HOT SAX: Efficiently Finding the Most Unusual Time Series Subsequence
ICDM '05 Proceedings of the Fifth IEEE International Conference on Data Mining
ACM Computing Surveys (CSUR)
| Hi-index | 0.00 |
Recently, wireless sensor networks have been proposed for assisted living and residential monitoring. In such networks, physiological sensors are used to monitor vital signs e.g. heartbeats, pulse rates, oxygen saturation of senior citizens. Sensor data is sent periodically via wireless links to a personal computer that analyzes the data. In this paper, we propose an anomaly detection scheme based on time series analysis that will allow the computer to determine whether a stream of real-time sensor data contains any abnormal heartbeats. If anomaly exists, that time series segment will be transmitted via the network to a physician so that he/she can further diagnose the problem and take appropriate actions. When tested against the heartbeat data readings stored at the MIT database, our ECG anomaly scheme is shown to have better performance than another scheme that has been recently proposed. Our scheme enjoys an accuracy rate that varies from 70-90% while the other scheme has an accuracy that varies from 40-70%.