Human knee abnormality detection from imbalanced sEMG data

The classification of imbalanced datasets, especially in medicine, is a major problem data mining. Such evident analyzing normal and abnormal subjects about knee from collected during walking. In this work, surface electromyography (sEMG) were walking the lower limb 22 individuals (11 with 11 without abnormality). Subjects abnormality take longer to complete task than healthy subjects. Therefore, SEMG signal length unhealthy that subjects, resulting imbalance sEMG data. Thus, development model for such datasets challenging due bias towards majority class signals are contribution multiple motor units at time their dependency on neuromuscular activity, physiological anatomical properties involved muscles. Hence, automated analysis an arduous task. A multi-step scheme proposed research overcome limitation. wavelet denoising (WD) used denoise signals, followed by extraction eleven time-domain features. oversampling techniques then balance under increasing training minority class. competency was assessed using various computational classifiers 10 fold cross-validation. It found improve performance all studied when applied