Using Fuzzy Sets to Model Paralinguistic Content in Speech as a Generic Solution for Current Problems in Speech Recognition and Speech Synthesis

Current problems in speech processing exist due to infinite variations of speech utterances. No two speech utterances are exactly alike, even if they are linguistically the same word. The difference is therefore, due to the paralinguistic content of the speech utterances. This leads to the conceptualization of the paralinguistic content of speech as arising from infinite variation. Infinite variation in paralinguistic content has been modeled using the interval [0, 1], the basis of fuzzy theory. Variability, that is, the ability to vary, has been identified as the property of natural systems, due to which infinite variation is possible. Thus, variability as a concept has been mapped to paralinguistic content. Further, each component of paralinguistic content has been mapped to a group of membership functions of fuzzy sets.