We present a new algorithm for the precise measurement of the duration of consecutive glottis cycles with a view to the analysis of jitter. Conventionally, jitter features take on the shape of the dispersion measurement in the durations of consecutive periods in an analysis interval of typically one second. Strictly speaking, this scheme is only valid provided that consecutive period values are purely random; otherwise, second-order statistics are required to complement dispersion measurements . The purpose of the analysis procedure presented here is i) to check the randomness of the glottis cycle aperiodicities and ii) to put jitter analysis back into a framework of time series analysis. The algorithm which we propose works both on laryngograms and acoustic signals. It achieves the required degree of precision by oversampling with on-line testing of the accuracy of the waveform reconstruction. It outputs several jitter measurements and displays the time series of the period values and the statistical distribution of the deviations of the instantaneous cycle durations from a running mean. Preliminary results obtained on healthy subjects are given.
Cite as: Schoentgen, J., Guchteneere, R.d. (1990) An algorithm for the measurement of jitter. Proc. ESCA Workshop on Speaker Characterization in Speech Technology, 175-180
@inproceedings{schoentgen90_scst,
author={Jean Schoentgen and Raoul de Guchteneere},
title={{An algorithm for the measurement of jitter}},
year=1990,
booktitle={Proc. ESCA Workshop on Speaker Characterization in Speech Technology},
pages={175--180}
}