Eric Alan Durant
Department of Electrical Engineering and Computer Science
The University of Michigan
Ann Arbor, MI, USA
Doctoral Committee:
Associate Professor Gregory H. Wakefield, Chair
Professor David J. Anderson
Associate Professor Yili Liu
Dr. Dianne J. VanTasell
Wednesday 31 January 2001
The perceptual tuning problem is encountered in many fields. It is the problem of tuning a system, which may have many interacting parameters, to provide output which is good as perceived by the system’s users. The system may be a radio, hearing aid, personal digital assistant, frozen custard machine, etc. Approaches to the perceptual tuning problem may be judged on ease of use, speed, robustness and generality, among other qualities.
In this research, we investigate algorithms and techniques to enable genetic algorithms (GAs) to be used successfully in a variety of perceptual tuning problems. To validate this research, we apply it to a difficult problem which has a rich history of research into its many aspects, and to which a variety of tuning algorithms have been applied with varying success. This problem is DSP-based hearing aid fitting.
Fitting modern DSP-based hearing aids, which have numerous parameters, is a challenge for even the experienced audiologist. Simple but proven fitting rationales are used, limiting the prescribed setting to a subset of the device’s capabilities. We will first investigate whether a genetic algorithm can guide a hearing-impaired subject towards perceptually good feedback canceller parameters in everyday conditions when used with his or her hearing aid prescription. Many previous studies of GAs for perceptual tuning [1, \2] and other adaptive hearing aid fitting methods [3] were conducted in the laboratory, limiting their applicability to real-world use.
We will use the results of this first investigation to extend our procedure to general hearing aid fitting, including frequency selective amplification and compression. Our approach will use prescriptive settings from existing rationales for linear and compressive aids as starting points, much as Neuman et al. [3] did for hearing aids with linear amplification. Linear rationales have been shown to work well for basic hearing aids, but compressive rationales remain a subject of major debate [4].
[1] Hideyuki Takagi and Miho Ohsaki, “IEC-based hearing aid fitting,” in IEEE International Conference on Systems, Man, and Cybernetics (SMC ‘99). 1999, vol. 3, pp. 657–662, IEEE.
[2] Paul R. Runkle, Optimization of Head-Related Transfer Functions Using Subjective Criteria, PhD thesis, The University of Michigan, 2000.
[3] Arlene C. Neuman, Harry Levitt, Russell Mills, and Teresa Schwander, “An evaluation of three adaptive hearing aid selection strategies,” Journal of the Acoustical Society of America (JASA), vol. 82, pp. 1967–1976, 1987.
[4] Harvey Dillon, “Tutorial: Compression? Yes, but for low or high frequencies, for low or high intensities, and with what response times?,” Ear and Hearing, vol. 17, no. 4, pp. 287–307, 1996.