This repository provides the codes and data used in our paper "Human Activity Recognition Based on Wearable Sensor Data: A Standardization of the State-of-the-Art", where we implement and evaluate several state-of-the-art approaches, ranging from handcrafted-based methods to convolutional neural networks. Also, we standardize a large number of datasets, which vary in terms of sampling rate, number of sensors, activities, and subjects.

• Scikit-learn
• Keras (Recommended version 2.1.2)
• Tensorflow (Recommended version 1.3.0)
• Python 3

1. Clone this repository
2. Run

   python <Catal2015|...|ChenXue2015>.py data/<SNOW|FNOW|LOTO|LOSO>/<MHEALTH|USCHAD|UTD-MHAD1_1s|UTD-MHAD2_1s|WHARF|WISDM>.npz

   For example

   python Catal2015.py data/LOSO/MHEALTH.npz

The raw signal provided by the original dataset was segmented by using a temporal sliding window of 5 seconds. Its format is (number of samples, 1, temporal window size, number of sensors)

Contributions to this repository are welcome. Examples of things you can contribute:

• Implementation of other methods. See template_hancrafted.py and template_convNets.py
• Accuracy Improvements.
• Reporting bugs.

The table below shows the mean accuracy achieved by the methods using the Leave-One-Subject-Out (LOSO) as validation protocol. The symbol 'x' denotes which was not possible to execute the method on the respective dataset.

Please cite our paper in your publications if it helps your research.

@article{Jordao:2018,
author    = {Artur Jordao,
Antonio Carlos Nazare,
Jessica Sena and
William Robson Schwartz},
title     = {Human Activity Recognition Based on Wearable Sensor Data: A Standardization of the State-of-the-Art},
journal   = {arXiv},
year      = {2018},
eprint    = {1806.05226},
}