Neuromorphic computing system for real-time signal monitoring and classification with ultra-low-power 2D devices

Introduction

The detection and classification of electrophysiological signals (EPSs), such as electroencephalography (EEG) and electromyography (EMG) recordings, are the gold standard in neuroscience. These techniques enable the identification of digital biomarkers capable of health monitoring, personalised medicine, and advanced brain-computer interfaces (BCIs).

Current Technology Limitations

The state-of-the-art technology in this field, however, still relies on bulky, inefficient microelectronic systems which depend on artificial intelligence (AI) in the cloud.

Proposed Solution

The energy efficiency and classification accuracy can be largely improved by neuromorphic computing with emerging materials and devices capable of mimicking the neural mechanisms in our brain.

Project Goals

This project aims at developing a novel class of neuromorphic systems based on reservoir computing (RC) in charge trap memory (CTM) based on 2D semiconductors.

Advantages of 2D-CTM Devices

1. 2D-CTM devices are able to extract features from EPSs at extremely low power.
2. They provide high accuracy of classification.
3. They thus offer efficient biomarkers for medical diagnosis and BCIs.

Application and Impact

The project will develop the RC system based on the 2D-CTM technology for a broad application space, with the goal of establishing a novel technology platform for scalable, low-power implantable/wearable chips for real-time EPS monitoring and classification.