DOC PREVIEW
MIT HST 723 - Speech recognition in noise as a function

This preview shows page 1-2-3-4-5-6 out of 17 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 17 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Speech recognition in noise as a function of the numberof spectral channels: Comparison of acoustic hearingand cochlear implantsLendra M. Friesen,a)Robert V. Shannon, Deniz Baskent, and Xiaosong WangDepartment of Auditory Implants and Perception, House Ear Institute, 2100 West Third Street,Los Angeles, California 90057HST. 723Gus F. Halwani4/8/2009Background. Speech recognition performance correlated with # of electrodes.(up to 4-7 electrodes). s/n ratio correlated with performance.(for electric and acoustic listeners). Electric listeners are particularly vulnerable to the effects of noise.(course spectral resolution?)Speech througha CI Simulator(1, 2, 4 channels)(4, 8, 32 channels)(original sound)(House Ear Institute, Shannon, Fu, & Galvin). What effect does....1) # of electrodes2) processing strategy3) s/n ratiohave on...1) medial vowel and consonant recognition2) monosyllable word recognition3) sentence recognitionin...1) electric listeners (using SPEAK, CIS, SAS)2) acoustic listeners (using different numbers of bands)?Methods...ListenerSpeechprocessingstrategyAge共years兲 GenderCIearN3 SPEAK 56 M RN4 SPEAK 40 M RN6 SPEAK 65 F RN7 SPEAK 55 M RN9 SPEAK 55 F LN14 SPEAK 63 M RN15 SPEAK 70 F LN17 SPEAK 71 F RN18 SPEAK 77 F RN19 SPEAK 70 M LC1 CIS 66 F LC3 CIS 56 M RC4 CIS 51 F LC5 CIS 38 M LC9 CIS 46 F RC2 SAS 72 M RC6 SAS 61 F RC7 SAS 82 M RC8 SAS 76 M R- No visual cues- Subjects 1m from speaker- Presentations at 65dB SPL- s/n ratio: no noise, 15, 10, 5, 0- (and -2.5, -5, -7.5, -10 for ALs)- Vowel and Consonant Recognition- Monosyllabic word recognition- Sentence recognitionChannel-Electrode Manipulation(SPEAK system)Channel-Electrode Manipulation(Nucleus systems) 2, 4, 7, 10, 20 electrodesChannel-Electrode Manipulation(Clarion systems)Results(Vowel Recognition)General decrease in noise, and electric listening asymptotes around 7-10 electrodes(Contrary to Fish et al.’s 4 electrodes) (normal hearing keeps benefitting... why?)Results(Consonant Recognition)General decrease in noise, and electric listening asymptotes around 7-10 electrodes(Contrary to Fish et al.’s 4 electrodes) (normal hearing keeps benefitting... why?)Results(Monosyllable Word Recognition)General decrease in noise, and electric listening asymptotes around 7-10 electrodes(Contrary to Fish et al.’s 4 electrodes) ‘Magic Number’EL Score > ALResults(Sentence Recognition)General decrease in noise, and electric listening asymptotes around 7-10 electrodes(Contrary to Fish et al.’s 4 electrodes) Less adv.for EL’sResults(# of channels, s/n ratio)snr makes big difference (preprocessing, comm strategy, better speech coding).Background. Speech recognition performance correlated with # of electrodes.(up to 4-7 electrodes). s/n ratio correlated with performance.(for electric and acoustic listeners). Electric listeners are particularly vulnerable to the effects of noise.(course spectral resolution?). Pulse Rate?(high fidelity encoding... hardware rate encoding not a limiting factor). Spectral Warping? (plasticity)Electrode Interaction Effects?. greater electrode spacing. changing the conductive medium. current focusing (stimulating electrodes simultaneously, differentially interacting to focus activation in one site) (van den Honert JASA 121 3703)Limiting FactorsConclusions. Increase in performance up to 7-10 electrodes. Electric listeners more variable than acoustic listeners. Different conditions? Cochlear vs Neural?. Looking Forward?(how to increase spectral resolution?)(Interaction result of too many electrodes?How to limit this?)Song (w/ male singer) through a CI simulation(4, 8, 16, 32 channels)(original sound)(House Ear Institute, Shannon, Fu, &


View Full Document

MIT HST 723 - Speech recognition in noise as a function

Download Speech recognition in noise as a function
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Speech recognition in noise as a function and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Speech recognition in noise as a function 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?