Presentation is loading. Please wait.

Presentation is loading. Please wait.

INTRODUCTION Human hearing and speech cover a wide frequency range from 20 to 20,000 Hz, but only a 300 to 3,400 Hz range is typically used for speech.

Similar presentations


Presentation on theme: "INTRODUCTION Human hearing and speech cover a wide frequency range from 20 to 20,000 Hz, but only a 300 to 3,400 Hz range is typically used for speech."— Presentation transcript:

1 INTRODUCTION Human hearing and speech cover a wide frequency range from 20 to 20,000 Hz, but only a 300 to 3,400 Hz range is typically used for speech communication devices including telephones (1). Traditionally, sound frequencies below 300 Hz have been considered redundant, but recent studies have shown that low-frequency sounds can significantly improve cochlear implant performance in noise, particularly when the noise is a competing voice (2-4). Here we used cochlear implant simulations in normal hearing listeners to study the size and mechanisms of this improvement. We found that largely unintelligible low-frequency sounds considerably enhanced speech recognition in noise.METHODS Sixteen normal hearing subjects participated in the study (8 for low-frequency EAS simulation test, 8 for high-frequency EAS). Each subject listened to a 4-channel CI simulation (5) with residual acoustic hearing at either low frequencies ( 2000, 4000, or 6000 Hz). The acoustic sound was combined with the CI simulation either monaurally to simulate the standard Electro- Acoustic-Stimulation (EAS) or binaurally (both diotic and dichotic methods included) to simulate the CI and hearing aid stimulation. HINT sentences (6), in the presence of a competing voice, were used to measure the speech reception threshold (SRT), in terms of signal-to-noise ratio (SNR). An additional control simulating a 5-channel CI was used to assess whether potential improvement associated with acoustic hearing was equivalent to the result of a CI simulation simply with more channels.  SRT: SNR at which subjects score 50% correct on sentences FIGURE 1: Schematic view of all the processing strategies and frequency divisions ( 1a ) ( 1b ) Acoustic stimulation 4000Hz Row 1 EAS CI simulation + acoustic cues Row 2 Control 1 5-channel CI simulation (Can improved performance be attributed to an effect of more channels?) Row 3 Control 2 4-channel CI simulation FIGURE 2: Speech spectrograms of original and processed Speech-in-Noise Stimuli  A (male voice, target) + B (female voice, masker) = C (combined original signal)  D (acoustic) + E (electric, 4-channel CI) = F (EAS) ( 2a ) ( 2b ) Acoustic stimulation 4000HzRESULTS Although extreme low- ( 4000 Hz) acoustic sounds had negligible intelligibility when presented alone, they both improved CI performance, including bilateral implant simulation, when combined. While the method of combination (monaural, diotic, or dichotic) had no significant effects, the degrees of improvement were different between low and high-frequency sounds. Performance of low-frequency sound conditions significantly surpassed that of the 5-channel CI control, whereas high-frequency sound produced similar performance to 5-channel CI. These results suggest a unique mechanism underlying the improvement of cochlear implant performance by addition of low-frequency information. FIGURE 3: SRT of EAS with low-frequency acoustic sounds A control study of perception of the residual low- and high-frequency sounds showed that the intelligibility was 0, 10, and 11% correct for sentences and 5, 51, and 75% correct for keywords, with the low-pass cutoff frequency at 250, 500, and 1,000 Hz, respectively; 56, 3, and 0% correct for sentences and 85, 18, and 14% for keywords, with the high-pass cutoff frequency at 2,000, 4,000, and 6,000 Hz. We found the improvement in SRT of EAS simulation (7-15 dB increase in all EAS conditions over one or bilateral implant ) similar to some previous studies done on real CI+HA subjects. (see FIGURE 4) FIGURE 4: Sentence recognition score as a function of SNR (3) This real CI subject study (by Kong et al.) used a male masker and female speech (3). The figure to the right shows the HA alone provides very limited perception, but very high perception when combined with CI. Notice the similarity between this real subject data and the simulation data in FIGURE 3. Y.Y. Kong et al FIGURE 5: SRT of low- and high- frequency EAS The 500 Hz (low-frequency acoustic) EAS simulation gave the best SRT result, outscoring the 5-channel CI control by 8 dB. On the contrary, the 4000 Hz EAS (high- frequency acoustic) is 4 dB less efficient than the 5-channel CI. The low-frequency acoustic sound shows more benefits in speech-in-noise test, in terms of the relative contribution to the combined speech intelligibility.DISCUSSION The low- frequency sound improved the SRT by 7-15 dB over the “one implant” and “bilateral implant’’ controls [p < 0.05]; this cannot be explained by any current theories. The high-frequency sound did not improve the SRT as much as the low-frequency counterpart did; the 500 Hz low-pass EA simulation gave a 9 dB benefit over the 4000 Hz high-pass EA simulation. CONCLUSIONS For simulation, EAS achieves better performance than a single cochlear implant or bilateral implants in speech-in-noise tests. Low-frequency acoustic sound ( 4000 Hz) does. The present result suggests a synergetic interaction between the low- and high- frequency sounds not in the ear but in the brain. We hypothesize that the low- frequency sound helps the brain segregate and group the high-frequency temporal envelopes into different sound sources. FUTURE DIRECTIONS Similar experiments are being conducted to see whether these simulation results can be extended to actual bimodal and bilateral cochlear implant users.REFERENCES 1. H. Fletcher, Speech and Hearing in Communication, Bell Laboratories Series (D. Van Nostrand Company, Inc.,New York, ed. Second, 1953). 2. C. W. Turner, B. J. Gantz, C. Vidal, A. Behrens, B. A. Henry, J Acoust Soc Am 115, (Apr, 2004). 3. Y. Y. Kong, G. S. Stickney, F. G. Zeng, J Acoust Soc Am 117, (2005). 4. C. von Ilberg et al., ORL J Otorhinolaryngol Relat Spec 61, (Nov-Dec, 1999). 5. R. V. Shannon, F. G. Zeng, V. Kamath, J. Wygonski, M. Ekelid, Science 270, (Oct 13, 1995). 6. M. Nilsson, S. D. Soli, J. A. Sullivan, J Acoust Soc Am 95, (Feb, 1994).ACKNOWLEDGEMENTS We would like to thank Ms. Abby Copeland and Dr. Ed Rubel for commenting on our manuscript. This experiment was supported by NIH Grant 2R01 DC LOWS ARE THE NEW HIGHS: IMPROVING SPEECH INTELLIGIBILITY WITH UNINTELLIGIBLE LOW-FREQUENCY SOUNDS LOWS ARE THE NEW HIGHS: IMPROVING SPEECH INTELLIGIBILITY WITH UNINTELLIGIBLE LOW-FREQUENCY SOUNDS Janice E. Chang 1, John Y. Bai 2, Martin Marsala 2, Helen E. Cullington 2, and Fan-Gang Zeng 2 1Department of Bioengineering, University of California, Berkeley, CA, USA 2Hearing and Speech Research Laboratory, University of California, Irvine, CA, USA A. Original HINT: "A boy fell from the window." B. Competing voice: "A pot of tea helps to pass the evening." F. Combined high-pass and implant simulation. C. Original signal + competing voice (SNR=0 dB) D. High-passed original mixed sound. E. Four-channel implant simulation of the mixed sound below 4000 Hz. B. Competing voice: "A large size in stockings is hard to sell." E. Four-channel implant simulation of the mixed sound above 500 Hz. D. Low-passed original mixed sound. F. Combined low-pass and implant simulation. C. Original signal + competing voice (SNR=0 dB) A. Original HINT: "A boy fell from the window."


Download ppt "INTRODUCTION Human hearing and speech cover a wide frequency range from 20 to 20,000 Hz, but only a 300 to 3,400 Hz range is typically used for speech."

Similar presentations


Ads by Google