Laboratory for Audio Communication and Acoustics

Research is underway to develop methods for enhancing speech sinals using artificial intelligence (AI) to improve speech intelligibility and quality in acoustic conditions where conventional algorithms often reach their limits. These innovative methods are frequently combined with classical algorithms to achieve the best possible results.

Uses:

• Voice control
• Text-to-speech
• In-car communication (ICC)
• Smartphones
• Smart homes
• Medicine (e.g. operating tables)
• Care (e.g. care beds)
• Industry (e.g. manufacturing)

Predictive maintenance with artificial intelligence (AI) involves the early detection of critical machine conditions. Sensors monitor the state of a machine, and the AI analyses this data to schedule maintenance proactively, preventing unexpected breakdowns. This approach aims to enhance sustainability and cost-effectiveness while providing advantages over traditional maintenance intervals.

Uses:

• Correctly assess system status
• Routine maintenance as required
• Recognise critical machine conditions
• Avoid sudden machine failures
• Avoid expensive production stops
• Evaluate large amounts of data from sensors with AI

New interaction options, such as gesture or voice control, along with facil recognition using artificial intelligence (AI), are at the forefront of this research. In contrast to conventional methods for controlling devices, this approach allows data to be collected and control commands to be transferred while the operator carries out another activity. The aim is to increase convenience, efficiency and safety. Sensor data fusion also plays a key role in this process.

Uses:

Driver State Monitoring Systems: These are able to determine the driver's condition and increase safety while driving by providing information or directly intervening in what is happening. Relevant impairments are, for example, physical impairment, fatigue, alcohol or drug consumption, distraction, unexpected and sudden illness. In addition the detection of emotions, stress levels, health, speaker verification for access control are other valuable applications in vehicles.

To increase road safety, vehicles will learn to hear in the future. In Acoustic Environment Perception, acoustic events outside the vehicle are recorded using airborne sound microphone arrays and processed in real time by an intelligent signal processing unit. The recognition of different events and hazardous situations outside the vehicle is realised with the help of artificial intelligence and machine learning techniques.

Uses:

Detection of an approaching siren:

• Display on the dashboard to form an emergency lane in good time.
• Switching on the hazard warning flashers and initiating an evasive manoeuvre.
• Automatic reduction of the volume of the car hi-fi system.
  
  Condition monitoring of the vehicle based on driving noise:
• early detection of critical signs of wear and tear.
• Detection of fault conditions due to unusual driving noises (e.g. nail in tyre).
  
  Audio-visual detection:
• Hazardous situations: Play road, pedestrian and cycle paths.
• Driver gives instructions, e.g. for parking, unparking or re-parking an autonomous vehicle.