Speech Based Optimization of Hearing Devices Alice E. Holmes, Rahul Shrivastav, Hannah W. Siburt & Lee Krause
The Problem Programming is based on electrically measured dynamic ranges of pulsed stimuli (non-speech) Current programming methods have numerous options
Purpose The goal is to understand speech, the tuning of the device should be based on speech and not tones. Development of a standard metric to understand the strengths and weaknesses of the individual device user. The complexity of problem requires an automated and intelligent process to optimize the device programming.
Overview Hearing Device, D Brain, B Input Signal, S inp Output Signal, S out Intermediate Signal, S int Almost nothing is known about the function B
What to optimize? Acoustic contrasts essential for speech intelligibility -- Minimize error function From patient experiments, we can get data for different values of the parameters and the corresponding errors –The dimensionality of this data is related to the number of independent programmable parameters –Many parameters, hence very high dimensionality leading to the “curse of dimensionality”
How to reduce the complexity of the problem? Artificial Intelligence Algorithms -- Patient-independent knowledge should be available (e.g. as “rules”) -- Patient-specific knowledge should be statistically extracted from the performance of each patient -- “Model field theory” approach to model relationships
Initial Clinical Trial 20 adults with –N24 or New Freedom implants –Freedom Processors Adjusted the following parameters –Rate –Loudness growth –Frequency allocation tables Outcome measures – –CNC lists in quiet – –BKB-SIN – –Subjective questionnaire
Subject Demographics Gender Male Female N=7 N=13 Age (Years) Mean S.D. Range Length of CI Use (months) Mean S.D. Range Type of CI N24 New Freedom N=3 N=17
Initial Clinical Trial The Optimization program (Clarujust™) was designed to interface with a customized version of Cochlear Corp. Custom Sound so that programming changes recommended by the algorithm could be tested seamlessly. All stimuli were presented through a direct connection to the speech processor and at a constant level across all test sessions (approximately 60 dBA). 3 Sessions – two weeks apart
Clarujust™ A series of VCV syllables were presented & verbal responses were recorded by the researcher. NWE for the processor setting was calculated The next combination of FAT, PR & LG was automatically recommended & tested. Procedure was repeated for 30 minutes
Procedures Outcome measures Clarujust™ routine Map with lowest net weighted error (NWE) was selected and programmed in to processer for use until next session
Subject Map Parameters
CNC Word Scores Significant difference among the three conditions using Greenhouse-Geisser analysis (p < 0.004). Further trend analyses indicated a significant ascending omnibus trend from baseline (p < 0.004) Pairwise comparisons significant differences between baseline and Opt 1 (p < 0.025) and between Baseline and Opt 2 (p < 0.015).
CNC Phoneme Scores Significant difference among the three conditions using Greenhouse-Geisser analysis (p < 0.008). Further trend analyses indicated a significant ascending trend from baseline (p < 0.015) Pairwise comparisons showed significant differences between base line and Opt1 (p < 0.003) and between Baseline and Opt 2 (p < 0.04).
BKB-SIN Scores Significant difference among the three conditions using Greenhouse-Geisser analysis (p < 0.03). Further trend analyses indicated a significant ascending quadratic trend from baseline (p < 0.009). Pairwise comparisons showed significant differences between baseline and Opt 1 (p < 0.03)
Subjective Results At the end of this clinical trial, 17 out of 20 patients preferred to continue using one of their optimized maps. Subjective ratings in various situations were also obtained from each subject (Holden, et al, J Am Acad Audiol 18:777– 793, 2007)
Summary The optimization method used in this study resulted in improved subject performance in all outcome measures. Speech perception was significantly better in word and phoneme identification with optimized maps. In addition, subjects performed better in noise using the optimized maps. Subjective tests suggest that patients preferred the optimized maps in their daily lives.
What is Next? Continue to refine process with CI technology Currently doing clinical trials with two hearing aid manufacturers –Three pilot subjects have been fitted with bilateral hearing aids using the optimization protocol Future applications –Hybrids –Audiologic rehabilitation –Cell phones –????
Thank you to the students involved Hannah Siburt Kevin Still Elyse Schwartz Bekah Gathercole
Acknowledgments This project is funded by Audigence, Inc. and the Florida High Tech Corridor Council. We wish to thank Cochlear Corporation for supplying the fitting software platform and for their extensive and timely technical support. We also want to thank our subjects for their willingness to participate in the experiment.