Auditory spectra smooth harmonic detail while preserving formant structure

Evidence

The linear power spectra and auditory spectra for a pure tone, white noise, and two speech signals are shown in Figure 13.1. Overall, the auditory spectrum is seen to differ from the power spectrum due to the use of the ERB scale, the asymmetric spreading of the excitation, and the emulation of the hearing threshold. The pure tone case demonstrates how the excitation spreads to adjacent frequency bands. The white noise case, in turn, visualizes the combined effect of the frequency warping and the emulation of the frequency-dependent sensitivity of hearing on the spectrum. That is, the level of the auditory spectrum increases with frequency, and the boost around 4 kHz resulting from the ear canal resonance is visible as well. In addition, the auditory spectrum differs from the physical one in terms of frequency resolution. The auditory spectra of the speech signals illustrate how the harmonic fine structure in voiced phonemes is visible in the physical spectra but is smoothed in the auditory spectra due to frequency warping. Additionally, the warping increases the visibility of low frequencies in the spectra. However, the formants in the speech signals are clearly also present in the auditory spectra. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 1589

Categorization

• XREF: Connects to the broader soundscape/noise chapters read earlier: frequency warping here parallels how the ear’s nonlinear sensitivity shapes what signals matter, echoing how lo-fi environments erase distinctions.