Improving the accuracy of hand sign recognition by chaotic swarm algorithm-based feature selection applied to fused surface electromyography and force myography signals

This study addresses the challenge of distinguishing ambiguous American Sign Language (ASL) gestures by integrating surface electromyography (sEMG) and force myography (FMG) signals. A Chaotic Salp Swarm Algorithm (CSSA) with six chaotic maps was employed for feature selection, followed by k-Nearest Neighbors (kNN) classification. The objective of this approach was to enhance the classification accuracy of 20 American Sign Language (ASL) hand signs using individual and combined sEMG and FMG signals. The experiment involved the collection of data from ten participants, with each participant repeating the ASL postures ten times. The results of the study demonstrated that the CSSA effectively reduced the number of features required for classification while slightly improving the classification accuracy. The fusion of sEMG and FMG achieved an average accuracy of 88.5%, surpassing the accuracy of individual sEMG (81%) and FMG (76.75%) signals. Moreover, after systematically eliminating eight ambiguous gestures, the model attained 100% accuracy. These findings underscore the potential of integrating sEMG and FMG signals with CSSA-based feature selection to enhance ASL gesture recognition. Furthermore, the study identifies specific ambiguous postures that must be excluded to achieve perfect classification.