Feature Learning for Accelerometer based Gait Recognition

TL;DR

This paper compares supervised and unsupervised deep learning methods, including autoencoders and end-to-end models, for gait recognition using accelerometer data, showing that autoencoders perform comparably to discriminative models.

Contribution

It provides a comparative analysis of feature learning approaches for gait recognition, highlighting the effectiveness of autoencoders and fully convolutional models.

Findings

Autoencoders perform close to end-to-end models in feature learning.

Fully convolutional models learn effective features regardless of training strategy.

Autoencoders and end-to-end models produce similar quality features.

Abstract

Recent advances in pattern matching, such as speech or object recognition support the viability of feature learning with deep learning solutions for gait recognition. Past papers have evaluated deep neural networks trained in a supervised manner for this task. In this work, we investigated both supervised and unsupervised approaches. Feature extractors using similar architectures incorporated into end-to-end models and autoencoders were compared based on their ability of learning good representations for a gait verification system. Both feature extractors were trained on the IDNet dataset then used for feature extraction on the ZJU-GaitAccel dataset. Results show that autoencoders are very close to discriminative end-to-end models with regards to their feature learning ability and that fully convolutional models are able to learn good feature representations, regardless of the training strategy.