Voice Identification Techniques

The problem of designing a machine which can identify automatically a speaker by processing samples of his speech has been investigated. Through construction of special purpose equipment and digital computer simulation, the fundamental operations of speech clue extraction and voice recognition have been implemented in a manner designed to render voice identification by a machine insensitive to the effects of noise and speech bandwidth truncation. Results of experiments indicate that the extraction of certain spectral, voice fundamental and speaking cadence characteristics of speech can lead to reliable identification of a speaker's voice, even in undesirable speaking, and speech transmission environments.