Krishnakant Saboo

Gesture Recognition using FMCW radar with low angle resolution


Guide: Mr. Sandeep Rao, Texas Instruments, Bangalore, India

Background

There are myriads of applications of gesture recognition. Though the camera is a good sensor, its applicability in situations with bad lighting conditions, harsh weather, or long ranges is limited. Radars can overcome some of the these limitations. We attempted to create a gesture recognition algorithm using Frequency Modulated Continuous Wave (FMCW) nano-radars with low angle resolution mounted on a chip.

Approach

The data being streamed from the nano-radar chip is 4-dimensional- (distance, velocity, magnitude, time). We developed a projection scheme for effectively representing the data corresponding to hand gestures performed in front of the chip. A gesture recognition algorithm using the projection as a feature and developed. It achieved a recognition accuracy of 87% for gestures within 1m of the chip and 84% for gestures done within 3m of the chip.

A US patent has been filed for the developed algorithm with me named as the first inventor.