Reduction in Classification Errors for Myoelectric Control of Hand Movements with Independent Component Analysis

Myoelectric control has been an important area of investigation by researchers during the past forty years. An important role has been played by myoelecric signals in prosthetic control, since it targets the amputees who lost their body limb either in an accident or in the war. Remarkable advances were achieved with the number of movements to be classified with a high accuracy. This paper presents Independent Component Analysis (ICA) as a pre-processing technique for myoelectric control. Three different window lengths were investigated in the current study (64 ms, 128 ms and 256 ms). Two classification schemes were applied based on Time Domain-Auto Regression (TDAR) features with Uncorrelated Linear Discriminant Analysis (uLDA) and Principle Component Analysis (PCA) for dimensionality reduction, and Linear Discriminant Analysis (LDA) classification. The ICA pre-processing technique increased the classification accuracy for different window lengths used from 88% to 93%. The results suggested that FastICA consistently improves the performance across all window lengths and classification schemes. Keywords—Myoelectric Control, Surface Electromyography, ICA, TDAR and LDA