イヌのリアルタイム情動推定システムの開発

抄録

Search and rescue dogs are widely used in disaster situations. The efficiency of their operations can be improved if the dogs' inner states or motivation to search are estimated. We developed a real-time emotion estimation system for canines based on a measured electrocardiogram. Interval time of heart beats were measured by our canine suit equipped with an electrocardiographic device. Temporal variability of the measured intervals, i.e., heart rate variability, was calculated, then processed as input to a classifier of the canine's emotional state. An estimate of the emotional state is presented in a real-time graph, and updated in every five seconds. The proposed system was evaluated in an outdoor experiment and we confirmed that it can be operated in real-time.