Gait Calibration System

A gait calibration system for personalized walking speed estimation for people with Multiple Sclerosis (MS).

Prerequisites

• bokeh (0.12.4)
• Flask (0.12)
• Flask-Bootstrap (3.3.7.1)
• Flask-Script (2.0.5)
• Jinja2 (2.9.4)
• matplotlib (1.5.3)
• numpy (1.12.0)
• pandas (0.19.2)
• Pillow (4.0.0)
• pyparsing (2.1.10)
• python-dateutil (2.6.0)
• PyYAML (3.12)
• scikit-learn (0.18.1)
• scipy (0.18.1)
• simplejson (3.10.0)
• Werkzeug (0.11.15)

How to install

cd /path/to/ms-gait-calibrate
pip install -e .

How to calibrate via web browser

1. Start the flask server python run.py
2. Check the server IP address such as ifconfig in Linux
3. Open web browser then go to http://<ip-address>:5000
4. Go to Upload page and upload files
   
5. Go to Prepare page to prepare files for training
   
   
6. Go to Calibrate page, click check button to scan for the files used for training, give the name of the model, and click Calibrate button. Note that this process can take a while as it will calibrate a model.
   
   
7. Go to Estimate page, select the uploaded file and the calibrated model, and click Estimate button
   
   

Additional helper scripts

These are additional scripts used to preprocess some CSV files.

How to segment a large CSV file

This script is used to segment a very large CSV file collected in home environment into multiple CSV files containing 1-h of acceleration data. No preprocessing is applied here.

python scripts/segment_csv_data.py --csv_file /path/to/csv_file --sampling_rate 100 --body_location lower_back --position center_right --output_dir /path/to/save/output_dir

How to extract walks from CSV files in the directory and save in an NPY file

Extract walks from CSV files in the specified directory. The extracted walks will be applied with transform_orientation to transform from x, y, z into fwd, hor, ver axes based on body_location and position. These walks will be then stored in NPY files as a list of Acceleration objects.

python scripts/extract_walk.py --data_dir /path/to/directory/csv_files --output_dir /path/to/output_npy_files

How to estimate walking speed using the trained model

Estimate walking speeds from CSV and NPY files that contain walks.

python scripts/estimate_speed.py --input_file /path/to/input_csv_or_npy_file --model_file /path/to/model_file --output_dir /path/to/output_dir

Visualize estimated walking speeds with their corresponding acceleration data

Visualize estimated walking speeds and their corresponding acceleration data of all walks. Each output is saved in html generated using Bokeh.

python visualize/speed.py --acc_file /path/to/input_csv_or_npy_file --speed_file /path/to/speed_file --output_dir /path/to/output_dir

Reverse data collected from the device attached upside-down

python scripts/reverse_csv_data.py --input_csv_file /path/to/input_csv_file --output_csv_file /path/to/output_csv_file

Licence

• For academic and non-commercial use only
• For commercial use, please contact akara.spt@gmail.com and p.matthews@imperial.ac.uk
• Apache License 2.0