PAAL ADL Accelerometry dataset v2.0

Abstract

The PAAL ADL Accelerometry dataset (v2.0) has been acquired with a high-quality wearable multisensor device, the Empatica E4. In this dataset, among the signals collected by the sensors embedded in the Empatica E4, only the acceleration has been extracted to monitor the users performing different activities of daily living (ADLs). To promote the real-life acquisition procedure, subjects acted in their natural environment, with no instructions about how and for how long to perform each activity (other than a minimum time). The device was worn on the dominant hand. The dataset includes <strong>24 different ADLs</strong> performed using real objects. Each activity was repeated between 3 and 5 times (on average) by <strong>52 healthy subjects</strong>, characterized by a gender balance <strong>(26 women and 26 men)</strong>, and a large age range (between 18 and 77 years, mean = 44.08 years and standard deviation = 17.06 years). The PAAL ADL Accelerometry dataset (v2.0) is composed of three files: users.csv: each line contains (participant id, gender, age) of each user performing the ADLs in the dataset. <em>N.B: gender labels are 'man' and 'woman'</em>. ADLs.csv: each line contains (ADL id, ADL name) data.zip: folder with 6,072 files of accelerometer data of users performing ADLs. The name of each file indicates the name of the ADL, the user id and the repetition. Each row in the files represents the continuous gravitational force (g) applied to each of the three spacial dimensions (x, y, and z). The scale is limited to [-2g, +2g]. The sampling frequency is 32 Hz, with a resolution of 0.015 g (8 bit). More information about the format here.