The prime objective of this piece of work is to devise novel techniques for computer udbased classification of Electrocardiogram (ECG) arrhythmias with a focus on less computational time and better accuracy. As an initial stride in this direction, ECG beat classification is achieved by using feature extracting techniques to make a neural network (NN) system more effective. The feature extraction technique used is Wavelet Signal Processing. Coefficients from the discrete wavelet transform were used to represent the ECG diagnostic information and features were extracted using the coefficients and were normalised. These feature sets were then used in the classifier i.e. a simple feed forward back propagation neural network (FFBNN). This paper presents a detail study of the classification accuracy of ECG signal by using these four structures for computationally efficient early diagnosis. Neural network used in this study is a well-known neural network architecture named as multi-Layered perceptron (MLP) with back propagation training algorithm. The ECG signals have been taken from MIT-BIH ECG database, and are used in training to classify 3 different Arrhythmias out of ten arrhythmias. These are normal sinus rhythm, paced beat, left bundle branch block. Before testing, the proposed structures are trained by back propagation algorithm. The results show that the wavelet decomposition method is very effective and efficient for fast computation of ECG signal analysis in conjunction with the classifier.
展开▼
