ECG SIGNAL PROCESSING USING ANN

Objective:- to classify the ECG signal as normal or abnormal

Artificial Neural Network can be used for classification of bio-medical signals. We used Artificial Neural Network to classify the ECG signals. The Wavelet Transform is a transformation that can be used to analyze the temporal and spectral properties of non-stationary signals like ECG. This work contains an overview of the uses of the Wavelet processing of the ECG for decomposition and its analysis using Artificial Neural Network.

Methodology

In the case of the (Daubechies 4) Wavelet decomposition, a pair of 4-coefficient FIR filters is used to separate the input signal into a hierarchical set of fast and slow components. A high pass filter computes the fast components and a low pass filter computes the slow components.

The preprocess information is provided to the Neural Network. The decomposed signal is given to feedforwrd network, trained using backpropogation algorithm. The data is provided to train Neural Network shows two-class.

Back propagation uses supervised learning in which the network is trained using data for which inputs as well as desired outputs are known. Once trained, the network weights are frozen and can be used to compute output values for new input samples.

The feed forward process involves presenting an input pattern to input layer neurons that pass the input values onto the first hidden layer. Each of the hidden layer nodes computes a weighted sum of its inputs, passes the sum through its activation function and presents the result to the output layer.

Conclusion

The Wavelet Transform provides method to divide the time- frequency (or time-scale) plane. The two major advantages offered by Wavelet approach was that it enables one to zoom into time discontinuities and that basis localized both in time and frequency can be calculated. The decomposed data is provided to (32 - 20 – 1) Neural Network.

We have used feedforword neural network with sigmoid node function, trained by error backpropogation algorithm. The main idea underlying backpropogation is gradient descent. Each weight is updated by a small quantity proportional to the negative of the partial derivative of the mean squared error with respect to that weight. Neural Network classify the signal as normal signal or abnormal signal.

Future Scope

It is possible to use backpropogation for many-input many-output function approximation, where each output node in the network corresponds to a dimensional of the out vector.