TY - CONF
T1 - Gait recognition using FMCW radar and temporal convolutional deep neural netowrks
AU - Addabbo, Pia
AU - Bernardi, Mario Luca
AU - Biondi, Filippo
AU - Cimitile, Marta
AU - Clemente, Carmine
AU - Orlando, Danilo
N1 - © 2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting /republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
PY - 2020/7/5
Y1 - 2020/7/5
N2 - The capability of human identification in specific scenarios and in a quickly and accurately manner, is a critical aspect in various surveillance applications. In particular, in this context, classical surveillance systems are based on video cameras, requiring high computational/storing resources, which are very sensitive to light and weather conditions. In this paper, an efficient classifier based on deep learning is used for the purpose of identifying individuals features by resorting to the micro-Doppler data extracted from low-power frequency-modulated continuous-wave radar measurements. Results obtained through the application of a deep temporal convolutional neural networks confirms the applicability of deep learning to the problem at hand. Best obtained identification accuracy is 0.949 with an F-measure of 0.88 using a temporal window of four second.
AB - The capability of human identification in specific scenarios and in a quickly and accurately manner, is a critical aspect in various surveillance applications. In particular, in this context, classical surveillance systems are based on video cameras, requiring high computational/storing resources, which are very sensitive to light and weather conditions. In this paper, an efficient classifier based on deep learning is used for the purpose of identifying individuals features by resorting to the micro-Doppler data extracted from low-power frequency-modulated continuous-wave radar measurements. Results obtained through the application of a deep temporal convolutional neural networks confirms the applicability of deep learning to the problem at hand. Best obtained identification accuracy is 0.949 with an F-measure of 0.88 using a temporal window of four second.
KW - deep learning
KW - gait recognition
KW - low-power radar
KW - micro doppler
UR - https://ieee-aess.org/conference/2020-ieee-international-workshop-metrology-aerospace
M3 - Paper
T2 - IEEE International Workshop on Metrology for Aerospace 2020
Y2 - 22 June 2020 through 24 June 2020
ER -