Research On Epileptic EEG Signal Classification Algorithm Based On LSTM Network And FPGA Implementation

In recent years,with the development of Brain Computer Interface(BCI)technology and deep learning technology,researchers are trying to apply these technologies to the field of medical rehabilitation and have made some achievements,among which the portable epileptic seizure prediction system based on deep learning network has become a research hotspot in this field.Epileptic EEG signal classification algorithm is a core part of the portable epileptic seizure prediction system,which directly determines the classification system performance.This paper investigates the key technologies involved in the portable epileptic seizure prediction system,including the design of epileptic EEG signal classification algorithm,hardware implementation and optimization based on FPGA and the building of verification platform.According to the characteristics of Epileptic EEG signals and the requirements for the implementation of portable epileptic seizure prediction system,this paper proposes to combine the wavelet transform algorithm with the Long-Short Term Memory(LSTM)network to extract and classify the features of epileptic EEG signals,which is finally implemented on the FPGA platform.The main contributions are listed as follows:(1)This paper proposes to use Discrete Wavelet Transform(DWT)to decompose signals and extract statistical features to reveal the time-frequency characteristics of epileptic EEG signals.Then,the extracted feature sequence is sent to LSTM network to mine the most discriminative features for classification.The algorithm was trained and tested in the CHB-MIT public dataset,and fragment-based evaluation and event-based evaluation criteria are used to evaluate the performance of the algorithm respectively.Finally,the algorithm is compared with other classification algorithms,and the results show that the performance of this algorithm is better.(2)In order to meet the requirements of epileptic seizure prediction system for the real-time and portability,this paper uses the FPGA platform to design the hardware circuit of epileptic EEG signal classification algorithm.In the hardware design,the IP core of epileptic EEG signal classification algorithm is designed on Vivado HLS including data type selection,DWT feature extraction and classification module based on LSTM model,and then the functional simulation is performed to verify the logic function of each module.Finally,the IP core on Vivado HLS is realized and the correct mapping from algorithm design to hardware implementation is completed.(3)In order to improve the calculation speed of hardware circuits and enhance the real-time prediction of epileptic seizure,this paper adopts the multiplexing optimization strategy and parallel optimization strategy to optimize the IP core of the classification algorithm according to the parallel characteristics of proposed algorithm.In order to ensure the integrity of system function,the filter design of epileptic EEG signals is completed based on FPGA.Finally,a hardware system with epileptic EEG classification algorithm hardware IP as the core is built in Vivado software,and the series of Xilinx’s Zynq7000 is used for the board level verification of the overall circuit.This paper investigates the epileptic EEG signal classification algorithm and proposes to combine the wavelet transform algorithm with the Long-Short Term Memory network to extract and classify the features of epileptic EEG signals.Then the hardware circuit of epileptic EEG signal feature extraction and classification algorithm is implemented on the FPGA platform,and the hardware circuit of the algorithm is optimized and improved by multiplexing optimization and parallel optimization strategy,which improves the calculation speed of hardware circuits and enhances the real-time prediction of epileptic seizure,and finally meets the requirements of portable epileptic seizure prediction system.