Optimized MK-ELM by BAS and its Application in sEMG Gesture Recognition

Abstract

In order to solve the problem of insufficient generalization performance and kernel parameter selection of the kernel extreme learning machine, a new model based on optimized multiple kernel extreme learning machine by beetle antennae search algorithm (BAS-MK-ELM) is proposed. In the model, the weighted combination of polynomial kernel function and Gaussian kernel function is selected as the kernel function of multiple kernel extreme learning machine (MK-ELM), and beetle antennae search algorithm is introduced to optimize the kernel function parameters to get the optimal kernel function parameters. This classification model is applied to sEMG gesture recognition, and four time-domain features of the pre-processed signalsl are extracted, and signal features are classified and recognized by BAS-MK-ELM. The experimental results show that the average recognition rate of the proposed classification model is higher than that of single-kernel extreme learning machine, multi-kernel support vector machine, MK-ELM and particle swarm optimization multi-kernel extreme learning machine, which verifies the effectiveness and superiority of the proposed classification model.