Automatic On-line One-channel Recognition of Masseter Activity

Abstract

The etiology of myoarthropathies of the masticatory system (MAP) is not fully understood. For the hypothetical association between the myogenous pain of MAP patients and masticatory muscle overuse to be proved, functional and parafunctional behavior of the masticatory muscles should be analyzed in normal and diseased subjects. The aim of this study was to test on-line the validity and reliability of an algorithm, applied to the electromyographic signal, to recognize various oral activities. The surface electromyogram of the masseter muscle was recorded in 12 subjects (seven females and five males, from 18 to 32 years old) who performed a series of functional and parafunctional activities (chewing soft food, chewing hard food, swallowing, laughing, speaking, and tooth grinding and clenching), as well as no activity. During the computer training phase, intra-individual classification functions of a multivariate discriminant analysis were calculated while each subject performed the described activities. During the test phase, each subject repeated the same activities, and the computer continuously classified them on-line. The percentage of correctly recognized activities was calculated for each activity and for each subject. No activity, chewing hard food, swallowing, laughing, grinding, and clenching were recognized correctly > 99% of the time. Chewing soft food was recognized correctly 97% and speaking 86% of the time. The sensitivity values for the recognition rates of the complete oral activities were, with one exception, > 0.82; the specificity values were > 0.95, and the K-values > 0.80. These results show that the algorithm had high sensitivity, specificity, and reliability in the classification of different oral activities under laboratory conditions.