Using Windowed Relative Deviation to Detect Possible Voice Pathology

A diagnostic method is presented that provides for analyzing pitch "jitter" in running speech. "Jitter" is typically measured with explicit voice tasks, namely sustained vowel phonation. However, some voice pathologies cannot be detected with sustained phonation. Further, it is not possible to ensure explicit voice productions from certain patients, including pediatric populations. In contrast, windowed relative deviation reports instantaneous pitch "jitter" as well as the overall "jitter" statistic commonly reported. Also, the width of the analysis window is related to the rate of pitch deviation, which provides a unique form of selectivity. Voice productions from a normal adult speaker and from an adult speaker with a known voice pathology were analyzed with this method. Voice productions from the normal speaker exhibited less than 1% pitch deviation during phonetic portions of the signal that were akin to sustained phonation. On the other hand, the speaker with a known pathology exhibited greater than 10% pitch deviation at quasi-periodic intervals within sustained phonation