Position-Information-Indexed Classifier for Improved Through-Wall Detection and Classification of Human Activities Using UWB Bio-Radar

Fugui Qi,Fulai Liang,Miao Liu,Hao Lv,Pengfei Wang,Huijun Xue,Jianqi Wang

DOI: https://doi.org/10.1109/lawp.2019.2893358

IF: 4.2

2019-01-01

IEEE Antennas and Wireless Propagation Letters

Abstract:Noncontact penetrating detection and classification of human activities based on micro-Doppler signatures (MDs) using ultrawideband (UWB) bio-radars are valuable tasks in various practical applications such as post-disaster search-and-rescue operations and urban military operations. However, for all classifiers, MD features of different-magnitude activities at different positions are likely to result in classification errors due to MD attenuation and confusions. This letter proposes a classifier improving method called position-information-indexed classifier (PIIC). It aims at enhancing the performance of various classifiers in terms of recognition and classification. This method fully exploits the position information acquired by UWB bio-radar to create a position-labeled modularized database of MD features. It also guides searching adaptively for optimal predict submodel of PIICs for activity classification at a random position. We report through-wall detection and classification experimental results related to five activities within a range of 6 m. These results, based on four typical classifiers, demonstrate that PIIC-based classifiers can avoid those classification errors in an effective manner. Moreover, all PIIC-based classifiers present a better classification performance with an average accuracy rise of 8.16% compared with those of overall-model-based classifiers. These performance evaluation experiments suggest that this method is strongly robust and stable, presenting wide applicability to various classifiers.