Conference proceeding
Assessing the performance of hearing aids using surveys and audio data collected in situ
2017 IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), pp.283-288
05/2017
DOI: 10.1109/INFCOMW.2017.8116390
Abstract
Measuring the performance of hearing aids in the real-world is challenging because it depends on both the subject's auditory abilities and the auditory context in which they use their hearing aids. The auditory context includes characteristics of the listening activity, social context, and acoustic environment. Ecological momentary assessment is the state-of-the-art method for evaluating the performance of hearing aids in the real-world. It may be used to assess the auditory context and its performance in that context. However, such techniques a introduces a significant data collection burden on study participants. A potential avenue for reducing the data collection burden is to use audio data collected in-situ to characterize the auditory context automatically. This paper presents our initial attempt at answering this challenge. Our results indicate that it is possible to predict the noise level and listening activity with an accuracy of 68% and 70%, respectively. Additionally, we show that a 4.6% reduction in the mean squared error of predicting the listening effort may also be achieved by using the audio data (without including subjective self-reports).
Details
- Title: Subtitle
- Assessing the performance of hearing aids using surveys and audio data collected in situ
- Creators
- Syed Shabih Hasan - University of IowaRyan Brummet - University of IowaOctav Chipara - University of IowaYu-Hsiang Wu - University of Iowa
- Resource Type
- Conference proceeding
- Publication Details
- 2017 IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), pp.283-288
- DOI
- 10.1109/INFCOMW.2017.8116390
- Publisher
- IEEE
- Language
- English
- Date published
- 05/2017
- Academic Unit
- Communication Sciences and Disorders; Computer Science
- Record Identifier
- 9984258850702771
Metrics
80 Record Views