Computational Auditory Scene Analysis

Definition

Computational implementation of Bregman's Auditory Scene Analysis principles for machine listening — algorithms that parse audio mixtures into constituent sources using acoustic grouping cues.

Key Ideas

• Implements ASA grouping principles computationally: harmonic grouping, onset synchrony, frequency proximity, common modulation.
• Typically operates on a time-frequency representation (cochleagram or spectrogram).
• Bottom-up: signal-driven grouping. Top-down: uses schemas/expectations.
• Pre-dates and partially inspired deep learning approaches — many CASA principles are implicitly learned by neural separation models.

Relationships

• Builds on ../concepts/auditory-scene-analysis (Bregman)
• See also ../concepts/common-fate-principle
• Contrast with ../entities/demucs-style deep separation — CASA is explicit, hand-designed; deep models learn grouping cues from data
• Some hybrid systems combine CASA front-ends with neural back-ends

Sources

None ingested yet — seed batch setup.