Abstract
In this paper, we propose a method of identifying human motions, such as standing, walking, running, and crawling, using a millimeter wave radar sensor. In our method, two signal processing is performed in parallel to identify the human motions. First, the moment at which a person’s motion changes is determined based on the statistical characteristics of the radar signal. Second, a deep learning-based classification algorithm is applied to determine what actions a person is taking. In each of the two signal processing, radar spectrograms containing the characteristics of the distance change over time are used as input. Finally, we evaluate the performance of the proposed method with radar sensor data acquired in an indoor environment. The proposed method can find the moment when the motion changes with an error rate of 3%, and also can classify the action that a person is taking with more than 95% accuracy.
Highlights
As the center frequency used by radar sensors increases, the size of the radar is getting smaller and its range resolution is getting higher
A microwave Doppler radar sensor was used to detect people moving behind the wall [9], and an impulse radio ultra-wideband (UWB) radar was used to identify the location of a person in a room [10]
Because the convolutional neural network (CNN) is trained with images, we used the cropped radar spectrogram described in Section 3.1 as input
Summary
As the center frequency used by radar sensors increases, the size of the radar is getting smaller and its range resolution is getting higher. A method of classifying arm motions by applying deep learning techniques has been proposed [14] Based on these studies, we propose a more advanced method that finds the moment when a person’s motion changes and at the same time discriminates the motion using a radar sensor in an indoor environment. Unlike studies using Doppler radars [3,12,14], the characteristics of the distance change over time of an object can be recognized by the FMCW radar system. Because we use the change in the statistical characteristics of the radar signal over time, we can determine the moment when the motion changes, which is not considered in most related studies. We accumulated over 500 spectrograms for each single motion and two consecutive motions
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
