The implementation of efficient hearing tests using machine learning
Keywords:Bayesian active learning, Machine learning, Hearing tests, Audiogram, Auditory filter
Time-efficient hearing tests are important in both clinical practice and research studies. Bayesian active learning (BAL) methods were first proposed in the 1990s. We developed BAL methods for measuring the audiogram, conducting notched-noise tests, determination of the edge frequency of a dead region (fe), and estimating equal-loudness contours. The methods all use a probabilistic model of the outcome, which can be classification (audible/inaudible), regression (loudness) or model parameters (fe, outer hair cell loss at fe). The stimulus parameters for the next trial (e.g. frequency, level) are chosen to yield maximum reduction in the uncertainty of the parameters of the probabilistic model. The approach reduced testing time by a factor of about 5 and, for some tests, yielded results on a continuous frequency scale. For example, auditory filter shapes can be estimated for centre frequencies from 500 to 4000 Hz in 20-30 minutes. The probabilistic modelling allows quantitative comparison of different methods. For audiogram determination, asking subjects to count the number of audible tones in a sequence with decreasing level was slightly more efficient than requiring Yes/No responses. Counting tones yielded higher variance for a single response, but this was offset by the higher information per trial.
Békésy, G. von (1947). “A new audiometer,” Acta Otolaryngol. 35, 411-422.
Brand, T., and Kollmeier, B. (2002). “Efficient adaptive procedures for threshold and concurrent slope estimation for psychophysics and speech intelligibility tests,” J. Acoust. Soc. Am., 111, 1857-1868.
Chu, W., and Ghahramani, Z. (2005). “Preference learning with Gaussian processes,” Proceedings of the 22nd International Conference on Machine Learning, Bonn, Germany, 137-144.
Cobo-Lewis, A. B. (1997). “An adaptive psychophysical method for subject classification,” Percept. Psychophys., 59, 989-1003.
Cox, M., and de Vries, B. (2015). “A Bayesian binary classification approach to pure tone audiometry,” arXiv:1511.08670.
Fletcher, H. (1953). Speech and Hearing in Communication (Van Nostrand, New York), pp. 1-461.
Glasberg, B. R., and Moore, B. C. J. (1990). “Derivation of auditory filter shapes from notched-noise data,” Hear. Res. 47, 103-138.
Houlsby, N., Huszár, F., Ghahramani, Z., and Lengyel, M. (2011). “Bayesian active learning for classification and preference learning,” arXiv:1112.5745.
Houlsby, N. M., Huszár, F., Ghassemi, M. M., Orbán, G., Wolpert, D. M., and Lengyel, M. (2013). “Cognitive tomography reveals complex, task-independent mental representations,” Current Biol., 23, 2169-2175.
Jensen, N. S., Hau, O., Nielsen, J. B. B., Nielsen, T. B., and Legarth, S. V. (2019). “Perceptual effects of adjusting hearing-aid gain by means of a machine-learning approach based on individual user preference,” Trends Hear., 23, 1-23.
Kontsevich, L. L., and Tyler, C. W. (1999). “Bayesian adaptive estimation of psychometric slope and threshold,” Vision Res., 39, 2729-2737.
Levitt, H. (1971). “Transformed up‐down methods in psychoacoustics,” J. Acoust. Soc. Am., 49, 467-477.
Minka, T. P. (2001). “Expectation propagation for approximate Bayesian inference,” Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence, Seattle, Washington, USA, 362-369.
Moore, B. C. J, and Glasberg, B. R. (2004). “A revised model of loudness perception applied to cochlear hearing loss,” Hear. Res. 188, 70-88.
Nielsen, J. B. B., Nielsen, J., and Larsen, J. (2014). “Perception-based personalization of hearing aids using Gaussian processes and active learning,” IEEE/ACM Trans. Audio, Speech, Language Process., 23, 162-173.
Patterson, R. D. (1976). “Auditory filter shapes derived with noise stimuli,” J. Acoust. Soc. Am., 59, 640-654.
Rasmussen, C. E., and Williams, C. K. I. (2006). Gaussian Processes for Machine Learning, MIT Press, Cambridge, MA, USA.
Rasmussen, C. E., and Nickisch, H. (2010). “Gaussian processes for machine learning (GPML) toolbox,” J. Mach. Learn. Res. 11, 3011-3015.
Schlittenlacher, J., Turner, R. E., and Moore, B. C. J. (2018a). “Audiogram estimation using Bayesian active learning,” J. Acoust. Soc. Am. 144, 421-430.
Schlittenlacher, J., Turner, R. E., and Moore, B. C. J. (2018b). “A hearing-model- based active-learning test for the determination of dead regions,” Trends Hear. 22, 1-13.
Schlittenlacher, J., and Moore, B. C. J. (2019). “Fast estimation of equal-loudness contours using Bayesian active learning and direct scaling,” Acoust. Sci. Tech. (in press).
Shannon, C. E. (1948). “A mathematical theory of communication,” Bell Syst. Tech. J. 27, 379–423, 623–656.
Shen, Y., and Richards, V. M. (2013). “Bayesian adaptive estimation of the auditory filter,” J. Acoust. Soc. Am., 134, 1134-1145.
Shen, Y., Sivakumar, R., and Richards, V. M. (2014). “Rapid estimation of high- parameter auditory-filter shapes,” J. Acoust. Soc. Am., 136, 1857-1868.
Shen, Y., Zhang, C., and Zhang, Z. (2018). “Feasibility of interleaved Bayesian adaptive procedures in estimating the equal-loudness contour,” J. Acoust. Soc. Am., 144, 2363-2374.
Song, X. D., Wallace, B. M., Gardner, J. R., Ledbetter, N. M., Weinberger, K. Q., and Barbour, D. L. (2015). “Fast, continuous audiogram estimation using machine learning,” Ear Hearing, 36, e326–e335.
Song, X. D., Garnett, R., and Barbour, D. L. (2017). “Psychometric function estimation by probabilistic classification,” J. Acoust. Soc. Am., 141, 2513-2525.
Stevens, S. S. (1956). “The direct estimation of sensory magnitudes: Loudness,” Am. J. Psychol., 69, 1-25.
Watson, A. B., and Pelli, D. G. (1983). “QUEST: A Bayesian adaptive psychometric method,” Percept. Psychophys., 33, 113-120.
Williams, C. K. I. and Barber, D. (1998). “Bayesian classification with Gaussian Processes,” IEEE Trans. Pattern Anal. Mach. Intell. 20, 1342–1351.
How to Cite
Authors who publish with this journal agree to the following terms:
a. Authors retain copyright* and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
b. Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
c. Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
*From the 2017 issue onward. The Danavox Jubilee Foundation owns the copyright of all articles published in the 1969-2015 issues. However, authors are still allowed to share the work with an acknowledgement of the work's authorship and initial publication in this journal.