Unsupervised habitual activity detection in accelerometer data

The activity of the user is one example of context information which can help computer applications respond better to the needs of the user in a seamless manner based on the situation without needing explicit instruction. With potential applications in many fields such as health-care, assisted living and sports, there has been considerable interest and work done in the area of activity recognition. Currently, these works have resulted in various successful approaches capable of recognizing common basic activities such as walking, sitting, standing and lying, mostly through supervised learning. However, supervised learning approach would be limited in that it requires labeled data for prior learning. It would be difficult to provide sufficient amounts of labeled data that is representative of free-living activities. To address these limitations, this research proposes motif discovery as an unsupervised activity recognition approach. Habitual activities would be detected by finding motifs, similar repeating subsequences within the collected accelerometer data. A 3D accelerometer sensor worn on the dominant arm is used to record, model and recognize different activities of daily living. The raw accelerometer data is then processed and discretized in order to perform motif discovery. Results have shown motif discovery to increase the performance in varying degrees (5-19%) depending on the discretization technique used