Summary

The World Health Organization estimates that 1.1 billion people worldwide are at risk of hearing loss caused by personal audio devices. Studies that investigate hazardous noise exposure have largely relied on participant self-report and laboratory testing, which can be unreliable and is often impractical when trying to understand music listening behaviour on a wider scale.

There has been a call to arms for the implementation of mobile applications that estimate an individual’s noise dose from recreational listening and present this data back to the user, in order to make them aware of their potentially hazardous behaviour.

The BTA Loudness Study mobile application was created to better understand media listening behaviour. By developing a native application for the Android platform, detailed metrics such as spectral information about playing audio as well as more general metrics such as track details, listening duration and volume changes can be monitored and analysed.

Developing a mobile phone app

The project successfully developed a mobile phone application capable of measuring these listening behaviours. The app is simple to downloaded, using the Google Play app store, and works “silently” in the background monitoring listening behaviour for pre-agreed lengths of time. Initial lab based experiments demonstrated that it was possible to measure RMS output from mobile phones with sub-second resolution allowing variations of sound level over entire songs to be measured. In addition, volume changes during listening could be monitored which could potentially be used to help estimate sound level output.

Before launch the app was extensively tested in laboratory settings. The app functioned as expected and did not intrude on the other functions of the phone. Testing revealed that the app only changed battery and data use by fractional amounts, demonstrating that it could be run without significantly effecting normal phone usage. In addition, battery usage was so low that battery monitoring mobile applications were unable to attribute any significant battery usage to the application, meaning users were unlikely to uninstall the app due to it effecting performance. RMS measured using the app was highly consistent with the sound output measured from headphones, demonstrating its ability to provide temporally reliable estimates of sound output from mobile phone listening.

Using the app in the real world

We next tested the app in real world settings by recruiting members of the public to download the app and testing its ability to monitor listening behaviour, without interrupting use, over extended periods of time. 52 Participants were recruited to download and use the app over a 28-day period, the app collected both real-time listening data and questionnaire data.

The majority of users were identified as using some form of in-ear earphone, this is worrying as these earphones are known to be louder and may pose more risk of ear infection. In addition, the app identified a small population of “extreme users” who listened to content almost every single day or who listened for more than 3 hours a day.

These data demonstrate not only the feasibility of the app but some interesting early findings that warrant further investigation. Both headphone type and listening behaviour data suggest that there may be extreme users in the population potentially impacting their hearing through their music listening habits.

 

The app also demonstrated its ability to measure alternate listening behaviour that may prove useful to further classifying and measuring potential risks in the population. For example, the data collected suggested that young people (<30s) tended to listen more frequently than older listeners, supporting the idea that young people may be particularly at risk. The app was also useful in identifying other listening behaviours such as usage profiles. Using the app we were able to identify day time listening and putative commuter listening profiles, suggesting it may be possible to infer additional information about listening habits using this approach.

Conclusion

Overall the study was a success, demonstrating that an easy to download app could be distributed through normal channels, i.e. Google Play, and function with minimal cost in performance to the user. The data generated through this method could allow large scale assessment of listening behaviours and potentially, with further work, yield estimates of sound doses. The data captured allowed us to determine participants that were exhibiting potentially hazardous listening behaviours. In the future, this will inform the design of behavioural interventions that will attempt to reduce this behaviour, resulting in a lower noise dose, and reduced risk of hearing damage.

Photo by William Iven on Unsplash