Module separation 

Direct/ambience signal separation algorithm.

This module implements the coherence-based separation of direct sound from room ambience using inter-bin phase analysis.

Algorithm Overview

The separation is based on the observation that:

• Direct sound has coherent phase relationships between neighboring frequency bins (the phase progression follows a predictable pattern related to time-of-arrival)
• Ambience/reverb has random, diffuse phase relationships

We use multiple cues:

1. Inter-bin phase coherence: Direct sound shows correlated phase between neighbors
2. Spectral flux: Transients (direct attacks) have higher positive flux
3. Temporal stability: Ambience is more temporally stable
4. Magnitude correlation: Direct sound often shows correlated magnitude envelopes

Structs

CoherenceAnalyzer

Coherence analyzer for bin-to-bin relationships.

SeparationConfig

Configuration for signal separation.

SignalSeparator

Signal separator that applies the coherence analysis to split signals.

StereoSeparator

Stereo separation for maintaining spatial information.