A Novel Personal Best- Control Binary Particle Swarm Optimization (NPbest-BPSO) based electromyography (EMG) signal feature selection and classification

Electromyogram signal features provide the inputs needed to understand the hand’s physical behavior or how it interacts with other objects, which are ultimately necessary for accurate electromyogram signal classifications. Though expanding in the quantity of electromyography feature selection ( f s e l e c t i o n ) corrupts arrangement execution, yet in addition, the same builds or insinuates the unpredictability of the classifier. The selection of features from the electromyography signals is a compelling procedure for dispensing with repetitive and unessential signal features. This article proposes an algorithm called Novel Personal Best controlled with Binary Particle swarm optimization algorithm (NPbest-BPSO) to tackle the electromyography signal feature selection for its corresponding signal classification. Initially, Discrete Wavelet Transform breaks down the sign into multiresolution coefficients. The feature is then extricated from every coefficient to shape the (feature) element vector. Subsequently, the NPbest-BPSO is utilized to assess the most revealing features from the benchmark feature dataset. Few existing algorithms, such as the genetic algorithm (GA), the binary differential evolution algorithm (BDEA) and the Binary Particle Swarm Optimization Algorithm (BPSO) were utilized to prove NPbest-BPSO has better performance compared with other few existing algorithms. Our test results show that NPbest-BPSO achieves a feature reduction rate of up to 97.22% (abridging nearly 92% of the original features) while still maintaining a mean classification accuracy of 92.1% across 15 amputee subjects, thus ensuring both dimensionality reduction and high signal fidelity. Henceforth, we intend for the NPbest-BPSO algorithm to be more adept at clinical-centric applications.