연속파 레이더를 활용한 작업자의 움직임 분석 방법

Abstract

In addition to security and surveillance, there is an increasing demand to analyze the behavior of workers in the workplaces. Such applications are implemented by recognizing specific motion through a vision system that works with machine learning and classifiers. However, image-based motion recognition systems may be limited in its use by the fact that they are vulnerable to malicious attacks and may cause privacy infringement. In this paper, we propose a method for analyzing the motions of the human body using electromagnetic waves to overcome these drawbacks. In particular, this study focuses on the use of continuous wave radar technology, which is affordable and enables small form factor in implementing application systems. When the continuous waves reflected from the target are transformed into data in the time-frequency domain, the change of Doppler shift over time can be easily observed and the characteristics related to the motions can be found. In previous studies, it has been common to perform a short-time Fourier transform of a radar signal in order to achieve this purpose, but this has limitations in observing the motion of the human body in which the change of the Doppler frequency is extremely small. In contrast, a short-time chirp Z-transform of radar signals was carried out to ensure high frequency resolution in the time-frequency domain in this study. Then, the time-frequency spectrum was converted into a gray scale image, and filters were applied to the image in order to emphasize the distinctive image components. The gradient information was extracted from the filtered image and the feature descriptors that characterized the motions were constructed from it. Feature descriptors are expected to be fed into an image classification algorithms to identify specific motions.