SaleNet: A low-power end-to-end CNN accelerator for sustained attention level evaluation using EEG

TL;DR

SaleNet is a highly compressed, low-power CNN designed for real-time sustained attention evaluation using EEG, achieving high accuracy and efficiency on FPGA hardware.

Contribution

The paper introduces SaleNet, a novel CNN architecture with advanced model compression techniques for EEG-based attention level classification.

Findings

Achieves 84.2% accuracy on a 6-subject EEG dataset.

Provides 183.11x model compression ratio.

Operates on FPGA with 0.11 W power consumption.

Abstract

This paper proposes SaleNet - an end-to-end convolutional neural network (CNN) for sustained attention level evaluation using prefrontal electroencephalogram (EEG). A bias-driven pruning method is proposed together with group convolution, global average pooling (GAP), near-zero pruning, weight clustering and quantization for the model compression, achieving a total compression ratio of 183.11x. The compressed SaleNet obtains a state-of-the-art subject-independent sustained attention level classification accuracy of 84.2% on the recorded 6-subject EEG database in this work. The SaleNet is implemented on a Artix-7 FPGA with a competitive power consumption of 0.11 W and an energy-efficiency of 8.19 GOps/W.