Optimized adaptive neuro fuzzy inference system (OANFIS) based EEG signal analysis for seizure recognition on FPGA. (April 2021)

• Record Type: Journal Article
• Authors: B. Indira, Priyadarshini; D. Krishna, Reddy
• Is Part Of: Biomedical signal processing and control. Volume 66(2021)
• Publication Date: 2021-04
• Subjects: Electroencephalogram (EEG) -- Seizure -- Discrete wavelet transformation (DWT) -- Adaptive neuro fuzzy inference system; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102484
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: To simplify the design structure of seizure recognition by integrating feature selection and classification model. To improve performance, certain best performing statistical, linear and non-linear features are extracted from each bands. To improve classification accuracy, ANFIS classification model is optimized using BPSO based feature selection process. OANFIS classifier determines best trade-off between classification accuracy and area usage due to optimized structure. Abstract: The activities of the human brain can be affected due to certain neurological disorder called seizure. Generally, the epileptic abnormalities can be identified by direct visual scanning. But this scanning consumes more time, and it is limited due to some technical artefacts. Hence, there is a necessity of an efficient computer-aided diagnosis (CAD) system for distinguishing the seizure signals from non-seizure signals automatically. In this paper, an Optimized Adaptive Neuro Fuzzy Inference System (OANFIS) classifier is proposed to detect the seizure automatically. The main aim of this work is to increase the accuracy of the classifier with less computational complexity in terms of area and power consumption. The proposed system initially extracts the features of the FIR filtered EEG signal using Discrete Wavelet Transform (DWT). Then, these features are learned by the proposed classifier, which improves the classification accuracy by selecting the optimal features using Binary Particle Swarm Optimization (BPSO) algorithm. To provide convenience and compactness, the proposed system is implemented in Xilinx working platform by developing Verilog code. The simulation outcomes demonstrate that the proposed system overtakes the existing approaches by achieving the classification accuracy of 99.25 % and consuming only 2.018 μ W power.