This thesis deals with intra-speaker correlation analyses of speech sounds, and the possible utilization of this correlation to speech recognition. Current approaches to phonetic classification, regardless of whether they use context-dependent or -independent models, achieve classification based on locally optimum criteria.
