FPGA implementation of Wavelet Neural Network for epileptic seizure detection

Abstract

This paper introduces an implementation of Wavelet Neural Network (WNN) with Particle Swarm Optimization (PSO) learning ability on Field Programmable Gate Array for detection of severity index of epileptic seizure detection. The Electroencephalography (EEG) signals were first pre-processed using discrete wavelet transforms (DWTs). This was followed by the feature selection stage, where five wavelet parameters and four representative summary statistics were computed. Four different activation functions were used in the hidden nodes of WNNs. The best combination to be used was the WNNs that employed Haar wavelet as the activation function, with Haar wavelet along with Heursure soft thresholding at the feature extraction stage. The PSO learning method is used in this paper. Twenty known epilepsy patients in nine defined groups are studied. The efficacy of the WNN is analyzed through sensitivity, specificity and classification accuracy. Higher the benchmark values will be the better classifier.