Smart Speaker Based on Detection of Millimeter Wave

Yu-Ping Liao, Feng-Kai Huang, Ying-Jieh Xia, Hsiang Cheng

Abstract: In order to resolve privacy concerns in smart homes, we propose a gestures controllable smart speaker with a millimeter wave radar. Furthermore, the radar can be turned to face wherever nearby hand-clapping sound comes from. By running deep neural networks (DNN) on a small, powerful computer-NVIDIA Jetson Nano Developer Kit, we can analyze and classify 5 dynamic arm gestures in real-time. The results of gesture recognition will cause the corresponding function of controlling music. The average accuracy is above 99%.