A data processing method for motion estimation considering network and sensor node loads

In a system that estimates human motions using acceleration sensors, a large amount of sensor data must be analyzed and processed appropriately. However, these systems present the following two problems: 1) advanced analysis and processing of sensor data are difficult because of the resource limitation of sensor nodes, 2) such analyses and processing burden the network because numerous sensor data are sent to the network. As described in this paper, we propose the method for sensor data analysis and processing using a host computer located in the neighborhood of sensor nodes (neighborhood host). This method is intended to achieve a good balance between reducing the network load and advanced sensor data analysis and processing. Moreover, this method incorporates reduction of the load to sensor nodes. We implement a prototype system to evaluate this method. This system estimates human motion using the acceleration sensor that generates sensor data frequently. We conduct some initial experiments using this prototype system. The experimentally obtained results show that the proposed method can reduce the network load by processing the acquired sensor data in the neighborhood of sensors. Moreover, power consumption of sensor nodes is reduced by controlling the time interval for data acquisition based on instructions of the neighborhood host.