Sound Event Detector

This platform provides a unified API (Application Programming Interface) that covers the necessary (cloud) network communication, data synchronization and accounting for an intelligent home management. We use cloud technology for data storage and to provide maximum network security.

Vision: The company JALUD Embedded s.r.o. (JE) has developed a prototype of Sound Event Detector (SED) that brings a new approach to the problematics of sound detecting.
Thoughts behind: The primary goal of the SED is to allow an immediate reaction of rescue or security team on the dangerous situation (gunshot, car crash, bomb blast, human scream etc.)

This solution enables energy efficient sound detecting running on devices with energy efficient processors and low-cost processors. The great advantage of our solution is that it faces no ethic issues as the sound data are processed in a real-time on the target device.

It offers amazing opportunities to spread devices of this type into various segments. SED offers significant enhancement to the existing monitoring systems, it can provide real time analysis of sound streams, detection and categorization of selected impulsive sound event.

The frequent public areas are usually covered by monitoring camera systems for getting visual information, but there is no possibility to monitor all camera systems concurrently. Moreover, the cameras cannot see everything and have a lot of blind spots.

Terrorism across Europe has killed 10,537 people in 18,803 reported attacks. Frequency of such attacks in EU has rapidly increased during last years.

Terrorism in EU

In case of gunshot injury the crucial is the time before the Ambulance get on the site. The period from the time of the injury till wounded will get an appropriate medical treatment takes usually about some time needed before the Ambulance is contacted, after that the operator has 5 minute to assess the current situation and the time of Ambulance arrival itself. Every second that can be saved matters, not only to improve the chances of survival, but also affects the quality of life for the future and recovery time.

Global video surveillance market forecast

Immediate reaction saves lives

Gun murders in the World

Immediate reaction saves lives

HW requirements:

Highly efficient and optimized algorithms that can run on low power ARM CPU as well as on MCUs

Immediate reaction saves lives

Technology and reference

Event classification is based on the most advanced ML&AI techniques like Convolutional and Recurrent Neural Networks, model responsible for detecting required sound event has been computed on 200GB of sound data.

JE has developed a system capable to detect the sound of a broken glass with 97% F-measure success rate and successfully delivered the SW to Jablotron Alarms a.s.

Immediate reaction saves lives