This is (almost) a mirror of the ZJU-GaitAcc dataset webpage at Gang Pan's site.

Overview

ZJU-GaitAcc is the first publicly available dataset of gait acceleration series. It contains the gait acceleration series of 175 subjects. 153 of them are present in 2 sessions (Session 1 & 2), and 22 of them are in only 1 session (Session 0). For each subject, 6 records are included in one session, where every record contains 5 gait acceleration series simutalneously measured at the the right wrist, left upper arm, right side of pelvis, left thigh, and right ankle, respectively. 

The detailed specification of this dataset can be found in the citation, where an evaluation protocol for identification and verification is also provided.

According to our agreements with the volunteers, we should not record or release any specific personal information other than the gait acceleration. We can provide only some rough statistics of the datasets:

  • About 2/3 are male, 1/3 are female
  • Age: 16 ~ 40
  • Height: 1.5 ~ 1.9 m

How to get it?

Option 1: The database is publically available for non-commercial use. To obtain a copy of the database, please download the ZJUGaitAcc Database  release agreement, sign it after carefully reading, and send a scan to Gang Pan (gangpan AT gmail.com). We will soon reply to you.

Option 2: Alternatively, you can sign an ONLINE agreement. You will be redirected to the download page immediately after completion.

Citation

Please refer to the data as ZJU-GaitAcc and cite this paper:
  • Yuting Zhang, Gang Pan, Kui Jia, Minlong Lu, Yueming Wang, Zhaohui Wu, “Accelerometer-based Gait Recognition by Sparse Representation of Signature Points with Clusters”, IEEE Transactions on Cybernetics, vol. 45, no. 9, pp. 1864-1875, September 2015.

Development Toolkit

We provide a development toolkit to facilitate the usage of the ZJU-GaitAcc dataset. Currently, there are functions for

  • reading the acceleration records from the dataset

Dowload: zju-gaitacc_dev_20140924.tar.gz

Format

To make the dataset readable without any specific (commercial) software, we stored the data in text files, which are organized in a directory tree. 
The folder structure of this dataset:
data_dir/session_(session id)/subj_(subject id)/rec_(record id)/
Inside a fold: 
  • (body location id).txt
  • cycles.txt
  • useful.txt
Body location id indicates:
  • 1 - the right wrist
  • 2 - the left upper arm
  • 3 - the right side of pelvis
  • 4 - the left thigh
  • 5 - the right ankle
Format of (body location id).txt:
  • 3-by-N matrix in ASCII-delimited format of MATLAB.
  • one line for one row
  • elements in the same row are separated by commas
  • 3 rows for the x,y,z components of acceleration
  • N columns for N acceleration samples
  • frequency: 100Hz
  • unit: g -- the gravitational constant
Format of cycles.txt:
  • 1-by-C matrix in ASCII-delimited format of MATLAB.
  • elements are separated by commas
  • each element indicates one cycle borders
  • C elements for C cycle borders in sequential order, where the cycle number is (C − 1)
Format of useful.txt:
  • 1-by-2 matrix in ASCII-delimited format of MATLAB.
  • elements are separated by commas
  • the first element indicates the start point of the useful segment
  • the second element indicates the end point of the useful segment

Examples

The following figure (get the PDF version) shows a record of gait acceleration, and the annotations for it.

Sample first halfsample second half

Disclaimer

THIS DATA SET IS PROVIDED "AS IS" AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.