UltraMic. QIF 2023 to 2024

UltraMic wins Qualcomm Innovation Fellowship 2023-2024

The computer is everywhere now, and people would like to make computers to understand the world around them so that they can activate the home/indoor automation system. To achieve that, the computer should be able to identify different activities that people are doing, which is activity recognition. Camera, as one of the most common ways of conducting activity recognition, has many weaknesses such as causing privacy concerns and performing badly in a dim environment.

UltraMic is the FPGA-based microphone array focusing on collecting the ultrasound and achieving activity recognition. Instead of using a camera, a microphone array will not be affected by the light condition and the ultrasound frequency range can better protect people’s privacy. The project is still under development. Currently, we are able to achieve a 2D localization based on a 50kHz sampling rate and a cross-shape microphone array structure composed of 13 microphones on our microphone array board. CHECK the video demo below ;).

In the demo video (the sound happens to be not recorded), I use my phone to play a 440Hz sine wave signal. You are able to see ten most recent system predictions on sound source location. Qualitatively speaking, the performance is quite good, right? Later, the demo doing ultrasound source localization with a 192KHz system sampling rate will be provided soon.

How I look from the third-person perspective

Life would be too smooth if it had no rubs in it.