Real-time Arm Gesture Recognition in Smart Home Scenarios via Millimeter Wave Sensing

• DOI: 10.1145/3432235
• Publication Date: 2020-12-18T15:39:14Z
• Authors: Haipeng Liu; Yuheng Wang; Anfu Zhou; Hanyue He; Wei Wang; Kunpeng Wang; Peilin Pan; Yixuan Lu; Liang Liu; Huadong Ma
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

"In air" gesture recognition using millimeter wave (mmWave) radar and its applications in natural human-computer-interaction for smart home has shown potential. However, the state-of-the-art works still fall short terms of limited space, vulnerable to surrounding interference, off-line recognition. In this paper, we propose mHomeGes, a real-time mmWave arm system practical home-usage. To end, first distill gesture's position dynamic variation, then custom-design lightweight convolution neural network recognize fine-grained gestures. Next, user discovery method focus on target human gesture, thus eliminating adverse impact interference. Finally, design hidden Markov model-based voting mechanism handle continuous signals at run-time, which leads real-time. We implement mHomeGes commodity also perform study, demonstrates that achieves high accuracy above 95.30% across various scenarios, regardless movements, concurrent gestures, physiological conditions, outer packing materials. Moreover, have publicly archived data-set collected during developing consists about 22,000 instances from 25 persons may an independent value facilitating future research.