Published on 25 Aug 2020

Putting a Brain on a Chip

Putting a Brain on a Chip

Portable and wearable devices such as health sensors could soon become smarter with a new invention by Nanyang Technological University (NTU).

Researchers from NTU’s School of Electrical and Electronic Engineering (EEE) have developed a smart chip that could imbue devices with artificial intelligence at minimal energy and cost.

The small and lightweight prototype, when mounted on a headgear, could help decode a monkey’s brain waves and predict how it wanted to move its fingers with 99.3% accuracy. This could pave the way for implants and portable devices to help paralysed people regain use of their limbs.

The chip could also enable surveillance cameras and sensors to automatically identify intruders and alert security staff, and empower health-monitoring devices to recognise when a person is about to suffer a heart attack or epileptic seizure, and call for help.

The invention taps on manufacturing imperfections that occur in the electronics sector. Transistors in electronics devices invariably deviate to some degree from the design and from one another due to manufacturing inconsistency.

The chip taps the resulting, unintended variations in the transistors’ threshold voltage. It uses mathematical formulas that multiply input signals such as brain waves, temperatures, pressure, sounds and images using the transistors’ varying threshold voltages. This fleshes out similarities and differences among the signals.

The chip can then better compare each signal to specified patterns. A doctor could study an epilepsy patient’s brain activity and programme the chip to recognise the signature of an impending episode, so that help can be activated when needed.

To achieve the same results, conventional smart chips have to use an additional component and more processes, so that their power usage is more than 100 times that of the NTU EEE chip.

The NTU EEE scientists are developing an implant integrated with their chip for brain-machine interfaces. They have also received a $250,000 grant from the Singapore-MIT Alliance for Research and Technology to improve their prototype.