Genetic algorithms for adaptive psychophysical procedures: recipient-directed design of speech-processor MAPs.

Area of Science:

• Biomedical Engineering
• Auditory Neuroscience
• Computational Intelligence

Background:

• Optimizing audio-processing designs, particularly for cochlear implants, is complex.
• Current methods may require extensive listener feedback or prior knowledge.
• Developing efficient listener-directed optimization is crucial for improving device fitting.

Purpose of the Study:

• To evaluate the efficacy of genetic algorithms (GAs) for listener-directed optimization of audio-processing designs.
• To test the hypothesis that cochlear-implant recipients can utilize a GA-guided adaptive psychophysical search.
• To assess the selection of effective speech processor Microphone Array Programs (MAPs) from a large set.

Main Methods:

• An adaptive psychophysical procedure was developed using genetic algorithms.
• Listeners selected preferred speech processor MAPs, which were then updated by the GA.
• Experiments were conducted with cochlear-implant recipients to analyze convergence and optimized MAP properties.

Main Results:

• The adaptive procedure demonstrated convergence to useful speech processor MAPs.
• Optimization was achieved within approximately twenty iterations for all participants.
• Results were consistent across five cochlear-implant recipients.

Conclusions:

• Genetic algorithms offer an efficient method for searching numerous audio-processing designs.
• This approach minimizes the need for extensive listener feedback or prior preference knowledge.
• The procedure shows potential as a clinical aid for fitting cochlear-implant speech processor MAPs.