TY - GEN
T1 - Radar-based human activity recognition using two-dimensional feature extraction
AU - Xiang, Fei
AU - Nie, Xiangfei
AU - Cui, Chang
AU - Nie, Wenliang
AU - Dong, Xichao
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - Aiming to solve the issues of high data dimensions with Frequency Modulated Continuous Wave (FMCW) radar image, slow extraction of feature information and complex classifiers in recognition algorithms, this paper propose a human activity recognition algorithm using two-dimensional feature extraction for FMCW radar. First, two-dimensional principal component analysis (2DPCA) is used to reduce the dimension of the radar Doppler-Time Map (DTM). On this basis, two-dimensional linear discriminant analysis (2DLDA) is used to extract the category feature information. Finally, K-Nearest Neighbor (KNN)classifier is used to achieve human activity recognition. The method proposed is verified by using the open dataset 'Radar signatures of human activities' created by the University of Glasgow. The human activity recognition rate reaches 96.40%. The results show that compared with the existing feature extraction algorithm in this field, the new method can effectively extract the key feature information of radar images and improve the recognition accuracy of human activity, meanwhile the running time is shortened.
AB - Aiming to solve the issues of high data dimensions with Frequency Modulated Continuous Wave (FMCW) radar image, slow extraction of feature information and complex classifiers in recognition algorithms, this paper propose a human activity recognition algorithm using two-dimensional feature extraction for FMCW radar. First, two-dimensional principal component analysis (2DPCA) is used to reduce the dimension of the radar Doppler-Time Map (DTM). On this basis, two-dimensional linear discriminant analysis (2DLDA) is used to extract the category feature information. Finally, K-Nearest Neighbor (KNN)classifier is used to achieve human activity recognition. The method proposed is verified by using the open dataset 'Radar signatures of human activities' created by the University of Glasgow. The human activity recognition rate reaches 96.40%. The results show that compared with the existing feature extraction algorithm in this field, the new method can effectively extract the key feature information of radar images and improve the recognition accuracy of human activity, meanwhile the running time is shortened.
KW - FMCW radar
KW - human activity recognition
KW - two dimensional-linear discriminant analysis
KW - two dimensional-principal component analysis
UR - http://www.scopus.com/inward/record.url?scp=85163199718&partnerID=8YFLogxK
U2 - 10.1109/ICCECE58074.2023.10135278
DO - 10.1109/ICCECE58074.2023.10135278
M3 - Conference contribution
AN - SCOPUS:85163199718
T3 - 2023 3rd International Conference on Consumer Electronics and Computer Engineering, ICCECE 2023
SP - 267
EP - 271
BT - 2023 3rd International Conference on Consumer Electronics and Computer Engineering, ICCECE 2023
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 3rd International Conference on Consumer Electronics and Computer Engineering, ICCECE 2023
Y2 - 6 January 2023 through 8 January 2023
ER -