sEMG Signal Gesture Recognition Method Based on Dung Beetle Optimized Random Forest

The control of manipulator by hand surface electromyography (sEMG) signal is a key area of human-computer interaction research. One of the challenges in gesture recognition is to ensure high accuracy while maintaining a fast recognition speed. This article uses recursive feature elimination with cross-validation (RFECV) to select relevant features, enhancing model performance. Therefore, we propose an optimized random forest (RF) model inspired by dung beetle algorithm that mimics four key behaviors of dung beetles: rolling, breeding, foraging, and stealing. This approach introduces four optimization methods to determine the optimal number of trees and tree depth, thereby improving both the global and local search capabilities. Experimental results demonstrate that the RF model based on dung beetle optimization (DBO-RF) significantly enhances the recognition accuracy of sEMG signals. Results from experiments on both our self-collected dataset and the public NinaproDB1 dataset indicate that the model exhibits superior performance and robustness compared to state-of-the-art deep learning models and traditional classification models. Additionally, experimental validation of the model's rapid recognition capability through robotic arm control provides new insights for research in the field of human-computer interaction.