Personal Identification Using Radar Measurements of Motions  

Background

Personal identification is important not only for security but also for healthcare and monitoring. In homes or care facilities where multiple people live together, identifying who performed an activity can help evaluate individual activity levels, daily routines, gait function, and changes in physical function.

Conventional methods include face recognition, fingerprints, iris recognition, wearable devices, and camera-based gait recognition. However, continuous use in daily environments can be limited by privacy concerns, the burden of wearing devices, lighting conditions, and camera installation constraints.

Radar-based personal identification can obtain velocity patterns from walking, standing up, sitting down, picking up an object, and other daily motions without requiring wearable devices. This enables privacy-preserving personal identification and individual healthcare monitoring without visual images.

Research Objectives

This research aims to extract individual differences from daily motions using time-velocity distributions, micro-Doppler features, motion duration, and velocity changes obtained by Doppler radar and micro-Doppler radar.

Main topics include:

• Personal identification based on gait motions

• Personal identification based on sit-to-stand and stand-to-sit motions

• Analysis of individual differences in daily motions such as picking up an object

• Feature extraction from time-velocity distributions and micro-Doppler images

• Personal identification using machine learning and deep learning

• Privacy-preserving individual monitoring and healthcare sensing

Related Work

Our previous work has investigated radar-based personal identification by analyzing individual differences in gait and daily human motions using Doppler radar and micro-Doppler radar.

During walking, velocity patterns generated by the legs and trunk differ among individuals. Motions such as standing up from a chair, sitting down, and picking up an object also include individual differences in motion speed, posture transition, and body usage. By extracting these features from radar signals, this research aims to realize privacy-preserving personal identification and individual physical-function monitoring without visual images.

In particular, our work demonstrated the feasibility of personal identification using sit-to-stand and stand-to-sit motions. To the best of our knowledge, this was among the first studies, including non-radar sensing fields, to clarify the possibility of person identification based on sit-to-stand motions.

Selected Publications

• Keitaro Shioiri and Kenshi Saho, "Exploration of Effective Time-Velocity Distribution for Doppler-Radar-Based Personal Gait Identification," Sensors, vol. 23, Article no. 604, January 2023.

• Kenshi Saho and Keitaro Shioiri, "Personal Identification Based on the Picking-Up Movement Measured Using a Doppler Radar," IEEJ Transactions on Electrical and Electronic Engineering, vol. 16, pp. 1286–1288, September 2021. 

• Kenshi Saho, Keitaro Shioiri, and Keisuke Inuzuka, "Accurate Person Identification Based on Combined Sit-to-Stand and Stand-to-Sit Movements Measured Using Doppler Radars," IEEE Sensors Journal, Vol. 21, pp. 4563-4570, February 2021.

• Kenshi Saho, Keisuke Inuzuka,and Keitaro Shioiri, "Person Identification Based on Micro-Doppler Signatures of Sit-To-Stand and Stand-To-Sit Movements using a Convolutional Neural Network," IEEE Sensors Letters, vol. 4, Article no.3500304, March 2020.

Keywords

radar-based personal identification, gait identification, gait recognition, micro-Doppler radar, Doppler radar, time-velocity distribution, sit-to-stand motion, stand-to-sit motion, daily activity recognition, non-contact personal identification, privacy-preserving sensing, human sensing, healthcare monitoring, radar signal processing