A stochastic model of fundamental period perturbation and its application to perception of pathological voice quality

Voice Pathology Detection Based eon Short-Term Jitter Estimations in Running Speech

“…One such property is that jitter in adjacent periods is correlated and thus, present time jitter could be predictable from past values [32] . In Endo and Kasuya [8] and in Schoentgen and de Guchteneere [32] , the jitter time series is modeled as an autoregressive process. Following that, it is shown that the frequency and bandwidth of the pole of the envelope is related to the rate of pathology perceived in the examined signal [8] .…”

“…In Endo and Kasuya [8] and in Schoentgen and de Guchteneere [32] , the jitter time series is modeled as an autoregressive process. Following that, it is shown that the frequency and bandwidth of the pole of the envelope is related to the rate of pathology perceived in the examined signal [8] . Therefore, it is straightforward to apply similar time series modeling techniques to the short-term jitter sequence estimated by the SJE.…”

“…Absolute jitter is a widely implemented 8) where N is the total number of pitch periods and u ( n ) is the pitch period sequence. This type of jitter estimation is implemented by two of the most established systems for acoustic voice quality assessment, the Praat [18] system and the MDVP [19] .…”

“…Also, averaging implies that pitch cycle perturbations are generated by an independent and identically distributed (gaussian) stochastic process. However, it has been shown that there is a correlation between successive values of jitter [8] . Therefore, this correlation should be removed before applying the average operator [2] .…”

Voice quality transform using jitter synthesis

“…Titze, Deshmukh 등 이 음성의 준주기성 검출에 관하여 연구하고 보고한 바 있다 (Titze, 1995;Deshmukh et al, 2005). 음성합성에의 적용 사례로 는 Hillenbrand가 합성음에서의 주기성의 변이에 대해 보고한 바 가 있고 (Hillenbrand, 1987) Endo & Kasuya 등이 장애음성의 합성 에 적용한 사례가 있다 (Endo & Kasuya, 1996) (Kreiman & Gerratt, 2003), 두 번째는 하모닉 상태의 음원을 생성 한 뒤 특정 전달함수를 통과시켜 F 0 의 변이를 생성하는 방법 (Endo & Kasuya, 1996) 그리고 세 번째로 특정 F 0 을 기준으로 난 수발생기에 따른 F 0 의 변이를 생성하여 구현하는 방법이 있다 (Alzamendi et al, 2013;Ruinskiy & Lavner, 2008…”