Aspect angle dependence and multistatic data fusion for micro‐Doppler classification of armed/unarmed personnel

Abstract

This paper discusses the analysis of multistatic micro-Doppler signatures and related features to distinguish and classify unarmed and potentially armed personnel. The application of radar systems to distinguish different motion types has been previously proposed and this work aims to further investigate the applicability of this in more scenarios. Real data have been collected using a multistatic radar system in a series of experiments involving several individuals performing different movements. Changes in classification accuracy as a function of different aspect angle between the direction in which the target faces and the line-of-sight of the radar nodes are analysed. Multiple data fusion methodologies are proposed, showing that significant improvement of the classification accuracy can be achieved when using separate classification at each node followed by a voting procedure to reach the final decision. This is beneficial especially at those aspect angles for which micro-Doppler detection is less favourable.

Feature Diversity for Optimized Human Micro-Doppler Classification Using Multistatic Radar

“…The former relates to the problem of classifying unarmed vs potentially armed personnel. In contrast to previous work [24][25][26], in these data there is no single target but two subjects who are simultaneously walking with similar speed and close in space, and one may (or not) be carrying a metallic pole representing a rifle. These data have been briefly analysed in [27], but without considering feature diversity and following the conventional approach of using the same features at each multistatic node.…”

“…Our previous work in [24][25][26][27] used a multistatic radar system to identify unarmed vs potentially armed personnel, initially in the simplified case of walking on the spot and then for actual realistic walking.…”

Centroid features for classification of armed/unarmed multiple personnel using multistatic human micro‐Doppler

“…Data with only a single person walking were recorded, with an equal number of repetitions with the person walking with free hands (hereafter referred to as "unarmed" case) and with the person holding a metallic pole representing a rifle (hereafter referred to as "armed" case). The pole was held using both hands in manner similar to that in which a real rifle would be held, and its size was comparable to that of a real rifle, hence this is expected to have a realistic effect on the walking gait of the person and on its micro-Doppler signature as shown in our previous works [17][18][19]. Two people took part In each of these recordings the individuals were walking towards the networked radar on the baseline, or perpendicular in the case of aspect angles 4 and 5, and no data were recorded for the individuals walking away from the radar.…”

“…Our previous work in [17,18] focused on the classification of armed vs unarmed personnel walking on the spot, investigating the effect of different aspect angles and of different ways of combining multistatic information on the classification accuracy. Empirical features extracted from the spectrograms, such as bandwidth, period, Doppler offset, and radar cross section (RCS) ratio, were used as input to the classifier and good results were achieved with classification accuracy of approximately 90% or above for the most favourable aspect angles and features.…”

Gait classification based on micro-Doppler features

“…In [6], the extraction of human features at long range based on micro-Doppler signals is studied. In [7], armed/unarmed personnel targets are distinguished based on multi-static micro-Doppler signatures. The method in [8] is capable of recognizing men and women.…”