AppWatchSecurity: Improving a Video Surveillance System by Integrating Smartwatch-Based Arousal Detection

Abstract

This paper proposes a prototype system that collaborates smartwatches with traditional video surveillance security. By combining concepts of user-centered design, ubiquitous wearable, psychophysiology and Internet of Things (IoTs), we present the upgraded video surveillance system where heart rate-based anomalies can automatically trigger the alarm. As a first prototype, the system was limited to library-like experimental setups and the anomaly was defined by arousal heart rate-unusually high heart beats. Using a smartwatch and a simple three-question questionnaire, we were able to collect referential arousal heart rate data from 25 healthy subjects together with their individual rating scores regarding three habit factors-smoking, drinking alcohol and eating fatty foods. According to our semi-quantitative user testings in a controlled library environment, the prototype was able to wirelessly connect and synchronize all devices, send the alarm, and perform real-time heart rate measurement as well as calculation. Based on confusion matrix evaluation, our anomaly detection gave promising results of 95% accuracy and 90% precision. However, major revision was required for the anomaly detection to cover unobserved factors, and there were serious usability problems regarding the smartwatch to be fixed.