Improving the Efficiency of City Surveillance with Acoustic Detection

The City of Plzeň operates an extensive camera system with more than 300 cameras, which requires high personnel costs and is prone to human operator errors. The main problem was the operators’ workload and the limited ability to identify real incidents in time, reducing surveillance efficiency and increasing costs.

Acoustic detectors (SED) were deployed to automatically recognize high-risk sound events (e.g., shouting, gunshots, breaking glass). Even when covering only 18% of the camera network, the SEDs managed to detect nearly as many incidents as the entire team of human operators. The solution is fully automated, easily scalable, and integrates into the city’s existing surveillance systems.

Operators of the city surveillance center, municipal police, security forces.

6

Convolutional neural networks, edge AI, embedded systems, acoustic analysis, API integration.

Head of functional/operational unit

Full implementation

Regular servicing is provided by the supplier (JALUD Embedded).

Surveillance center operators, IT department technical support.

Integration into the existing VMS, minimal IT infrastructure increase, low operating costs.

Increased efficiency of city surveillance
Reduced operator workload
Timely and reliable incident detection
Enhanced citizen safety without infringing on privacy

61 incidents detected acoustically (vs. 89 incidents by human operators)
False alarms reduced: 5% (SED) vs. 34% (video analysis)
Operating costs reduced by up to 95% compared to human operation

Positive – confirmed high efficiency and cost savings.

Easy integration into existing infrastructure
Clearly measurable results (false alarms, costs, incidents)
Close cooperation with municipal police

Overcoming initial distrust in AI technologies
Coordination with municipal IT systems

Start with a pilot deployment in a limited part of the infrastructure to clearly demonstrate effectiveness and make it easier to expand to the entire network.