William & Mary

2018 Park Graduate Research Award

yma.jpg
The annual Park Graduate Research Award was given in 20017/08 to Yongsen Ma. Yongsen was nominated by his advisor, Gang Zhou, for his paper SignFi: Sign Language Recognition Using WiFi, which is published in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT). The paper will be presented at the 2018 ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp 2018). This research combines WiFi and neural networks for recognizing American Sign Language (ASL) gestures. It provides a non-intrusive way to recognize nearly 300 ASL gestures with high accuracy and low cost. The rationale of the paper is that WiFi signals are impacted by nearby persons that are making sign gestures. Recently, many papers use WiFi to recognize hand and finger gestures. But existing WiFi-based ASL recognition methods are tested on no more than 25 ASL gestures. This paper demonstrates that existing recognition algorithms have low accuracy and high cost when there are nearly 300 ASL gestures. To address this problem, the paper proposes a convolutional neural network as the recognition algorithm. It has 94% accuracy for 276 ASL gestures performed by one user in two environments. For 150 ASL gestures performed by five users in a lab environment, the accuracy is 86%. The proposed algorithm has very low test cost. It takes only 0.62 milliseconds to recognize a new gesture