Abstract
Temporal patterns of the fundamental frequencies of Japanese speech voice are measured, and thereby distributions, power spectrum and transitional probability of the fundamental frequency are calculated. Results induced are as follows; (1) as the mean value of the fundamental frequency become higher, the variance increases, and the variances in skilled speakers such as announcer are generally larger than those of normal speakers. ( 2) 98% of energy of pitch signal can be transmitted in 5 c. p. s. band width, provided gaps between pitch patterns are filled in smoothly. (3) simple relations exist between utterance speed and power spectrum. (4) differences of succeeding pitch signals (transitional probability) can be approximated to exponential distribution and negative difference is predominant than positive. Then the estimate of optimum characteristics of the filter in pitch signal transmission, considering the effect of linear distortion of the filter upon perception, and also the probability of the duration time of the singular fundamental frequency are discussed here.
