Our engineers go to extreme measures to make sure our Sound Packs™ are as good as they can possibly be and we often get asked how we achieve the performance levels we do. In conjunction with our advanced patent pending sound recognition technology used in our CoreTrainer™ & CoreLogger™ technology, we also go to great lengths to collect 1,000’s of hours of audio recordings. These are collected in a wide range of environments using a variety of different recording equipment, from security camera microphones to external microphones. We also collect many 100's of samples of the specific sounds we are interested in which enables our system to characterise the sound through the profile to be stored in our Sound Packs™. To show you the length our engineers go through to make sure our products are of the highest quality, we want to share some of the stages of generating one of our Security Sound Packs™, Glass Break.

The first stage our engineers undertake is to collect a range of recordings for glass windows being broken (intrusion/perimeter detection sounds) for laminate, plate, wired & tempered glass types commonly found in commercial and residential areas. Different thicknesses and sizes of glass, as well as, where appropriate, a range of different ‘implements’ are used to generate the recordings. A variety of security cameras with built in microphones, external microphones and other devices are then used to capture the audio.

The short video below is just a very small proportion of the glass we smashed in the process of generating our Glass Break Sound Pack™, as well as the separate tests we conducted to validate its performance.

Our engineers made 100’s of recordings of different types of glass being broken while also evaluating commercial standalone glass break detectors as a comparison to our Sound Pack’s™ performance. What you are left with, after many days of continuous glass breaking, are many boxes full of broken glass.

These recordings are then grouped by type of glass and then the many days’ worth of recordings are fed into our CoreTrainer™ running on a high performance super computer. This super computer is made up of 3,584 individual computer processors with roughly 30 Terra Flops per second worth of computing power at CoreTrainer’s™ control. The CoreTrainer™ chews through the audio recordings until it can summarise the sounds in an extremely concise light weight set of parameters we call a Sound Pack™. The Sound Pack™ is then coupled with our classification engine, CoreLogger™. The CoreLogger™ technology is designed to run on a modern security camera or on a video management system and due to the accuracy and compactness of the description stored in the Sound Pack™ it does so without requiring large amounts of a device’s resources.

Our engineers take this Sound Pack™ and test it against 1,000s of hours of challenging audio from every conceivable acoustic environment the Sound Pack™ is required to operate under, as well as other similar sounds, to check that it can operate effectively and reliably with minimum false positives. Finally, the Sound Pack™ is then tested on a real end device to ensure a complete circle of testing and quality.

All of this effort creates a Sound Pack™ that is not only functional, but that can be used to protect and alert the broader security solution to the presence of an intrusion, or attempted intrusion, as and when it is happening in a reliable and dependable manner.

Would you like to know more about developing Sound Packs™? Read about how we developed our Gunshot Sound Pack.

Application Areas

Please use the following links to find about how Audio Analytic's sound recognition and classification software, including other Sound Packs, can help enhance your products or services.

• Audio Analytics for Camera, DVR or NVR manufacturers
• Audio Analytics for entertainment and communication devices
• Sound Analytics for acoustics and measurement