Gait Re-Identification using IMU Sensor Data

This project was done at EHWA PAI lab.

Project Overview

The goal of the project was to re-identify criminals using CCTV and gait information.

We analyzed gait patterns using IMU sensor data and developed a method to determine whether two gait data sequences belong to the same person. The project took two IMU data inputs (3D joint coordinate sequences) and transformed them into 128-dimensional feature vectors through FFT transformation and CNN. We then used Contrastive Learning to calculate the similarity between vectors and, based on the final similarity score, determined whether the two data points belong to the same person. I was responsible for data preprocessing, surveys, and modifying the model architecture.

gait-re-id