featurelearning

Feature Learning for Accelerometer based Gait Recognition

Code repository of paper: Feature Learning for Accelerometer based Gait Recognition, submitted to Journal

Used datasets

• Dataset used for evaluations

  • [ZJU-GaitAcc] - http://www.ytzhang.net/datasets/zju-gaitacc/
    • session 0 - 22 subjects
    • session 1 - 153 subjects
    • session2 - 153 subjects

• Dataset used for feature learning *[IDNet] - http://signet.dei.unipd.it/research/human-sensing/ * 50 subjects with various number of sessions * Resampled at 100 Hz

Segmentation

• FRAME-based: length = 128 samples (Sampling frequency 100 Hz)

Features

• RAW - use raw accelerometer data as features - 3 x 128 = 384 (ax - ay - az)

• SUPERVISED feature extraction

  • Convolutional end-to-end model (FCN) trained on IDNet

• UNSUPERVISED feature extraction - autoencoders

  • Fully Convolutional (FCN) autoencoder trained on IDNet

Verification - based on a single gait segment (FRAME)

• OneClass SVM (OCSVM) for each user
• Two protocols:
  • SAME-DAY: using data from a single session
  • CROSS-DAY: training - session 1, testing - session 2

Code

• The main_gait.py python file contains the necessary code to run an experiment.
• The TRAINED_MODELS folder contains the end-to-end models as well as the autoencoders trained in different settings (with or without augmentation)
• The plots folder contains the source codes necessary to create the figures in the paper.
• The util folder contains the following:
  • augment_data.py - functions used for data augmentation
  • autoencoder.py - code for training and evaluating autoencoders
  • classification.py - code for user identification (classification)
  • fcn.py - Fully Convolutional end-to-end model
  • model.py - code for training the end-to-end model
  • normalization.py - functions for data normalization
  • oneclass.py - code for user verification
  • plot.py - different utility plots
  • utils.py - utility functions