Smartphone Inertial Sensors and CCTV Facial Dataset for Human identification
Description
This multimodal biometric dataset combines smartphone inertial gait signals with facial recordings for human identification research. Contains data from 400 participants (39% female, 61% male, ages 18-49) performing 10 walking trials each at Koya University, Kurdistan Region, Iraq. Dataset includes 4,000 inertial sensor recordings (accelerometer, gyroscope, magnetometer at 30 Hz) capturing distinctive gait signatures with high consistency across trials. Additionally provides 20,000 facial image frames from 100 selected participants using dual-camera CCTV setup: frontal and side views with minimal motion blur. Synchronized inertial-visual data enables multimodal fusion experiments and temporal alignment studies. Collection methodology ensures realistic real-world conditions with natural smartphone holding positions, making this dataset valuable for practical biometric system development.
Files
Steps to reproduce
Data collection used Samsung Galaxy A53 smartphones and dual Dahua IPC-HFW2849S-S-IL 8MP cameras in a 12-meter indoor corridor at Koya University. 400 participants (18-49 years, mixed gender) each performed 10 walking trials while holding the smartphone naturally in their dominant hand. Inertial sensors (gyroscope, magnetometer) recorded at 30 Hz simultaneously with dual-camera video capture (frontal CAM2, side CAM1) for 9-13 seconds per trial, generating timestamped CSV files and AVI recordings. Processing involved extracting 10 frames per camera view per trial from videos and applying filtering plus gait cycle extraction to inertial data. Due to storage constraints, visual data was processed for 100 selected participants, yielding 4,000 CSV files for all participants and 20,000 JPEG frames (100 participants × 10 trials × 2 views × 10 frames). Standard video processing tools and signal processing techniques were used for data filtering and frame extraction.