MindReading: An Ultra-Low-Power Photonic Accelerator for EEG-based Human Intention Recognition

Abstract

A scalp-recording electroencephalography (EEG)-based brain-computer interface (BCI) system can greatly improve the quality of life for people who suffer from motor disabilities. Deep neural networks consisting of multiple convolutional, LSTM and fully-connected layers are created to decode EEG signals to maximize the human intention recognition accuracy. However, prior FPGA, ASIC, ReRAM and photonic accelerators cannot maintain sufficient battery lifetime when processing real-time intention recognition. In this paper, we propose an ultra-low-power photonic accelerator, MindReading, for human intention recognition by only low bit-width addition and shift operations. Compared to prior neural network accelerators, to maintain the real-time processing throughput, MindReading reduces the power consumption by 62.7\% and improves the throughput per Watt by 168\%.

Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.