Sensing social interactions through BLE beacons and commercial mobile devices

Area of Science:

• Mobile social sensing
• Human-computer interaction
• Wearable sensing technologies

Background:

• Wearable devices offer high-resolution data for behavior analysis.
• Detecting social interactions via wearables is an emerging field.
• Privacy concerns necessitate focusing on non-sensitive data like physical proximity.

Purpose of the Study:

• To present the SocializeME framework for collecting proximity data and detecting social interactions using Bluetooth Low Energy (BLE) signals.
• To analyze the performance and limitations of BLE signals for social interaction detection.
• To release a comprehensive dataset for the research community.

Main Methods:

• Developed the SocializeME framework for heterogeneous mobile device data collection.
• Collected over 820,000 BLE signals across 11 hours, simulating real-world social interactions.
• Tested various configurations including user posture (standing, sitting) and device placement (hand, pockets).

Main Results:

• Evaluated BLE signal quality metrics like RSS, packet loss, and channel symmetry under different conditions.
• The SME-D algorithm achieved 81.56% accuracy and 84.7% F-score in detecting social interactions.
• Identified technical limitations and performance variations influenced by body and device positioning.

Conclusions:

• The SocializeME framework and SME-D algorithm demonstrate a viable approach for detecting social interactions using BLE signals.
• The collected dataset provides a valuable resource for advancing mobile social sensing research.
• Understanding the impact of physical configurations is crucial for improving the accuracy of wearable-based interaction detection.