Learning from audiological data collected in the lab and the real world

Author: Dr Niels Pontoppidan, Eriksholm Research Centre, Denmark

Abstract: While audiological data is often discussed as a single matter, it encompasses a wide range of data that addresses independent steps in the research, development, and usage of hearing devices. Early large-scale collections of audiograms are a good example of the start of this era of audiological data giving rise to the first models for normal hearing and age-adjusted normal hearing, and later creating insights on how different occupations affect hearing.

On the instrument side, around 20 years ago, hearing instruments started storing aggregated information about the sound environments which they had been worn in and, the corresponding sound pressure levels, and the use of programs programmed into the hearing instruments. This audiological data along with recordings of everyday sounds, and insights from the wearers of hearing instruments suggested to increase the signal to noise ratios in audiological test paradigms. In the past few years hearing instruments connected to phones to store timestamped records of sound environments, operations, and user feedback. With sufficient coverage of everyday use, the detailed logging data enables a new type of studies where the participants evaluate the alternatives in their natural environments, while the researchers can also get some information about when and which alternative was tried in which situation.

Moreover, yet another type of audiological data goes directly into developing and tuning of audiological algorithms for the hearing instruments. Here the audiological data encompasses speech, everyday sounds, music, and various noises, and by defining the desired signals tuning the algorithms to prioritize speech, segregation, music, and or comfort.

Research in the last decades with the audiological data led to many important discoveries, and today, as the area of data emerges the focus turns to maturing those discoveries along the dimensions of coverage, applicability, bias, and privacy into solutions that improve the lives for people with hearing problems.