Time Series Analysis of Jitter in Sustained Vowels

This paper proposes a new equation to show the time domain features of jitter: Jitter = (Ti+1-Ti)/Ti. The subjects are tonal language speakers. Jitter is extracted from EGG signals. The results show that: (1) jitter does not obey Gaussian distribution but multimodal distribution; (2) the first peak position near zero of the jitter distribution curve increases with the pitch; (3) each peak position is an integral multiple of the first peak position; (4) jitter values are around 0, and in most cases, adjacent values have opposite algebraic sign; (5) the zero value percentage and zero-crossing rate of jitter increase with the pitch; (6) the variance of pitch keeps in the allowance of just noticeable difference, which can be attributed to the effect of audible feedback, and it affects the variance of jitter further.