifts perceived virtual source position toward the earlier sound.
• Interaction with the user is what distinguishes virtual reality audio from traditional reproduction
• Interactive audio applets teach sound better than static text can
• Interactive game audio requires dynamic rendering systems beyond traditional linear production.
• Interactive rendering differs from reproduction because user actions alter the sound.
• Interaural coherence measures similarity of ear canal signals
• Interaural level differences and time shifts trade off to control lateralization
• Interaural level differences only emerge at frequencies above roughly 400 Hz.
• Interaural time difference ranges roughly 0-700 microseconds across azimuth
• Interaural time differences are computed by cross-correlating the two ear signals
• Internal models of reality must trade completeness for tractability.
• Interval dissonance increases with position higher in the harmonic series
• Introspection alone cannot serve as accepted scientific auditory observation
• Inverted-polarity (antiphasic) stereo signals cancel lows and smear image
• Ion concentration differences across the basilar membrane create an electric potential.
• IPA’s complexity drove development of computer-friendly phonetic notation systems
• ITD spans 0-700 microseconds across azimuth independent of frequency
• Jeffress coincidence detection explains how neurons encode interaural time differences.
• Just intonation leaves many intervals audibly mistuned like the D-A fifth
• Just-noticeable SPL differences stay constant near 1 dB across wide ranges
• Known excitation signals sidestep the need for physically impossible impulses
• Lack of hearing diagnostics once caused deaf children to be mislabeled intellectually disabled
• Language knowledge beyond acoustics is required for automatic speech recognition.
• Languages differ widely in vowel category counts, from two to over ten.
• Languages range from two to over ten vowel categories
• Languages vary in whether timing or syllable quantity carries meaning
• Limited human spatial resolution allows plausible upmix from fewer source channels
• Linear audio distortion is easy to equalize while non-linear distortion resists compensation
• Linear prediction models signals as all-pole IIR systems predicting samples from prior ones
• Linear prediction order scales with sample rate to capture speech formants.
• Linear time-invariant systems cannot generate new frequency components
• Listener personality biases reported auditory thresholds across subjects
• Listeners adapt to room-specific spatial cues through repeated exposure and template learning
• Listeners cannot report auditory attributes on an absolute scale
• Listeners perceptually subtract room coloration from source signals, called binaural decolouration.
• Listening room specifications exist to make psychoacoustic test results comparable across sites
• Listening test stimuli often differ across multiple unknown perceptual attributes simultaneously
• Listening-test statistics must inform experimental design from the outset
• Logarithmic quantization suits speech but distorts music beyond tolerance
• Logarithmically swept sines measure distortion and match human auditory frequency scaling
• Long-term averaging obscures transient attack and release sounds
• Longer masker sounds produce stronger forward masking
• Loud signals degrade frequency resolution but 10-15% bandwidth is a fair design rule
• Louder playback can bias listeners into rating a device better
• Louder playback is unconsciously rated as better quality in audio listening tests
• Louder signals need disproportionately more dB to double perceived loudness
• Loudness above 60 dB is encoded by cross-frequency spread, not single-neuron rate
• Loudness at high sound levels arises from excitation spreading across adjacent frequency bands
• Loudness is a subjective percept standardized against 1-kHz tone levels in phons
• Loudness models need 100-200 ms time constants to capture temporal integration and post-masking.
• Loudness perception sums auditory bands and signals distance by comparison
• Loudness rises once noise bandwidth exceeds the critical band at constant SPL.
• Loudspeaker level differences become inter-ear phase differences via cross talk
• Loudspeaker signals can be computed as virtual microphone signals with first-order directional patterns.
• Loudspeaker-room interactions remain the weak link in audio reproduction chains
• Loudspeakers and room acoustics remain the weak link in audio quality
• Low interaural coherence makes sound events seem diffuse and surrounding
• Low-frequency attenuation depends critically on proper seal and construction
• Low-frequency IACC peaks are broader because phase changes more slowly with time lag
• Low-frequency ILD serves as a distance cue and coherence detector
• Low-frequency interaural level differences cue source distance and signal coherence.
• LP inverse filtering flattens signals into spectrally whitened residuals
• LTI systems cannot create new frequency components from their input signals
• Machine hearing mirrors human perception by comparing feature vectors against known patterns
• Machine pattern recognition mirrors Gestalt grouping via feature vectors and similarity metrics
• Machines lead humans in logic tasks but lag in pattern recognition.
• Male speaker identity survives when the voice fundamental is filtered out
• Mannequin heads with ear microphones capture realistic spatial audio for recording
• Masking noise raises the loudness threshold and fades audible tones
• Mathematical source-filter models of speech enable practical synthesis and coding algorithms
• MDCT excels for coding but lacks robustness to time-frequency domain edits
• MDCT exploits inter-frame aliasing cancellation to achieve critical sampling
• Mean Opinion Score scales standardize audio-quality evaluation on a 1-to-5 range
• MEG localizes brain activity better than EEG because skull distorts signals less
• Method-of-limits thresholds are distorted by subjects’ response tendencies
• MFCCs remain the dominant feature representation for speech recognition systems
• Microphone construction determines directional sensitivity patterns like cardioid and dipole
• Microphone directional patterns range from omnidirectional to dipole to cardioid based on sensing principle.
• Microphone directivity shapes how reverberant a convolved signal sounds
• Microprocessor electronics made cochlear implants possible
• Middle ear acts as a band-pass filter attenuating 6 dB per octave
• Middle-ear impedance matching boosts pressure transfer roughly 30-fold
• MIR metadata is largely assigned by expert listeners, motivating automation as music creation accelerates
• Mobile phone codecs adapt their bit rate across eight levels to optimize call quality
• Modal density scales with spatial dimension, from flat in 1D to squared in 3D.
• Modal frequency density scales with dimensionality from constant to f-squared
• Modeling time-varying loudness requires emulating temporal integration and post-masking.
• Modern synthetic voices learn from single-speaker speech databases rather than hand-crafted phonetic rules.
• Modifying frame gains in overlap-add processing reveals previously cancelled aliasing artifacts as audible distortion.
• Modifying STFT bins creates edge transients that synthesis windows imperfectly suppress
• Monaural spectral cues alone permit sound localization with degraded accuracy
• Mono dominates because close-miked telephony sidesteps room coloration
• More reverberation relative to direct sound makes listeners judge a source as farther away.
• Most audio reproduction chain components are now perceptually flawless
• Most modern cochlear implant patients achieve telephone conversation ability
• Most real instruments are nonlinear and resist simple filter-cascade modelling
• Most sound arises from vibrating mechanical objects in a medium
• MTF measurement uses amplitude-modulated noise bands across 0.63 to 12.5 Hz
• Multi-stimulus quality tests fail when samples vary in several perceptual dimensions at once
• Multidimensional scaling maps perceived sound dissimilarities without labeling the perceptual dimensions
• Multidimensional scaling reveals dissimilarity structures without labeling perceived attributes
• Multidisciplinary problems cannot be mastered within a single knowledge domain
• MUSHRA only yields reliable results when sample differences are easily perceptible
• Music and speech are utility sounds distinguished by their primary role
• Music databases need automatic high-level content querying beyond raw audio signals.
• Music information retrieval serves both casual listeners and music professionals needing structured database queries
• Musical eras are distinguished by their systematic handling of sensory dissonance
• Musical instrument design evolved as early product sound quality experimentation
• Musical instrument sounds divide into attack, decay, sustain, and release phases
• Musical instruments couple excitation to resonances; human speech does not
• Musical notation is a lossy code requiring stylistic knowledge to realize
• Musical notation is a simplified code requiring stylistic knowledge to realize
• Musical pitch perception spans roughly 30 Hz to 4 kHz.
• Narrowband noise can carry clearer pitch than some periodic signals
• Natural and engineered systems organize as vertical control hierarchies
• Natural background sounds are synthesized statistically for virtual environments
• Neural networks trained on one subject’s data can predict human sound-elevation localization
• Neural networks trained on single-subject data can accurately model human elevation localization
• Neural refractory periods after strong onsets disable directional hearing for delayed sounds
• Neuron-by-neuron modelling of audition is computationally infeasible now
• Neurotransmitter loss in the synaptic cleft produces auditory adaptation.
• No practical impulse source is both powerful and perfectly omnidirectional
• No single ideal listening room acoustic exists because humans adapt to room differences
• No single microphone technique works everywhere, so engineers adapt per venue
• No single standard method measures perceived tonality accurately
• No tuning system can make all diatonic intervals have simple ratios
• No tuning system can make all diatonic scale intervals perfectly consonant
• Noise barriers block high frequencies but let low frequencies pass around them
• Noise barriers only effectively block high-frequency traffic sound, not low frequencies
• Noise combines measurable harm with subjective disturbance
• Noise control should prioritize source reduction then path attenuation before personal protection
• Noise damage creates a 4-kHz notch while aging affects high frequencies
• Noise engineering now targets annoyance and quality, not just hearing loss
• Noise exposure and age produce distinct hearing-loss signatures
• Noise is defined as sound that annoys, disturbs, or risks hearing damage
• Noise is defined by annoyance, disturbance, or hearing harm rather than loudness alone
• Noise is defined by its disturbing effect relative to a desired signal
• Noise is masked more easily by tonal parts than tones are masked by noise
• Noise is the most common cause of work-related impairments
• Noise research shifted from hearing-loss risk to subjective annoyance
• Noise suppression algorithms can degrade speech quality in non-stationary noise
• Noise with a sharp spectral peak is perceived as pitched.
• Non-linear acoustics governs high-pressure waves like shocks and cavitation
• Non-linear and time-variant systems generate frequency components absent from the input
• Non-redundant time-frequency transforms introduce aliasing artifacts when signals are modified in the frequency domain.
• Nonlinear equalization methods remain underdeveloped in audio engineering
• Nonlinear spatial analysis distributes measured impulse responses across loudspeaker arrangements
• Normal hearing yields a temporal resolution factor near unity
• Notched-masker tests prevent off-frequency listening to measure filter width
• Object-based audio decouples soundtracks from specific loudspeaker layouts
• Object-based audio renders sound per-loudspeaker via metadata at playback time
• Objective measures based on hearing models substitute for tedious listening tests of distortion.
• Objective models of subjective annoyance remain limited to specific noise types
• Objective sound quality models have limited domains of validity and cannot replace subjective testing
• Off-frequency listening can distort auditory filter measurement unless masked
• One sound can mask another even when their spectra do not overlap.
• Only part of loudspeaker input power becomes sound energy
• Opposing traveling waves of the same frequency create standing waves with nodes.
• Opposing waves do not annihilate because their energy just shifts form.
• Otoacoustic emissions objectively reveal auditory system functioning and defect location
• Outdoor wall and ground reflections produce repetition pitch in real settings.
• Outer hair cell damage compresses dynamic range, causing recruitment and distorted loudness perception
• Outer hair cell damage destroys the cocktail party effect
• Outer hair cell damage widens critical bands and worsens speech-in-noise
• Outer hair cells provide active frequency-selective amplification in the cochlea
• Overtone singing selects harmonics with sharp vocal-tract resonances to create melodies
• Paired comparison against anchors reduces bias in absolute judgments
• Paired comparison lets flash profiling skip time-consuming attribute training.
• Panning between non-symmetric loudspeaker pairs distorts perceived spatial width.
• PAQM estimates audible quality degradation by comparing specific loudness across critical bands.
• Peak dissonance between two tones occurs at a quarter critical band separation
• PEAQ uses a trained neural-network cognitive model to predict human MOS ratings.
• Perceived auditory distance follows a compressive power function of actual distance
• Perceived auditory distance has an acoustic horizon under 10 meters
• Perceived instrument character is shaped by the time evolution of sound amplitude
• Perceived loudness can be measured by iterative ratio-matching across subjects
• Perceived loudness depends on frequency content and duration, not just sound pressure level
• Perceived loudness of concurrent tones hinges on frequency separation via three distinct mechanisms
• Perceived loudness of time-varying sounds is set by short-time peak values.
• Perceived roughness from beating depends on basilar-membrane critical bands.
• Perceived sharpness is stable across subjects and rises with spectral centre of gravity
• Perceived-loudness matching often needs separate listening tests to achieve
• Perception is a brain interpretation that need not use all sensory detail
• Perception is an active process of filtering and organizing, not passive reception
• Perceptual attributes map onto multiple interacting physical sound properties
• Perceptual auditory models will eventually replace traditional audio metrics
• Perceptual hearing properties map onto objective signal analysis quantities
• Perceptual pitch ranking and direct electrical stimulation yield concordant electrode discrimination
• Perceptual structure analysis indirectly recovers individual feature-usage in auditory perception.
• Perceptual thresholds blur because listener internal noise varies by personality
• Perceptually distinct instrument timbres are practically infinite yet cluster into families
• Perfect sound reproduction requires an LTI system with unity transfer function
• Perfect sound reproduction tolerates arbitrary signal delay when reproducing recordings.
• PESQ works as a signal-difference device rather than a true auditory model
• Phase alignment determines whether sounds sum by amplitude or by power
• Phase randomization alters timbre more than 4 dB amplitude randomization
• Phase-based localization sensing has a hard frequency ceiling near 1.6 kHz
• Phonation types differ by glottal airflow and spectrum shape
• Phonetic information concentrates either in a phone’s steady core or its transition boundary
• Physical accuracy of a sound field does not predict perceived accuracy
• Physical, biological, and engineered systems share describable general properties
• Physically realizable systems cannot respond to future input values
• Piano hammers rebound and strike strings multiple times per note
• Piano hammers strike strings two to three times per keypress
• Pinna asymmetry creates front-back and elevation cues for directional hearing.
• Pitch depends only on spectrum and periodicity, not volume or direction.
• Pitch discrimination is best near 500 Hz and degrades at both extremes
• Pitch is not uniquely tied to a single physical frequency
• Pitch maps to repetition frequency but is not a simple physical property.
• Pitch perception can be modeled by autocorrelation across auditory filter bands
• Pitch perception requires a tone lasting at least two signal periods
• Pitch shift from uniform frequency change depends on overtone structure
• Place theory cannot account for level-independent pitch, so other mechanisms must operate
• Placing all string damping at terminations yields efficient digital synthesis
• Plosive consonants begin with silence then a burst when the tract opens.
• PLP coefficients describe speech spectral features compactly and speaker-independently
• Plucked string excitation is independent of vibration so it models easily
• Plucked strings vibrate autonomously after excitation, simplifying their modelling.
• Pointing gestures can localize sound sensations where words fall short
• Population coding across many neurons achieves precise and robust auditory processing
• Post-noise tinnitus and distortion signal auditory overload and hearing-loss risk
• Pressure differences across the eardrum attenuate low-frequency sound and cause pain
• Product sound quality depends on task context, not just pleasantness.
• Properly sampling continuous signals yields discrete and digital signals
• Prosody denotes speech features not tied to segmental structure
• Psychoacoustic models explain experimental results without addressing physiology
• Psychoacoustic tonality measures tone-character, distinct from musical tonality terms
• Psychoacoustics developed largely through communication technology challenges
• Psychoacoustics studies higher auditory functions where physiology is unknown
• Psychometric functions map continuous stimulus scales onto probabilistic detection judgements
• Psychophysical functions map continuous physical sound scales onto varied perceptual scales.
• Psychophysical scales are built by adjusting sounds to match reference ratios
• Psychophysical scales are built from repeated ratio-adjustment judgments plus an anchor point
• Pure-tone audiometry measures hearing thresholds at six standard frequencies
• Pythagoras established the inverse relation between pitch and string length
• Quadraphonic sound failed due to delivery problems and too few speakers.
• Quality and impairment are measured with distinct MOS variants.
• Quantization noise drops by 6 dB per added bit while hearing spans 130 dB
• Raw audio recordings lack semantic searchable metadata about musical content
• Ray-tracing and image-source methods fail on low-frequency diffraction at corners
• Real loudspeakers show flat magnitude response only below 800 Hz with a midrange dip
• Real-time auditory models enable speech recognition, sound reproduction, and hearing aids
• Real-world signals change over time, so fixed filters need adaptation
• Real-world signals vary over time so filters must adapt to conditions.
• Recommendable reverberation times range from 2 seconds down to 0.35 seconds by room type
• Recorded audio is engineered for plausibility, not faithful reproduction of reality
• Recovery from temporary hearing shift takes roughly double the exposure time
• Recruitment makes loudness grow faster in impaired ears than normal ears.
• Regularization trades inversion accuracy against loudspeaker dynamic-range demands
• Relative loudness change provides the key auditory cue for source distance
• Relative magnitude judgments can be anchored into absolute psychophysical scales
• Relative motion between a sound source and listener shifts perceived pitch
• Repertory grid technique lets each subject generate their own descriptive dimensions via triadic comparisons.
• Required audio fidelity scales with the application’s communicative demands
• Research advances only through sustained iteration across abstraction levels
• Resonance can amplify desired sound or cause harmful machine vibration
• Resonance can be engineered to strengthen sound or damped to prevent malfunction
• Resonance frequencies produce longer build-up and decay than off-resonance
• Resonators color spectral and temporal content of transmitted sound.
• Reverberant rooms favor directional loudspeakers to control sound colouration
• Reverberation destroys the phase information that could encode meaning
• Reverberation low-pass filters speech modulation while noise raises its floor
• Reverberation time and signal-to-noise ratio map onto the same speech intelligibility scale
• Reverberation time is the key design variable separating concert hall types
• Reverberation time maps directly onto equivalent signal-to-noise ratio
• Reverberator pre-delay controls perceived source distance independently of loudness ratio
• Reversing a signal’s polarity changes its perceived bass level
• Reversing a single harmonic’s phase makes that frequency sound much louder.
• Reversing the phase of a single harmonic can drastically change its perceived loudness
• Rhythm derives from natural body repetitions like heartbeat and walking.
• Rigid-wall tube terminations fully reflect waves while open ends reflect with phase inversion.
• Ringing ears indicate damage even below the 85 dB threshold
• Robust spatial hearing despite reverberation has driven decades of binaural modeling research.
• Room acoustics can be simulated by statistically tracing sound rays until their energy decays
• Room acoustics contaminate loudspeaker listening tests needing controlled environments
• Room acoustics smear directional perception and mask recorded reverberation
• Room impulse response measurements cannot be made with true perfect impulses from practical sources.
• Room intelligibility can be predicted from impulse response and noise levels
• Room reflections and reverberation shape whether voice communication succeeds or fails
• Room reflections can be modelled as sources in mirrored rooms
• Room reflections shape how well speech communicates across a space.
• Room reverberation erases the audibility of phase changes in signals.
• Room reverberation improves rather than impairs distance judgment
• Room reverberation suppresses the comb-filter colouring of amplitude panning
• Room sound decay is exponential, appearing linear on a dB scale
• Roughness between note partials determines perceived musical dissonance
• Roughness from harmonics requires fundamental frequencies below 100 Hz
• Rule-based speech synthesis survives only where computational power is scarce.
• Sampling below twice the highest frequency causes aliasing distortion.
• Scene analysis principles apply to cross-modal audiovisual speech mixtures.
• Sensation level expresses tone level relative to one individual’s own threshold
• Sensory efficiency has driven animal evolution and survival
• Sensory scales can be calibrated for equal intensity across different attributes
• Sensory science quantifies perceptual attributes through structured descriptive analysis
• Separate sensory modalities fuse into one coherent perception unless cues conflict
• Shaping quantization noise to sit 13 dB below the signal renders it inaudible
• Sharp caustic edges demand frequencies far above human hearing range
• Sharpness, roughness, and narrowband temporal components increase noise annoyance
• Shifting harmonic partials upward raises perceived pitch despite unchanged envelope modulation
• Short sounds decay in perceived loudness faster than long sounds.
• Short-time Fourier analysis converts signals into spectrograms via windowed transforms
• Shorter synthesis units lower concatenation cost but syllables are the optimal balance
• Shoulder reflection delays versus direct sound vary inversely with source elevation
• Shoulder reflections create a distinguishable 1-ms delay in head-related impulse responses
• Shoulder reflections imprint identifiable arcs on measured head-related impulse responses.
• Signal cropping is equivalent to applying a rectangular window
• Signal processing can compensate for loudspeaker and headphone imperfections.
• Signal processing compensates reproduction deviations through equalization
• Similar-looking phenomena across domains still need domain-specific scientific concepts.
• Simple integer frequency ratios between notes produce the greatest measured consonance
• Simple integer frequency ratios minimize partial interference and thus dissonance
• Simple integer frequency ratios produce local consonance maxima between tones
• Simple stimuli fail to represent how hearing processes real complex sounds
• Simultaneous cochlear neuron firing across frequencies produces buzzy timbre
• Singing and speech differ systematically in pitch control, spectrum, vibrato, and range
• Single auditory nerve fibres encode only a narrow 20-40 dB dynamic range.
• Single auditory neurons encode only a narrow 20-40 dB dynamic range
• Single loud impulses can instantly tear the eardrum or break ossicles
• Single neuron auditory responses are mapped through controlled stimulus repetition
• Single-channel noise reduction exploits speech’s temporal, spectral, and statistical properties
• Single-ear listeners can still localize sound sources but with degraded accuracy
• Single-neuron auditory responses are mapped invasively in animals only
• Slight response deviations go unnoticed without an audible reference
• Sloping or rising hearing loss can create contradictory pitch codes across frequency regions
• Slow sound-level changes are heard as fluctuation and fast ones as roughness
• Solids support shear forces and thus transversal waves, unlike fluids.
• Sonic interaction design studies sound perception in active embodied contexts rather than passive listening
• Sopranos retune formants to match harmonics to boost loudness, not intelligibility
• Sound as physical event and sound as auditory perception are distinct
• Sound attributes are independent variables manipulated to elicit dependent responses.
• Sound audibility depends on both its level and its frequency
• Sound cannot persist in time because it leaves no physical trace
• Sound colouration emerges from the combined loudspeaker and room properties
• Sound denotes both a physical wave and the subjective percept it evokes
• Sound design in interfaces is an extension of product sound quality
• Sound diffraction arises when a wavefront’s edge acts as a secondary source
• Sound events arranged in time and frequency are perceptually organized into streams.
• Sound field control can deliver different audio to different zones of a room.
• Sound from a source falls 6 dB per distance doubling until reverberation takes over
• Sound impulsiveness resists simple psychoacoustic measurement
• Sound intensity is a vector describing the directional flow of sound energy.
• Sound is a mechanical wave requiring a medium and a vibrating source
• Sound level meters trade average level against tracking speed via time response.
• Sound localization blur is smallest in front and largest at the sides
• Sound localization evolved in nocturnal mammals and remains vital for prey and predator.
• Sound localization fuses independent sensory evidence into one coherent percept
• Sound perception is shaped by the listener’s expectations and context of use
• Sound pressure is the small pressure deviation caused by a sound wave
• Sound pressure is tiny pressure deviation, and human hearing spans a specific Pa range
• Sound propagation can be approximated statistically by tracing discrete rays
• Sound quality cannot be assessed except through psychophysical experiments with human subjects
• Sound quality is fundamentally subjective and requires perceptual testing to characterize
• Sound quality judgments depend on listener context and expectations
• Sound reinforcement permits only milliseconds of delay while recorded reproduction tolerates seconds
• Sound reproduction’s endpoint is the human listener, not fidelity per se
• Sound reproduction’s unifying trait is its endpoint in the human listener
• Sound source location is encoded in waveform distortions between the ears
• Sound storage digitized fully but human ears force analog endpoints
• Sound technology work needs acoustics, perception, and signal processing jointly.
• Sound width perception depends on signal type, not physical source size
• Sound within one critical band is perceived as a single source
• Sound’s physical force is negligible yet its informational effects matter greatly.
• Sounds mask both preceding and following sounds but forward masking dominates.
• Source directivity is usually omitted from virtual audio for computational reasons
• Spaced microphones sound warmer because incoherent signals spread interaural cues
• Spaced mics sound warm and ambient because incoherent signals exceed loudspeaker angle.
• Spatial audio is synthesized by controlling band energies and inter-channel interdependencies.
• Spatial audio synthesis quality is measurable against a real-source reference.
• Spatial hearing is calibrated through learning to each listener’s head and pinna geometry
• Spatial hearing is described in three distinct coordinate systems with different strengths.
• Spatial hearing is worst near 1.7 kHz where head wavelength matches.
• Spatially broad sound sources should alter their perceived extent with listener distance
• Speakers shift glottal closure timing to control the spectral brightness of their voice.
• Speakers use pitch variation to signal statements versus questions
• Speakers vary pitch to signal emotion, questions, and focus.
• Spectral band replication transmits only lows and predicts highs from them
• Spectral elevation cues predict human sound-localization performance via machine learning
• Spectral masking by loud sounds is exploited to design lossy audio codecs.
• Spectral-cue adaptation in hearing requires visible sound sources as a reference
• Spectrograms require intensity maps or waterfall plots because they are three-dimensional
• Speech audiometry tests intelligibility by varying signal level and speech pace
• Speech coding exploits production models for low bit rates without databases
• Speech coding is fundamentally a quality-versus-bitrate optimization tradeoff.
• Speech coding is the most mature speech technology because it needs no databases.
• Speech elements near masked threshold carry intelligibility only under binaural listening.
• Speech intelligibility alone cannot capture loss of speaker identity in codecs
• Speech intelligibility depends on conserving envelope modulation through a channel
• Speech intelligibility estimates compress 98 MTF values into one index
• Speech intelligibility harms open-plan office quality, unlike in auditoria
• Speech intelligibility in public spaces is a regulated engineering requirement
• Speech intelligibility is the standard criterion for hearing function
• Speech intelligibility measures the proportion of correctly identified transmitted messages
• Speech intelligibility metrics can be predicted from impulse response and noise
• Speech intelligibility tolerates group delay changes up to about 100 ms
• Speech is a continuous flow that still requires discrete units to be understood
• Speech is a continuous signal flow, not discrete concatenated units
• Speech is the original form of linguistic communication, predating writing.
• Speech is uniquely human among species’ communication abilities.
• Speech processing can fail independently of general hearing ability
• Speech processing deficits can impair understanding independent of hearing
• Speech production can be modeled mathematically to derive useful signal processing algorithms
• Speech production maps onto a three-stage cascaded electrical circuit analogy
• Speech production reduces to a source-filter model with glottal or noise sources
• Speech production reduces to lungs as compressor driving a glottal oscillator through resonating tracts
• Speech quality can be predicted algorithmically using signals that mimic human auditory processing.
• Speech recognition picks the model maximizing the probability of observed features
• Speech recognition shifted from time warping and shallow nets to HMMs then deep networks
• Speech redundancy masks up to 30 percent consonant loss in auditoriums
• Speech remains understandable at high levels because timing carries formant info
• Speech segments are classified by abstract classes, concrete sounds, and transition units
• Speech sound generation was mechanically modeled as early as the 18th century
• Speech synthesis selects units by minimizing target and concatenation costs.
• Speech tracts lack coupling that brass instruments synchronize with.
• Speech transmission quality hinges on undistorted information transfer
• Speech-to-speech translation is limited by language processing, not signal processing
• Spherical harmonics act as basis functions for building any microphone pattern
• Spot miking captures each instrument with its own near microphone for independent channels
• Squaring the pressure signal makes noise and reverberation measurable on a common scale
• Standard Ambisonics microphones use four capsules to capture omnidirectional and three-dipole signals.
• Standard audio sampling rates range up from 8kHz speech to 192kHz ultra-high fidelity
• Standardized concert hall acoustics metrics have known technical limitations compared to human perception
• Standing waves arise from counter-propagating reflections with no net energy flow
• Static loudness models fail on time-variant sounds, motivating frame-based temporal integration models
• Stereo loudspeakers create directional illusion through summed phase shifts mimicking time delays
• Stereo signals can be processed for headphones to recreate natural binaural sound fields
• Stereo upmixing began with sum-and-difference signal processing decades ago
• Stereo-to-binaural upmixing splits direct and ambience signals for separate spatial processing
• Stereophonic two-channel reproduction became widespread after the two-channel record format arrived.
• STI reliability is judged against its 0.15 category step size
• STI simulation tools persist because fixing underperforming PA systems is costly
• Stiffness makes transversal waves dispersive and their modal tones inharmonic
• String length ratios explain musical intervals since Pythagoras
• Stronger tones in the cochlea actively mask neighboring weaker tones via two-tone suppression.
• Studying acoustic communication requires both basic understanding and engineering application goals.
• Subjective listening tests map manipulated acoustic variables to measured responses.
• Substantial hearing loss reduces ability to react to sonic warnings
• Subtractive microphone designs produce cardioid and dipole patterns via port spacing and delay
• Summing a signal with its delayed copy creates perceived pitch.
• Surround recordings pair spaced mics one-to-one with each loudspeaker
• Synchronized phase harmonics produce a ‘buzzy’ sharpness perceptible without room reflections
• Synchronized visual motion captures perceived sound direction within 30 degrees
• Synthesis and recognition demand language knowledge databases that coding does not.
• Synthetic speech dates to late-18th-century acoustic-mechanical machines
• Systems are analyzed through complementary structure and function concepts
• Tactile frequency discrimination is crude, often off by an octave
• Tactile vibration accompanies most real sound and fuses with it unconsciously
• Tactile vibration raises perceived loudness and lowers preferred bass levels
• Tactile-audio desynchrony is as critical as audiovisual desynchrony
• Telecom device sound quality depends on network, room, and feedback conditions jointly
• Temporal change is a defining feature of natural sound perception
• Temporal masking makes silence following a sound seem shorter
• Test method means the logic selecting sound attributes across successive tasks
• The 0 dB HL audiometric zero varies across frequency by design
• The 1930s vocoder pioneered reducing speech to slowly-varying parameters to cut bandwidth
• The 1948 Jeffress model explains ITD coding via neural delay-line coincidence detectors
• The absorption coefficient measures the fraction of incident sound energy absorbed.
• The auditory filter width is what matters for sound and voice technology design.
• The auditory periphery is non-linear and cannot be captured by one transfer function.
• The auditory system can be modeled as computing autocorrelation functions per frequency band
• The auditory system cancels room coloration to recover source signals
• The auditory system compensates for room acoustics to recover the original source signal
• The auditory system continually relearns to decode pinna-imposed spectral cues.
• The auditory system estimates source signals by ignoring the room’s spectral contribution
• The auditory system highlights spectral changes to help listeners compare sounds
• The auditory system largely ignores signal phase, driving spectral analysis conventions
• The Bark scale is linear below 500 Hz and logarithmic above it.
• The Blumlein XY pair captures reverberance fully but causes phase artefacts from side sound.
• The brain measures interaural delay via coincidence-detector neurons and delay-line axons
• The brain organizes simultaneous sound events into separate auditory streams via auditory scene analysis.
• The brain pools many noisy auditory fibers to form clean perception
• The cochlea actively amplifies sound and emits echoes that reveal inner-ear health
• The cochlea splits sound into narrowband signals that the brain processes individually
• The cochlea’s sharp frequency selectivity needs an unexplained second filter.
• The cochlear amplifier requires nonlinear, level-dependent elements to model.
• The cochlear nucleus functions like a preamplifier and initial spectral analyzer
• The condenser microphone and its electret variant dominate modern microphone technology
• The ear achieves spatial hearing through signal analysis rather than directional receptors
• The ear canal acts as a quarter-wavelength resonator amplifying 3-4 kHz by 10 dB
• The ear canal’s quarter-wavelength resonance boosts speech-critical frequencies around 3-4 kHz.
• The ear detects binaural delays finer than individual neuron timing supports
• The ear emits measurable echoes requiring active neural involvement
• The ear lacks the eye’s directional receptors, relying on canal analysis
• The ear performs three sequential conversions from air waves to neural signals
• The ear sits 8 cm off-center, so source-ear distance changes with azimuth
• The ear’s 1-2 ms time resolution makes monaural temporal cues unusable for localization.
• The ear’s 120-dB and 20 Hz-20 kHz working range exceeds man-made devices
• The ear’s analysis resembles Fourier frequency decomposition of sound
• The ear’s time-frequency resolution cannot be modeled by fixed-length frames
• The Eustachian tube pressure-balances the middle ear against altitude changes
• The eye has directional sensitivity that the cochlea entirely lacks
• The first objective speech quality measure failed on mobile-network timing variations
• The Fourier transform measures how a signal matches each sinusoid pair
• The Fourier transform trades costly time-domain convolution for simple frequency-domain multiplication
• The glottal waveform can be estimated indirectly through inverse modeling of microphone recordings
• The glottal waveform’s closing phase is abrupt and carries most excitatory energy.
• The hardest part of speech-to-speech translation is language processing, not signal processing.
• The head shadows and reflects sound to create level differences between the ears.
• The hearing system groups frequency bands together when their level variation is similar
• The hearing threshold is measured from sounds reaching the subject’s binaural position.
• The Helmholtz resonator models confined air as a spring with an air-mass plug
• The human auditory system evolved specifically for spoken-language analysis
• The human auditory system is communication acoustics’ hardest link to model
• The human auditory system is the hardest link in the communication chain.
• The human ear is largely insensitive to signal phase
• The IACC maximum value estimates interaural coherence between the two ear signals
• The International Phonetic Alphabet attempts a universal notation for every speech sound
• The listening experience intended by engineers is what gets conveyed, not raw sound
• The mel scale was derived by asking listeners to double perceived pitch
• The method of limits is inherently prone to response bias
• The middle ear evolved as an impedance transformer to bridge air-water sound energy loss
• The neural mechanism extracting perceived distance from room response remains unknown
• The ORTF technique blends coincident and spaced microphone characteristics
• The outer ear’s resonant cavities enable elevation perception where other cues fail
• The peripheral hearing system acts as a quantitative spectral analyser
• The phon defines loudness level equal to dB SPL at 1 kHz
• The power spectrum is blind to the relative phase of a sound’s components
• The public’s subjective opinion is the final truth on sound quality
• The rate of beating between close partials determines whether they sound consonant or rough
• The similarity between artificial and biological neural networks is only superficial.
• The singer’s formant arises from a quarter-wave resonator in the larynx
• The speech tract is modeled as a transmission line with pressure and volume-velocity variables
• The superior colliculus is the only mammalian brain nucleus with a topographical auditory map
• The sweet spot for stereo imaging spans only a few tens of centimetres
• The technically precise term for frequency response is magnitude response
• The telephone band of 300-3400 Hz remains the standard compromise for speech intelligibility.
• The transmission channel changes from acoustic path to electrical system in device-mediated speech.
• The vacil unit anchors fluctuation strength to a 4 Hz modulated tone
• The vocal apparatus functions as a pressure source driving a resonant filter system
• The vocal tract can be simulated as a cascade of acoustic transmission line segments
• Theory is needed to break through deceptive appearances but cannot dictate action
• Three-dimensional VBAP extends pair panning to position sources on loudspeaker triangles
• Thresholds and auditory scales link physical sound attributes to perceived auditory events
• Thresholds are measured by presenting many stimulus levels and fitting a psychometric function
• Thresholds can be measured by tracing stimulus change until perception flips
• Throat singers create melodies by shaping vocal-tract resonances
• Timbre distinguishes sounds sharing pitch, loudness, and duration.
• Time-frequency modulation and amplitude modulation are distinct operations
• Time-frequency transforms make signal processing and interpretation easier
• Time-varying sound loudness is set by integrated loudness peaks, not moment-to-moment level
• Time–frequency audio processing exploits the ear’s own spectral analysis structure
• Tinnitus can arise from unstable outer hair cell oscillation in the cochlea
• Tinnitus can originate from blood pressure changes in the ear
• Tinnitus is a phantom auditory event with multiple causes
• Tinnitus is treated by masking or habituation, not cured
• Tone-on-tone masking can be detected through beating and combination tones alone
• Tone-to-noise ratio fails when multiple tones share a critical band
• Tonotopic frequency ordering persists from cochlea to auditory cortex
• Total harmonic distortion measures all components a fundamental generates, including non-harmonic ones.
• Tracking audiometry maps hearing thresholds but ignores response bias
• Trained listener panels trade real-world representativeness for analytic test resolution
• Trained listeners improve test sensitivity but misrepresent general audience opinion
• Trained panels can bypass consensus vocabulary because perception varies less than description
• Trapped compressible air and moving air form a mass-spring acoustic resonator
• Triadic comparison tests can derive individual perceptual structures of sound
• Tube termination conditions determine which harmonics a wind instrument produces
• Two brain models differ on whether localization uses narrow or wide receptive fields
• Two competing loudness models predict time-variant loudness by peak tracking or frame averaging
• Two identical sounds within 30–40 ms are fused into one located auditory event.
• Two loudspeakers can synthesize a credible 5-channel surround illusion
• Two tones presented together generate an audible combination (difference) tone not physically present.
• Two-tone suppression lets strong signal components override weaker nearby ones
• Typical one-metre speech at 60-70 dB is optimally tuned for auditory communication
• Uncontrolled loudness differences confound audio quality listening-test results
• Uncorrelated noise from surround speakers creates enveloping source perception.
• Uniform masking noise must drop 3 dB per octave above 500 Hz.
• Unit-selection synthesis achieves naturalness at the cost of huge databases and inflexibility
• Unlike the eye, the ear has no primary directional sensitivity
• Untreated hearing loss in children was once misdiagnosed as intellectual disability
• Unvoiced speech arises from turbulence at constrictions or sudden pressure releases
• Valid psychoacoustic listening tests require blind conditions and carefully worded tasks.
• VBAP predicts perceived source azimuth accurately but not personal elevation
• Verbal elicitation dominates sensory science via consensus or individual vocabularies
• Very general theories guide thinking but resist direct application to specific cases
• Vibration generates acoustic waves while sound radiation drains vibration energy
• Virtual audio engines render sounds based on avatar position, not just a fixed mix.
• Virtual sound source distance is limited by reproduction technology, not perception
• Virtual source elevation perception is personal but azimuth is predictable
• Vision dominates audition except where visual events are absent or ambushed
• Vision dominates hearing in localization, identification, and size judgment
• Visual color affects perceived loudness of the same sound
• Vocal tract acoustics reduce to a cross-sectional area function along one axis
• Vocal tract length differs by sex and its resonances form speech formants.
• Vocal tract transmission depends solely on the cross-sectional area function.
• Voice has become a natural interface for human-machine communication
• Voice-based computer interfaces required granting machines language processing and world knowledge
• Voiced and unvoiced speech differ in whether source is glottal oscillation or frication noise
• Voiced speech originates from airflow through the vibrating vocal folds across the glottis
• VoIP call delays exceed those of regular mobile calls
• Vowels keep the vocal tract open and vary little by context, unlike consonants
• Wave field synthesis is nearly always generated synthetically, not from real recordings.
• Wave field synthesis needs loudspeakers spaced at most half a wavelength apart.
• Wavelet and constant-Q transforms trade time and frequency resolution non-uniformly
• Wavelet and constant-Q transforms use non-uniform time-frequency resolution unlike short-time Fourier analysis
• Waves split into reflected and refracted parts at impedance-mismatched interfaces
• Western major scales consist of five whole tones and two semitones in fixed arrangement
• Western orchestral tuning pitch has risen historically for a brighter sound
• Whether spectral adaptation dynamics are purely peripheral remains experimentally unresolved
• WHO defines disabling hearing loss as a 40 dB threshold shift for adults
• Wideband speech coding spans 50 Hz to 7000 Hz for brighter fuller quality