Multidimensional Motif Discovery in Physiological and Biomedical Time Series Data

Providing personalized diagnosis and therapy requires monitoring patient activity using various body sensors. Sensor data generated during personalized exercises or tasks may be too specific or inadequate to be reviewed and evaluated using supervised methods such as classification. We propose multidimensional time series motif discovery as a means for patient activity monitoring, since such motifs can capture repeating patterns across multiple dimensions of the data, and can serve as conformance indicators. Previous studies pertaining to mining multidimensional motifs have proposed offline algorithms and lack the capability of processing and mining motifs from multiple dimensions concurrently. In this paper, we propose an efficient approach to multidimensional motif discovery in body sensor generated time series data for monitoring performance of patients during therapy. We present two alternative models for multidimensional motifs based on motif co-occurrences and temporal ordering among motifs across multiple dimensions, with detailed formulation of the concepts proposed. The proposed method uses an efficient hashing based record to enable speedy update and retrieval of motif sets, and identification of multidimensional motifs. We also demonstrate the utility of this approach by applying it to synthetic as well as human motion data captured by on-body sensors in both unsupervised motif discovery and content based query resolution tasks. The approach is shown to be effective for (a) tracking repetitions during therapy sessions; (b) finding naturally occurring patterns; and (c) query resolution.