Analysis of Electromyography Signals with Recurrence Plots in Detection of Knee Disorders

The problems that occur in the knee and cause symptoms such as movement restriction and pain are disorders that require priority to be diagnosed and treated because they make it difficult for people to move. Since these problems usually originate from the muscle tissue surrounding the knee and therefore from muscle functions, the evaluation of electromyography (EMG) signals used to measure muscle activity through the skin in the detection of knee disorders has become an important area of study in recent years due to the easy acquisition of these signals. In this study, EMG signals recorded during walking, sitting and standing from 11 patients with different knee anomalies and 11 healthy individuals from 5 muscle groups responsible for bending and stretching movements were evaluated with the features obtained from recurrence graphs for the detection of knee disorders. The quantities obtained with the recurrence analysis were classified using K-Nearest Neighbor (KNN), Support Vector Machine (SVM) and Neural Network (NN) algorithms. The classification performances obtained for the detection of knee problems were evaluated in terms of movement type and muscle type. The results obtained have the highest accuracy value in the walking position and for the rectus femoris (RF) muscle as KNN 100%, SVM 95.5%, NN 95.5%. The study reveals that this methodology has a potential in the field of biomedical signal processing.