Self-organizing map analysis of sensor networks for human movement tracking

This article describes sensor networks for human movement tracking and data analysis that use self-organizing maps (SOMs) in indoor environments. These sensor networks consist of motion sensor nodes. In our system, the relationship between neighboring nodes is determined by the similarity of the sensor responses. We use sensor nodes with passive binary infrared motion sensors in experiments designed to track human movement, and examine the effectiveness of SOM analysis. An SOM enables us to acquire the relationships between neighboring nodes without the need for positioning techniques. We describe two types of input vectors for the SOM algorithm. First, we simply consider the order relation of the sensor responses used as input vectors into the SOM. Next, we examine the intervals of the sensor responses for the input vectors. We conclude that SOM analysis has great potential for human movement tracking applications.