Abstract

Info

Title:: Epilepsy detection method based on multi-scale adaptive residual network

Author(s):: ZHANG Peiling; HOU Kang; School of Physics and Electronic Information, Henan Polytechnic University, Jiaozuo 454003, China

Keywords:: electroencephalography signal; epilepsy; empirical mode decomposition; multi-scale adaptive residual network; attention mechanism

Abstract:: A novel approach based on multi-scale adaptive residual network (MSAR) is proposed to address the issues of single input data and inadequate feature extraction in current epilepsy detection approaches. The first 5 orders intrinsic mode functions for electroencephalography signal is obtained using empirical mode decomposition, and the decomposed the first 5 orders intrinsic mode functions are input into MSAR which incorporates CBAM-Residual and multi-scale adaptive convolutional network to extract multi-scale time-frequency information as well as fine-grained features of the signal. Subsequently, the signal features extracted by MSAR are fused and input into the fully connected layer to realize classification. The proposed approach obtains a classification accuracy of 98.94% on the CHB-MIT dataset, which is a notable improvement above the existing methods.