With the growing age of the baby-boomer generation and the rising cost of healthcare, it is becoming exceedingly difficult to give proper care and daily supervision to the elderly. The ability to detect anomalies in a subject's behavioral patterns based on minimal sensor data, will allow for a low cost sensor network to be placed into the subject's environment to provide this supervision. In this paper, we will discuss our approach for anomaly detection using a clustering algorithm on a minimal spanning tree and why this approach is more robust than statistical analysis. We will discuss the possible applications for such a system and will show that our method has a high success rate on both field and synthetic data.