Feature extraction of knee joint sound for non-invasive diagnosis of articular pathology

The aim of this paper is to classify the vibroarthrographic (VAG) signals according to the pathological condition using the characteristic parameters extracted by the time-frequency transform, and to evaluate the classification accuracy. VAG and knee angle signals, recorded simultaneously during one flexion and one extension of the knee, were segmented and normalized at 0.5 Hz by the dynamic time warping method. Also, the noise within the time-frequency distribution (TFD) of the segmented VAG signals was reduced by the singular value decomposition algorithm, and a back-propagation neural network (BPNN) was used to classify the normal and abnormal VAG signals. A total of 1408 segments (normal 1031, patient 377) were used for training and evaluating the BPNN. As a result, the average classification accuracy was 92.3 plusmn 0.9 %. The proposed method showed good potential for the non-invasive diagnosis and monitoring of joint disorders.