How it works

Acoustic Tunnel Monitoring

Events are detected acoustically via the microphones installed in the tunnel.
Abnormal sounds are recognized and an alarm is transmitted to the control room within 0.7 seconds.

Principle

The operating noise inside a tunnel is characterised by a specific mix of engine, rolling and airflow noise produced by vehicles passing through. Microphones installed in the tunnel detect acoustic anomalies such as a vehicle-wall and vehicle-vehicle collisions, tyre bursts, voices and shouts, etc.

Special detection algorithms make it possible to automatically identify these sounds and assign them to specific alarm classes. One big advantage of acoustic detection methods is that accidents or critical incidents in tunnels are virtually always accompanied by a characteristic sound. These sounds occur concurrently with the incident and can thus be immediately detected.
The detection of an incident allows measures to be immediately and automatically activated. For example, the system may, depending on the alarm or accident class, trigger an acoustic alarm in the tunnel control room and display the video camera image of the affected section on a central screen giving the tunnel operator a real-time overview of the situation. This saves valuable time in providing first aid to people involved and in alerting drivers approaching the scene of the accident.

Incident detection

Incident detection involves several steps. Abnormal sounds are first distinguished from normal tunnel sounds for each microphone channel. The machine learning method used models normal conditions and detects any deviation from them. This enables the system to detect abnormal and unknown sounds.

Alert classes
  • Crash and tyre burst
  • Tyre squealing
  • Door banging
  • Car horn honking
  • Voices/speech/shouts/screaming people
Features
  • Live streaming of sounds and voices
  • Sound localisation by assignment to nearest
    microphones (100 – 150m apart, depending on
    camera positions)
  • Alarm in the tunnel control room with category, event
    and camera image
  • Tunnel operator obtains information even when visibility
    is poor
  • Ring buffer for all microphone signals