Sensor Platform for Healthcare in a Residential Environment (SPHERE) Physical Activity Sensor Dataset
Uses classification algorithms such as Support Vector Machine (SVM), K-Nearest Neighbour (KNN), and Long Short-Term Memory (LSTM) to detect fall activities with data gotten from wearables. The sphere.WEAR.json file contains the raw data collected from the accelerometers.
The SPHERE dataset contains 20 physical activities performed by 2 actors wearing 2 wearables each. One wearable on the wrist and the other on the waist region. Each wearable serves as an accelerometer sensor and is uniquely identified with a unique id. The activities are:
1 - Fall 2 - Fall while wiggling hands 3 - Pick something on the floor 4 - Descend staircase 5 - Climb staircase 6 - Jump 7 - Kneel down 8 - Lie down 9 - Lie still on bed 10 - Drop something on the floor 11 - Raise hand then put it down 12 - Run up stairs 13 - Run on flat ground 14 - Sit down properly 15 - Sit still 16 - Stand still 17 - Dive on the bed 18 - Dive on the sofa 19 - Try to keep balance 20 - Walk on flat ground
The raw data collected by the accelerometers is stored in Json format. Each item contains a unique id ($oid), an accelerometer id (uid), a time stamp ($date), a list (e) that contains six sets of acceleration values in three directions (x,y,z) and the time offsets (t), as well as 5 other attributes that are irrelevant for this study.