基于无线信号和改进TCN的手势识别方法

Gesture recognition method based on wireless signal and improved TCN

针对传统手势识别方法环境适应能力较弱、特征提取能力较差等问题,提出一种基于无线信号和改进时间卷积网络(TCN)的手势识别方法GRT(gesture recognition with improved TCN)。将无线信号中提取到的信道状态信息幅值数据使用巴特沃斯低通滤波、离散小波变换以及数据归一化进行预处理,凸显信号特征;对TCN进行改进,设计密集连接结构并优化时间模块,降低网络计算量;引入多输入网络进行训练,深度提取数据中的特征信息并生成模型,实现10种手写阿拉伯数字手势的分类识别。实验结果表明,GRT方法平均识别精度为98.3%,相较其它方法具有更好的识别效果。

Aiming at the problems of traditional gesture recognition methods such as weak environmental adaptability and poor feature extraction capabilities,a gesture recognition method GRT(gesture recognition with improved TCN)based on wireless signals and improved temporal convolutional networks(TCN)was proposed.The channel state information amplitude data extracted from the wireless signal was preprocessed using Butterworth low-pass filtering,discrete wavelet transform,and data normalization to highlight the signal characteristics.The TCN was improved,and the dense connection structure was designed and optimized.The temporal module effectively reduced the amount of network calculations.A multi-input network was introduced for training,and the feature information in the data was deeply extracted and a model was generated to realize the classi-fication and recognition of ten handwritten Arabic numerals gestures.Experimental results show that the average recognition accuracy of the GRT method is 98.3%,which has better recognition effects than other methods.