A Dual Decision Fusion Approach for sEMG-based Gesture Recognition with Variable Time Windows

Natural user interface (NUI) facilitates intuitive control, enabling users to operate unmanned vehicles more naturally without relying on traditional remote controllers. NUI has thus been widely adopted across various domains, with gesture control emerging as a prominent interaction method. However, existing gesture recognition methods predominantly rely on fixed time windows, limiting their adaptability to dynamic and variable-duration gestures. To address this limitation, this study proposes a non-negative matrix factorization (NMF)-based dual decision-level fusion method that integrates a multilayer perceptron (MLP) and dynamic time warping (DTW) with K-nearest neighbors (KNN) using a weighted averaging strategy to enhance gesture recognition across multiple time frames. Experimental results demonstrate that the proposed method achieves 98.04%, 95.15%, and 95.33% accuracy for 1-second, 3-second, and 5-second time windows, respectively. The results indicate that accuracy remains consistent across different time frames, confirming the robustness of the proposed method in handling variable time windows. These findings suggest that the proposed approach is well-suited for real-world applications, including human-machine interaction and gesture-controlled systems.