Sparse codes for speech predict spectrotemporal receptive fields in the inferior colliculus.

We have developed a sparse mathematical representation of speech that minimizes the number of active model neurons needed to represent typical speech sounds. The model learns several well-known acoustic features of speech such as harmonic stacks, formants, onsets and terminations, but we also find more exotic structures in the spectrogram representation of sound such as localized checkerboard patterns and frequency-modulated excitatory subregions flanked by suppressive sidebands. Moreover, several of these novel features resemble neuronal receptive fields reported in the Inferior Colliculus (IC), as well as auditory thalamus and cortex, and our model neurons exhibit the same tradeoff in spectrotemporal resolution as has been observed in IC.

Efficient coding in human auditory perception.

Natural sounds possess characteristic statistical regularities. Recent research suggests that mammalian auditory processing maximizes information about these regularities in its internal representation while minimizing encoding cost [Smith, E. C.