PCA and LDA for EMG-based control of bionic mechanical hand

Electromyography (EMG) has some good abilities for bionic mechanical hand´s control and researchers have proposed many kinds of methods for EMG classification. Principal Components Analysis (PCA) which is an ideal tool for dimension reduction tool was introduced for EMG classification. Linear Discriminant Analysis (LDA) performs outstandingly on classification. This paper does a comparative study on PCA and LDA for EMG classification, mainly including LDA for raw EMG, LDA for features, PCA and LDA for raw EMG and PCA and LDA for features. Here five time-domain features and four frequency-domain features are selected. The five hand motions including hand closing, hand opening, index finger pinching, middle finger pinching and hand relaxing are selected for classification. The result shows PCA and LDA for features obtain 99.0% motion success rate and 99.8% success rate of classification. The bionic mechanical hand got a good performance.