Behavior recognition and disaster detection by the abnormal analysis using SVM for ERESS

Many lives have been lost for many years in all parts of the world by the sudden disasters such as fire and terrorism. Main causes of the damage expansion in these disasters include the escape delay of the evacuees. To support evacuation safely and quickly is one of the effective measures to reduce the victim by the disaster. So, we develop the system named Emergency Rescue Evacuation Support System (ERESS) as a system to detect a disaster using handheld terminals quickly and to guide to safety zone. This system automatically detects a disaster by analysis of the information of terminal holders, and sharing information with neighboring terminals. This paper focuses on behavior analysis of terminal holders and disaster detection which is big characteristics of ERESS. We propose an activity recognition using Support Vector Machine (SVM) and a disaster detection method by the abnormal analysis using SVM. The results of the performance evaluation by two experiments show the validity of the proposed method.