Efficient audio coding discards predictable signal components and encodes only novelty
Evidence
To maximize the information content of the signal, we should aim to remove the predictable parts, because they can be recalculated at the receiving end, and only encode the unpredictable part, because this represents the new information. — David M. Howard, Jamie Angus, Acoustics and Psychoacoustics, 2602
Categorization
- DEFINE: Clarifies the core rationale behind correlation removal in both lossless and lossy coding.
- XREF: Parallels information-theoretic ideas (Shannon entropy as unpredictability) and cognitive novelty-seeking; connects to how auditory perception privileges change over steady state.
Related
- Lossless audio coding works by predicting and stripping signal correlations (similarity 0.87)
- Perceptual coding works by maximizing the surprise value of coded signals (similarity 0.86)
- Audio compression maximizes information by encoding only unpredictable signal parts (similarity 0.86)
- Coded audio carries maximum information when the signal is maximally surprising (similarity 0.84)
- Lossy coding exploits psychoacoustic models to make removed information inaudible (similarity 0.83)
- Lossy audio codecs improve encoder-side while keeping decoder standards fixed (similarity 0.83)
- Perceptual coders exploit masking to cut audio storage and bandwidth (similarity 0.82)
- Without pitch, loudness, and timbre summaries, raw partial data would overwhelm consciousness (similarity 0.82)
- Instrument filtering adds predictability that audio coders exploit as redundancy. (similarity 0.82)
- Audio symbols with high probability carry less information and need fewer bits (similarity 0.82)
- Audio compression exploits masking to discard inaudible signal parts (similarity 0.81)
- Psychoacoustic models govern which audio information can be discarded unnoticed (similarity 0.81)
- Entropy encoding maximizes information carried by each bit pattern (similarity 0.81)
- Audio coding aims to reconstruct signals while other processing modifies them (similarity 0.80)
- Beyond audibility, spectral shape of loss drives which speech cues fail (similarity 0.80)