Objective: A patients activity patterns can be informative about her/his health status. Traditionally, this type of information has been gathered using patient self-report. However, these subjective self-report data can suffer from bias, and the surv
eys can become burdensome to patients over long time periods. Smartphones offer a unique opportunity to address these challenges. The smartphone has built-in sensors that can be programmed to collect data objectively, unobtrusively, and continuously. Due to their widespread adoption, smartphones are also accessible to most of the population. A main challenge in smartphone-based activity recognition is in extracting information optimally from multiple sensors to identify different activities. Materials and Methods: We analyze data collected by two sensors in the phone, the accelerometer and gyroscope, which measure the phones acceleration and angular velocity, respectively. We propose an extension to the movelet method that jointly incorporates both data types. We apply this proposed method to a dataset we collected and compare the joint-sensor results to those from using each sensor separately. Results: The findings show that the joint-sensor method reduces errors of the gyroscope-only method in distinguishing between standing and sitting. Also, the joint-sensor method reduces errors of the accelerometer-only method in classifying vigorous activities, such as walking, ascending stairs, and descending stairs. Conclusion: Across activities, for the given method, combining data from the two sensors performs as well as or better than using data from a single sensor. The method is transparent, personalized to the individual user, and requires less training data than competitor methods.
