Rick M. Roark, PhD, Steven D. Schaefer, MD, Ben C. Watson, PhD

Using a wavelet transform, time-frequency representations of electromyography (EMG) signals of the thyroarytenoid (TA) muscle were obtained for 10 normal controls and 10 subjects with spasmodic dysphonia (SD). The TA signals were recorded during the performance of five tasks that ranged from vegetative to complex linguistic. In previous research, statistical pattern recognition algorithms were applied to normalized mean differences of 24 time-frequency measures to classify subjects of the database into two groups for each task. Results of more detailed investigations of the time-frequency differences of TA EMG signals between study populations during quiet breathing are reported here. Subjects with SD are noted to have higher mean frequency content of distinguishing time-frequency features, but exhibit less variability in these statistical measures than the normal control population. This group of SD subjects appeared to breathe in more stereotypic fashion than the normal control group, thus permitting more successful classification of subjects during quiet breathing than Valsalva's maneuver or whispered /i/tasks. Additionally, comparisons of time-amplitude and time- frequency representations are made during one of two occurrences of vocal spasm noted during quiet breathing. Results indicate dynamic changes in frequency content of the TA EMG signal prior to and coincident with observed spasmodic amplitude behavior.