Acoustic Gunshot Recognition
Effective Detection In High Noise Environments
Safety Dynamics specializes in the use of smart sensors for threat recognition and localization with its gunshot recognition product SENTRI (Sensor Enabled Neural Threat Recognition and Identification).
SENTRI employs a Dynamic Synapse Neural Network (DSNN) based on neurobiological principles of brain signal processing, and allows, like the human brain, accurate pattern recognition of acoustic signals even in the presence of high noise.
A SENTRI unit recognizes gunshot events in a radius of 600 feet and locates the event with a powerful pan/tilt/zoom video camera in less than one second. Events are localized to within +/- 1 degree or 10 feet of the event.
Features and Developments
The SENTRI Microphone Array is comprised of four aluminum alloy ¾-inch diameter arms that are anchored into a machined aluminum base. Attached to the ends of each arm is a Knowles MR-8406 condenser microphone. The microphone array has the ability to perform in a harsh environment. Vibration, weather, noise, and height of the microphones are all factors in the design and testing of the array unit. Safety Dynamics has trained DSNNs to recognize gunshots various calibers and not to respond to other loud noises, like truck and bus backfires; fireworks, and motorcycles. The SENTRI system was adapted to both: (1) discriminate the features that allow separate classification of handguns and other rifles, and (2) find the common set of features that allow handguns, machine guns, M1 attack rifles, M16 attack rifles, and multiple artillery to all be classified as belonging to the common set of firearms. This class system concept was fundamental to allowing the DSNN to then classify other common sets like security-breach or intrusion.
The SENTRI System can integrate with multi-sensor Technology, or traditional surveillance Technology such as video cameras, to detect a threat, often before a serious event occurs. Enabling of acoustic recognition takes place in a three-phase process. First, the targeted sound, such as an impulsive sound, is recorded and "trained" on the super-computer that isolates the sound signature amidst background noise. Second, the output parameter is uploaded onto the execution platform. Third, the run-time parameters are adjusted to adapt to requirements for distance, discrimination, and false positives. The SENTRI system consists of a TMS320C6713 DSK with the DSP running at 225 MHz. A 4-channel ADC daughter card connects to the DSK via the External Peripheral Interface. Two Analog Devices AD7655 16-bit, 1 MSPS A/D converters, perform data acquisition. Each A/D converter samples the analog signal coming from two Burr-Brown PGA2500 digitally controlled microphone preamplifiers.