Ultrasound Doppler Radar

Current depth sensors (LIDARs, Kinect, stereo cameras, etc.) are expensive, power hungry and often demand dedicated hardware, cooling systems and precise calibration. On top of this, further processing needs to be done to get a velocity profile of the objects being sensed. In this work, we have developed an Ultrasound-Doppler Radar to investigate the application of Doppler Effect and Time-of-Flight to simultaneously estimate the speed and location of targets in a 360 degree horizontal field of view around a circular array of ultrasound piezoelectric transducers. The array consists of 8 pairs of ultrasound loudspeakers and receivers. To improve the spatial angular resolution and the framerate of our sensing, we employ offline loudspeaker and microphone beamforming. The time-delay and Doppler stretch factors (indicating the position and velocity respectively) are estimated together each frame. The device is low form-factor, inexpensive, low power consuming and enables a variety of interesting applications for indoor and outdoor, stationary and mobile Robotics sensing and Human-Computer-Interaction.

Speaker Details

Supreeth Krishna Rao is a 2nd year Masters student pursuing Robotics Engineering at the Worcester Polytechnic Institute. His research interests include Computer Vision (Video + Audio), Machine Learning (Deep learning for recognition/scene understanding) and Human-Computer/Robot-Interaction. Supreeth joined MSR Redmond as a research intern in the Audio and Acoustics Research group. His research is available on LinkedIn: https://www.linkedin.com/in/supreethrao

Date:
Speakers:
Supreeth Krishna Rao
Affiliation:
Worcester Polytechnic Institute
    • Portrait of Jeff Running

      Jeff Running

Series: Microsoft Research Talks