Automatic Measurement and Classification of Pressure Peaks When Estimating Subglottal Pressure by Oral Pressure During p-Occlusions

The subglottal pressure is an important parameter in voice production. It is strongly correlated with vocal loudness and fundamental frequency, but also with several voice disorders. Subglottal pressure is rarely measured directly but rather estimated by measuring the oral pressure during p-occlusions. The purpose of this study is to describe and evaluate a newly developed method for automatic measurement and classification of pressure peaks used to estimate subglottal pressure by oral pressure, in order to simplify measurement of subglottal pressure in research and in clinical settings. The method detects peaks in oral pressure signals and measures their height and estimates their quality by a score. Automatically obtained results were compared with results from four experienced researchers' manual measurements and quality ratings of 120 oral pressure peaks. The method is shown to produce measurements of peak height very similar to the average of the manual measurements performed by the researchers, Pearson correlation r > 0.99. The average of Pearson correlation coefficients between the automatic peak quality score and the rated peak quality was slightly lower than the average of the Pearson correlations between raters (r = 0.52 vs r = 0.67). The method seems to produce nearly identical measurements and similar quality ratings to those of human raters and thus has the potential of replacing manual measurements of oral pressure peaks. The method can greatly improve measurement speed and has the benefit of producing operator-independent results.