Robots improve hearing aids

- EN - DE- FR- IT
Complete system with mobile robots, electronics, audio interface and artificial
Complete system with mobile robots, electronics, audio interface and artificial heads.
It is extremely time-consuming to measure how sound behaves in a room. The Lucerne University of Applied Sciences and Arts and the hearing aid manufacturer Sonova have therefore developed robots that can take over this task. This serves to improve hearing aids in rooms with a lot of background noise.

A restaurant with a lot of reverberation and background noise is sometimes a problem even with healthy hearing. People with hearing problems are often completely lost because they are unable to filter out the essential acoustic information. Hearing aids can be adjusted according to the environment. However, these audio filters only offer very general default settings and therefore always reach their limits. This makes social contact extremely difficult for people with a hearing impairment - they are excluded from shared conversations in a restaurant, and it is difficult for them to contribute their ideas in a business meeting because they have to concentrate primarily on understanding what others are saying.

Hearing aid manufacturers are working hard to improve these filtering options. However, this requires a more precise knowledge of how rooms behave in detail at different positions. "In room acoustics, we talk about a room having a certain reverberation. But this statement is not enough to describe the differences in room acoustics from point to point," says Armin Taghipour, acoustics expert at Lucerne University of Applied Sciences and Arts. Hearing is always about the relationship between two points: the source of the sound and the person hearing the sound. Lucerne University of Applied Sciences and Arts and the hearing aid manufacturer Sonova are therefore working together on a project funded by Innosuisse to gain a better understanding of sound propagation in a room.

Robots measure the sound in a room

First of all, there is no other way to do this than to measure the behavior of the sound in the entire room. How does it change when the person speaking approaches the person hearing? When they walk around them? When they move away? If it all happens in a corner, or in the middle of the room? "If these detailed measurements are carried out manually by people, they are extremely time-consuming," says Armin Taghipour. That’s why he and his team, led by Pascal Jund, Tobias Walker and Manuel Isenegger, have developed robots at Lucerne University of Applied Sciences and Arts that can do this autonomously. Thanks to specially developed software, they are able to move around the room independently in the desired way if the boundaries - here a wall, there a table - are defined. They can measure how the sound behaves at each location.
In contrast to previous practice, the robots can autonomously measure both the volume of background noise and the room acoustics. For example, if someone is speaking in a canteen, the device can measure the effect on every point in the room. The results from different rooms can then be compared - high, low, large, small, carpeted rooms or rooms with wooden paneling.

Complicated interaction

What sounds quite simple - robots measuring the acoustics of a room - requires the cooperation of specialists from a wide range of fields. Sonova contributed its knowledge of hearing acoustics, audio technology, signal processing and electroacoustics, while HSLU added its expertise in robotics, software development, room acoustics, audio signal analysis and processing.
Data is now being collected in various rooms as part of a student project. This data creates the basis for simulations. These can then be used to modify audio recordings so that they sound as if they had taken place at the measurement location. "For Sonova, this data should provide the basis for creating new algorithms. After all, the most important prerequisite for machine learning to work is large amounts of data," says Hannes Wüthrich, Project Manager at Sonova. In the case of written texts or images, it is often possible to draw on numerous existing data sets - but not so in the case of acoustic data. These must first be generated, as can now be done with the help of the data collected by robots. Ultimately, the use of robots should lead to an improvement in hearing aids and better integrate people with hearing loss into their environment.

Path Planner’ operating software: The black point cloud represents the contour of the room. It is used by the robots for navigation. The red path delimits the work area. Outside the red path is the forbidden zone, which cannot be approached by the robots. In this scenario, the robot on the left will remain stationary while the robot on the right moves towards it along the indicated path and performs a measurement at each point.