Abstract
Analysis-by-synthesis procedures for automatic reduction of speech spectra describe the goodness of fit between an input-speech spectrum and an internally generated spectrum by an error score. In vowel analysis, the error score typically is derived from a computed-error curve specified by values corresponding to 24 input filters. Values of error scores from a set of utterances by three talkers are presented and compared to similar scores resulting from manual spectrum-matching procedures. Variations in error score attributable to particular vowels, contexts, and talkers are investigated, and the error-score magnitude is related to accuracy in formant location. [This work was supported in part by the U. S. Army, the U. S. Air Force Office of Scientific Research, and the U. S. Office of Naval Research; in part by the U. S. Air Force (Electronic Systems Division) under Contract AF 19(604)-6102; in part by the National Science Foundation (Grant G-16526); and in part by the National Institutes of Health (Grant MH-04737-02).]
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
