Preliminary investigation of indoor event detection using active sound probing

摘要

Event detection of indoor objects such as doors has a variety of application including intruder detection, HVAC control, surveillance of independently living elderly persons. Therefore, numerous researches have been done on this topic by the UbiComp research community. In this research, we propose a method to accurately detect events in indoor everyday objects such as doors and windows without using distributed ubiquitous sensors. In our approach, we perform event recognition conjoining the Doppler shift that occurs by the moving indoor objects and the acoustic characteristics of the environment acquired by the impulse response. Analysis of the time series data of the Doppler shift, using directional high frequency sine wave to acquire information of the direction of the occurred event, using stereo microphones to record the reflected wave can be stated as the specialties in this research. Also, by incorporating the knowledge about the state transitions of an object in to a recognition model, we will try to increase the accuracy of our recognitions.