SmartAberration: An annotated multimodal smartphone IMU dataset for aberration detection

Abstract

Smartphone IMU sensors are being used more and more to track different aspects of human activity during everyday life. The sensors built into smartphones yield objective quantifications of normal and anomalous behaviors. The dataset, referred to as SmartAberration, consists of motion data collected from five participants, each engaging in 10 activities, for a total of 50 activity recordings. The data set contains both normal activities and ones that are indicative of potentially suspicious actions, including running in off-limits areas, punching, or kicking. These are of specific interest because they have the potential to indicate security threats or interruptions of normal patterns of movement. The SmartAberration dataset records motion signals derived from the smartphone IMU data, such as parameters of acceleration, velocity, and angular velocity. Each sample is labeled in terms of the actual activity being undertaken, allowing behaviors to be classified based on their signal. This publicly accessible dataset provides a wide range of opportunities for researchers to utilize intelligent data analysis methods in order to automatically discover and type human activities. It provides a singular platform for creating systems that can continuously monitor and evaluate normal and abnormal behaviors in real-time, without the need for self-reporting or manual observation. This technology could make security, health monitoring, and behavioral analysis applications more powerful, enabling proactive interventions as well as enhanced comprehension of human activity patterns.