Single-Channel sEMG Achieves High Accuracy in Gesture Classification

Traditional hand gesture recognition using surface electromyography (sEMG) typically relies on multiple sensor channels and computationally intensive models, which limits its use in low-power, embedded applications. This research explores whether accurate classification of ten distinct hand gestures is possible using only a single sEMG channel, paired with more lightweight machine learning architectures. The methodology involved transforming raw sEMG signals into a comprehensive set of features, encompassing time-domain, frequency-domain, higher-order-crossing, and relative-intensity metrics. To manage feature redundancy, Pearson correlation filtering was applied, and highly correlated features were removed. Dimensionality reduction techniques like LDA and PCA were also selectively employed. Three different classifiers—a feed-forward neural network (NN), k-nearest neighbors (KNN), and a support vector machine (SVM)—were systematically evaluated. The results indicate that combining time and frequency features with Pearson filtering and a compact neural network can achieve impressive accuracy, reaching up to 90%. These findings highlight the significant potential of single-channel sEMG systems for developing cost-effective and low-power gesture recognition solutions.