基于CNN和超声传感的手势识别及辅助身份认证

Gesture recognition and auxiliary identity authentication based on CNN and ultrasonic sensing

基于超声波的非接触式传感扩展了移动设备的交互范围,提出了一种超声传感系统,它可以在多种表面材料上工作,只需要一对声音收发器。该系统使用了多普勒频移和信道脉冲响应来捕捉微妙的手部动作;然后通过卷积神经网络(CNN)从测量的信号数据中提取判别特征,形成对用户手势动作进行识别和身份验证的超声传感系统。在智能手机上实现了超声传感系统,并进行了大量的真实环境实验来评估其性能。实验结果表明:仅使用一个收发器对,它的性能可以与多收发器对实现相媲美;还可以用于多种材料表面,且不需要捕获用户的生物信息即可对用户手势识别,以及辅助身份验证的准确性超过96%。

Ultrasonic-based non-contact sensing expands the interactive range of mobile devices and propose an ultrasonic sensing system that can work on a variety of surface materials and only requires a pair of sound transceivers.The system uses Doppler frequency shift and channel impulse response to capture subtle hand movements,and then uses a convolutional neural network(CNN)to extract discriminative features from the measured signal data to form ultrasonic sensing system for recognition and identity verification of user gestures.An ultrasonic sensing system is implemented on a smartphone,and a large number of real environment experiments are performed to evaluate its performance.The experimental results show that using only one transceiver pair,its performance can be comparable to that of multiple transceiver pairs,and it can also be used on a variety of material surfaces,and can recognize user gestures and assist identity without capturing the user’s biological information.The verification accuracy exceeds 96 %.