Separation of speech from interfering sounds based on oscillatory correlation

A multistage neural model is proposed for an auditory scene analysis task-s egregating speech from interfering sound sources. The core of the model is a two-layer oscillator network that performs stream segregation on the basi s of oscillatory correlation. In the oscillatory correlation framework, a s tream is represented by a population of synchronized relaxation oscillators , each of which corresponds to an auditory feature, and different streams a re represented by desynchronized oscillator populations. Lateral connection s between oscillators encode harmonicity, and proximity in frequency and ti me. Prior to the oscillator network are a model of the auditory periphery a nd a stage in which mid-level auditory representations are formed. The mode l has been systematically evaluated using: a corpus of voiced speech mixed with interfering sounds, and produces improvements in terms of signal-to-no ise ratio for every mixture. The performance of our model is compared with other studies on computational auditory scene analysis. A number of issues including: biological plausibility and real-time implementation are also di scussed.