Featured Based Segmentation Method for Buliding Millimeter Wave Radar Gesture Recognition Data Sets

Abstract

Gesture recognition based on artificial neural network is an important application of the millimeter wave radar. In addition to extracting gesture features and constructing neural networks, the establishment of effective dynamic gesture data sets is also the direction worth paying attention to in gesture recognition research. In order to solve some problems caused by fixed radar observation time and multiple measurements of single gesture during the establishment of gesture data set, six gestures were tested in a single radar observation. Combining the relationship between actual gesture motion and speed, the data file containing six gestures is divided into six sets containing only one specific gesture which will be used in the gesture recognition system of millimeter wave radar based on neural network finally.