Pedestrian classification based on radial velocity features of UWB Doppler radar images

Imaging and identification of people using low-complexity radar is a promising technique for surveillance systems. Many studies have been conducted on the development of target identification methods using time-frequency micro-Doppler signatures [1]-[4]. These conventional methods can realize classification of motion types such as running, sitting, walking and walking without arm-motion. However, for accurate classification using these methods, reliable and long-term data or time-consuming procedures are needed, and their accuracy and real-time capability are thus inadequate. To resolve this problem, the use of shape and motion information is an option. For this purpose, we have proposed an ultra wide-band (UWB) Doppler radar imaging method with interferometry [5]. This method achieved adequate and real-time 3-dimensional human imaging with three receiving antennas in a realistic situation. This paper presents the imaging results on a variety of pedestrians using the UWB Doppler radar imaging method, and proposes their classification parameters based on a radial velocity feature of the estimated image. We experimentally verify that an effective classification for three types of pedestrians is achieved.