Acoustically Cued Video Surveillance
VAUDEO is the concept of using the combination of Acoustic Vector Sensors and video cameras. Microflown AVISA’s Acoustic Vector Sensors can detect and localise incidents by their acoustic signature, automatically cueing PTZ (pan-tilt-zoom) cameras to incidents of interest. Types of incidents would be gunshots, screams, crashes, bangs, aggressive shouting behaviour, alarms, etc. The benefit of this system is the reduced workload on video surveillance operators, with potential security issues identified more quickly and more often than is currently possible with human operators alone.
Although not widely known, at any point in space, a sound field can be described completely by its two physical quantities. The scalar value sound pressure is well known and well measured. But the other quantity in acoustics, the vector value acoustic particle velocity, only recently became a directly measurable quantity with the invention of the Microflown sensor. Our sensors are the only commercially available sensors in the world, capable of measuring acoustic particle velocity. Acoustic Vector Sensors provide information on the direction of arrival (DOA) of a sound source. With our sensor, the direction of a sound source can be directly measured with the information being used to cue cameras or other narrow aperture sensors.
Microflown’s Acoustic Vector Sensors can be integrated with PTZ cameras in and around airports, train stations and other public areas where security is a concern. When the Microflown sensor detects an abnormal acoustic pattern, associated with some kind of event or incident, the PTZ camera is steered toward the event, capturing the potential security risk immediately. Types of incidents which would trigger the system are: gunshots, explosions, screams, banging, running, aggressive shouting, crashes, alarms, etc. As long as the tannoy positions are known, airport and rail announcements can be ignored as acoustic events.
- The video cameras are (semi)autonomously controlled, assisting video operators in identifying unusual behaviour or incidents.
- The acoustic vector sensor is small enough that the camera it is integrated with would not be easily distinguishable from a standard PTZ camera.
- The system could be combined with pattern recognition software (video) further enhancing the value of the system.
- Because of the directional nature of the acoustic vector sensor, it is possible to listen in very specific directions, thus should a situation be identified where it would be useful to listen to, say a conversation, the operator would be able to covertly listen to the conversation, with no noise intrusion from other directions.
This system has many benefits over existing acoustic detection systems. Conventional microphones measure the acoustic pressure, but the Microflown sensor measures the acoustic particle velocity: the vibration of the air. Measuring the particle velocity in all three directions, we measure the direction of the source directly. Pressure-based source localization system use the phase difference between the pressure at different locations to determine the source location. Because of this working principle, the arrays are large and their frequency range is limited. The patented Microflown sensor does not have these drawbacks. The sensor elements take up only a few cubic millimeters of space, and the source location can be determined throughout the audible frequency range.
Traditional acoustic detection systems have trouble with reflections, because they rely on time differences to compute directions. Reflections do not interfere with Acoustic Vector Sensors in the same way, so mounting a camera system on a wall or on a ceiling is viable, whereas this configuration would not be feasible with a microphone based system.
In essence, our technology allows video cameras to act more like humans. When we hear something we localise it, look at it and decide if it is interesting and warrants further inspection. At the moment cameras are deaf and rely on looking in the right place at the right time. For example, if you wanted to protect an area using humans, which we regularly do using security guards, it is unlikely you would employ deaf security guards.
Published at the 2010 Workshops on Database and Expert Systems Applications, J. Kotus, B. Kunka, A. Czyżewski, P. Szczuko, P. Dalka, R. Rybacki
Published at the Advances in Soft Computing 2010, J. Kotus, A. Czyzewski
Published at the Information Technology (ICIT) 2010, A. Czyżewski, J. Kotus
(external link to Gdansk University of Technology)