WISDM Smartphone and Smartwatch Activity and Biometrics Dataset

Overview

The WISDM Smartphone and Smartwatch Activity and Biometrics dataset contains raw motion sensor data collected by the WISDM Lab at Fordham University from Android smartphones and smartwatches during controlled activity sessions.[web:119][web:120][web:123]

Data Collection

• 51 subjects performed 18 different activities of daily living (e.g., walking, jogging, climbing stairs, sitting) for 3 minutes each, generating about 54 minutes of data per subject.[web:119][web:123]
• Each subject carried a smartphone in a front pants pocket and wore a smartwatch on the wrist while accelerometer and gyroscope data were recorded at 20 Hz.[web:119][web:123]
• Data are organized into four directories: phone-accelerometer, phone-gyroscope, watch-accelerometer, and watch-gyroscope, each containing one file per subject.[web:119]

Variables and Format

• Each record follows the format <subject-id, activity-code, timestamp, x, y, z>, storing tri-axial linear acceleration or angular velocity depending on the file.[web:119][web:123]
• Activity labels encode 18 activities, and subject identifiers allow building both activity recognition and user-identification (biometric) models.[web:119][web:123]
• In addition to raw time series, the authors provide windowed feature examples derived with a 10-second sliding window.[web:119][web:123]

Use Cases

• Supervised human activity recognition from smartphone and smartwatch motion sensors.[web:119][web:123][web:129]
• Motion-based biometric authentication using subject-labeled sensor sequences.[web:123][web:126]
• Benchmarking time-series and deep learning models for wearable IoT analytics.[web:119][web:123]