ADAPTIVE WRLS-VFF FOR SPEECH ANALYSIS

The purpose of this correspondence is to show that an ;adaptive weight ed recursive least squares algorithm, with a variable forgetting facto r (WRLS-VFF) will adjust the size of the data segment to be analyzed a ccording to its time-varying characteristics, as during the transition s between vowels and consonants. The algorithm ran accurately estimate : the vocal tract formants, anti-formants,and their bandwidths, be use d for glottal inverse filtering, perform voiced (V)/unvoiced (U)/silen t (S)classification of speech segments, estimate the input excitation (either white noise or periodic pulse trains), and estimate the instan t of glottal closure.