Machine Learning
Fine-Grained Activity Recognition by Aggregating Abstract Object Usage
ISWC '05 Proceedings of the Ninth IEEE International Symposium on Wearable Computers
A sensory grammar for inferring behaviors in sensor networks
Proceedings of the 5th international conference on Information processing in sensor networks
| Hi-index | 0.00 |
To develop an energy-saving environment control system, we propose wireless sensor/actuator networks which classify the user's behaviors for environment control such as lighting and configure the network according to sensor node selection. In our system, a wireless remote control node monitors the user's environment control actions and occupancy sensor networks detect the user's movement simultaneously. The system learns the relationship among the responses of the remote control node and the occupancy sensor networks to classify the user's behaviors only with the occupancy sensor networks. The system chooses informative sensor nodes for the behavior classification based on the information gain criterion. These chosen nodes have high sensing cost. Sensor network routing is controlled based on the sensing cost and communication cost metric. In the resultant sensor networks, the sensing performance is the same as that in the original network, but the resources are successfully allocated to the nodes. In addition, less informative and redundant nodes are identified. We demonstrate tracking environment control behaviors and sensor node selection using the sensor/actuator networks testbed.