Hand gesture recognition method based on distance-speed features
暂无分享,去创建一个
The invention provides a hand gesture recognition method based on distance-speed features. The method comprises the steps of 1, adopting a frequency-modulated continuous-wave radar as a gesture sensor, intercepting and arranging received beat signals to obtain a radar echo signal two-dimensional matrix; 2, subjecting the radar echo matrix obtained in the step 1 to two-dimensional FFT respectively along a fast time a dimension and a slow time dimension; 3, based on an R-D graphic sequence, training a convolutional neural network; 4, inputting the real-time R-D sequence into a convolutional neural network model and conducting the recognition test, wherein the test accuracy of the final convolutional neural network is larger than 98%; 5, realizing the operation of a computer terminal and the control of a material object. According to the technical scheme of the invention, the hand gesture recognition is enabled even in the absence of light and without the wearing of a sensor, so that the interaction between the computer terminal and the material object is realized. The real-time detection is additionally conducted, so that the uncertain location problem of data interception is ingeniously solved. The real-time gesture recognition and control is realized. Therefore, the method has better practical value.
[1] Youngwook Kim,et al. Hand Gesture Recognition Using Micro-Doppler Signatures With Convolutional Neural Network , 2016, IEEE Access.