Classification of Arrhythmias with LDA and ANN using Orthogonal Rotations for Feature Reduction

This paper presents a new approach for feature reduction by using orthogonal rotations. Wavelet coefficients for beat segments are taken as features which are reduced by factor analysis method using orthogonal rotations. LDA (Linear Discriminant Analysis) and ANN (Artificial Neural Network) classifiers are used for classification. The signals are taken from MIT-BIH arrhythmia database to classify into Normal, PVC, Paced, LBBB and RBBB. The performance of classification output has been compared by the performance parameters. Both the classifiers have given best overall accuracy for equimax rotation. 96% accuracy is achieved with LDA classifier,99.2% accuracy is achieved using ANN.