Accelerometer Signals Classification for Activity Recognition: Smartphone-Based Accelerometer Dataset

Description

Rehabilitation and elderly monitoring for active ageing can benefit from Internet of Things (IoT) capabilities in particular for in-home treatments. In this task is proposed to recognize two the most important functions - Activity (AR) and Movement (MR). The major goal of this project – to detect one of the following patient states:

• idle,
• still,
• walking,
• running,
• going up/down the stairs

Based on the analysis the signals from smartphones.

Related Works The initial research is described in the paper: Enabling IoT for In-Home Rehabilitation: Accelerometer Signals Classification Methods for Activity and Movement Recognition[1].

Data

This is a dataset of accelerometric sample acquired for Activity Recognition (AR) algorithm. We collected raw measurements (one for each Cartesian axis: x, y, z). Since our algorithm required the framing of the signals, the frame duration has been set equal to 4 s. For the Activity Recognition (AR) case a set of about 14 hours has been employed. Acquisitions were performed by 8 users who kept the smartphones in four different positions and orientations: a) facing towards the user b) towards the opposite side, c) pointing up, d) pointing down.
The whole database, already exported in Matlab environment, is downloadable and available for possible further experiments and comparisons.

Reference publication:

1. I. Bisio, A. Delfino, F. Lavagetto, A. Sciarrone, “Enabling IoT for In-Home Rehabilitation: Accelerometer Signals Classification Methods for Activity and Movement Recognition" , IEEE Internet of Things Journal, doi: 10.1109/JIOT.2016.2628938.
2. Labelled AR /MR accelerometer data: http://www.dsp.dist.unige.it/images/download/AccDB.rar