FPGA-based ANN Design for Detecting Epileptic Seizure in EEG Signal

Abstract

This study aims to represent an FPGA (Field Programmable Gate Array) design of Artificial Neural Network (ANN) for Electroencephalography (EEG) signal processing in order to detect epileptic seizure. For analyzing brain's electrical activity, feedforward ANN model is used for classification of EEG signals. The designed ANN output layer makes a decision whether the person has epilepsy or not. In the proposed system, the ANN model is programmed and simulated on Xilinx ISE editor via computer and then, EEG signal data are transferred to FPGA-based ANN emulator core. The Core is trained on data which are patient's data and healthy person's data. After training, test data is loaded to ANN Emulator Core to detect any epileptic seizure of person's EEG signal. The main advantage of FPGA in the system is to improve speed and accuracy for epileptic seizure detection.

FPGA based Artificial Neural Network Processor for Detection of Epileptic Seizure

“…Actual/realtime implementation on FPGA is another challenging task owing to the limitations in terms of limited number of interfacing pins of FPGA, actual data capture and its processing, etc. The resource usage in [25] is also much more compared to our work. Our work presents real time implementation on FPGA by utilizing minimal resources (as presented in comparison Table II below) and comparable accuracy which makes it unique among other works.…”

“…Also, we are getting comparable accuracy by using hardware architecture of lesser complexity. Whereas, in [25], researchers has emulated the epilepsy detection on FPGA based ANN core emulator without actual implementation on FPGA hardware. Actual/realtime implementation on FPGA is another challenging task owing to the limitations in terms of limited number of interfacing pins of FPGA, actual data capture and its processing, etc.…”

Thermal error prediction for heavy-duty CNC machines enabled by long short-term memory networks and fog-cloud architecture

“…On the fog layer, the processor FPGA CycloneII EP2C5T144C8N, is deployed. FPGA exhibits good performances in executing the LSTM networks locally as it is of robust flexibility, reconfigurability and efficient parallel computing (Ahsan et al, 2012;Karakaya et al, 2018;Zairi et al, 2019;Sarić et al, 2020), while it is more cost effective in comparison with those processors deployed on the cloud layer. Training on the LSTM networks and FEA computations are computationally intensive, so that the LSTM training/re-training tasks and FEA are assigned to the cloud layer to leverage its better computational speed and larger memory.…”