IR-UWB Radar Sensor for Human Gesture Recognition by Using Machine Learning

In this paper, we propose a human gesture recognition algorithm using impulse radio ultra-wideband (IR-UWB) radar. The radar signal is transmitted into a three dimensional space, however, the received signal is only expressed in one dimensional. Therefore, it is difficult to classify 3-D gestures by analyzing specific features, such as power, peak value, index of peak value, and other values of received signal. To resolve this problem, a new human gesture recognition algorithm using machine learning is proposed. Two machine learning technics are used in this paper. One is unsupervised learning technic which is used for extracting features from received radar signal is principal component analysis, and the other one is supervised learning which is used for classifying gestures. The features are extracted by using the principal component analysis (PCA) method, then neural network method is used for training and classifying gestures using the extracted features. In training and classifying step, other method can be used, such as supporting vector machine (SVM), however, this method is hard to recognize noise gesture which means untrained gesture. To resolve this problem, we use neural network method in this paper, then in order to classy noise gestures and trained gestures, a noise determining algorithm is used.This work was supported by the ICT R&D program of MSIP/IITP. [B0126-16-1018, The IoT Platform for Virtual Things, Distributed Autonomous Intelligence and Data Federation/Analysis