Modulation Spectrum in Speech Processing

摘要

The work questions adequacy of the short-term spectral envelope as the dominant carrier of the phonetic identity of a given speech instant and suggests temporal dynamics of components of the spectral envelopes as more reliable means for carrying the linguistic context of the speech message. It summarizes some research which employs processing of the so called modulation spectrum of speech. In argues that dominant linguistic information bearing components of modulation spectrum are found in the vicinity of 4 Hz and demonstrates that suppressing slow and fast components of trajectories of spectral envelopes can be useful in alleviating effects of acoustic environment in processing of speech.