Current browse context:
eess.SP
Change to browse by:
References & Citations
Electrical Engineering and Systems Science > Signal Processing
Title: TinyRadarNN: Combining Spatial and Temporal Convolutional Neural Networks for Embedded Gesture Recognition with Short Range Radars
(Submitted on 25 Jun 2020 (v1), revised 26 Oct 2020 (this version, v2), latest version 16 Mar 2021 (v3))
Abstract: This work proposes a low-power high-accuracy embedded hand-gesture recognition algorithm targeting battery-operated wearable devices using low power short-range RADAR sensors. A 2D Convolutional Neural Network (CNN) using range frequency Doppler features is combined with a Temporal Convolutional Neural Network (TCN) for time sequence prediction. The final algorithm has a model size of only 46 thousand parameters, yielding a memory footprint of only 92 KB. Two datasets containing 11 challenging hand gestures performed by 26 different people have been recorded containing a total of 20,210 gesture instances. On the 11 hand gesture dataset, accuracies of 86.6\% (26 users) and 92.4\% (single user) have been achieved, which are comparable to the state-of-the-art, which achieves 87\% (10 users) and 94\% (single user), while using a TCN-based network that is 7500x smaller than the state-of-the-art. Furthermore, the gesture recognition classifier has been implemented on a Parallel Ultra-Low Power Processor, demonstrating that real-time prediction is feasible with only 21 mW of power consumption for the full TCN sequence prediction network, while a system-level power consumption of less than 100 mW is achieved. We provide open-source access to all the code and data collected and used in this work on tinyradar.ethz.ch.
Submission history
From: Moritz Scherer [view email][v1] Thu, 25 Jun 2020 15:23:21 GMT (11837kb,D)
[v2] Mon, 26 Oct 2020 14:44:10 GMT (12404kb,D)
[v3] Tue, 16 Mar 2021 15:33:18 GMT (13668kb,D)
Link back to: arXiv, form interface, contact.