Abstract: With the rapid development of smart phones, people’s life has been improved. User experience of mobile device can be improved in many scenarios with the assist of in-air gesture recognition. In this paper, AGRS, an in-air gesture recognition system that can be applied to off-the-shelf smart devices is proposed. In AGRS, the loudspeaker is used to emit a continuous 20 kHz tone and the microphone is used to receive the gesture-reflecting ultrasonic wave. AGRS can recognize the position of mobile device by the gyroscope. After the gesture has been separated from noise, AGRS adapts CFAR algorithm. The characteristic of ultrasonic can be extracted by using the Doppler shift and be handled by FFT conversion. Finally, suitable classifier has been chosen to recognize gestures of AGRS. The experimental results demonstrate that AGRS is able to achieve gesture recognition accuracy of above 95%.

Key words: ultrasonic, gesture recognition, Doppler

摘要： 现今智能手机发展迅猛，人们生活质量得到大幅提高。为了更有效地利用智能手机资源，提高用户体验，提出了一种使用超声波的智能手机手势识别系统（AGRS系统）。该系统使用移动设备自带的扬声器发射20 kHz的超声波信号，使用话筒接收反射信号。AGRS系统可通过陀螺仪辅助判断当前手机摆放状态。系统使用虚警率以降低手势误识别率。AGRS利用声波的Doppler效应提取特征值，用FFT算法处理声波信号，最后选择适合的分类器对手势进行识别。实验结果证明AGRS系统手势识别率超过95%。

关键词: 超声波, 手势识别, 多普勒效应