Behavior identification of search and rescue dogs based on inertial sensors
Shintaro Narisada, Sota Mashiko, Shunta Shimizu, Yu OHORI, Keisuke Sugawara, Sakuma Shumpei, Ichinari Sato, Yohei Ueki, Ryunosuke Hamada, Shumpei Yamaguchi, Tatsuya HOSHI, Kazunori Ohno, Ryo Yoshinaka, Ayumi Shinohara, Takeshi Tokuyama · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2017
We propose an underlying system that can infer and visualize search and rescue (SAR) dogs’ behavior. The system is aimed at identifying “run”, “walk”, “stop”, “sniff” and “bark” behaviors of SAR dogs robustly from inertial sensors data, and visualizing the results for the users. In the system, we apply Short-Time Fourier Transform (STFT) to the sensors data, and use a random forest algorithm for learning investigation activities of SAR dogs. We performed an experiment on our system and got the results that some behaviors can be identified precisely. We also developed an on-line visualization system for streaming data of behavior probabilities.