A Design of Acoustic Security System in Near Field based on Paired Microphones and Automatic Video Camera

Many conventional security systems use visual information from a video camera. However, these systems may not be able to acquire the important scenes in the blind area of the video camera. Acoustic security systems can support conventional visual security systems with acoustic events detection. In our research, we focused on acoustic security systems in the near field, and we designed a prototype of a three-dimensional acoustic security system using paired microphones and an automatic video camera. This system detects a sound event and automatically focuses the video camera on the detected sound source in real time. Conventional automatic video camera steering systems localize a sound source assuming it is in the far field. Moreover, they detect only speech because this type of system was developed for videoconferencing. However, detection errors increase in the near field. Thus, we designed a prototype acoustic security system that robustly detects a sound event, localizes the sound source in the near field, and then automatically steers the video camera to the detected sound event in real time. We carried out evaluation experiment in a real office environment. An evaluation experiment in a real office environment showed that the proposed security system could robustly detect a sound event and quickly steer a video camera toward it.