Fang Du, Dr. Christina L. Runge, Dr. Yi Hu. April 21, 2018 Effect of High-Rate Pulse Trains on Speech Perception for Cochlear Implant Recipients Fang Du, Dr. Christina L. Runge, Dr. Yi Hu. April 21, 2018
Outline Introduction Problem Statement Design and Implementation Testing and Data Analysis Conclusion
Introduction What is cochlear implant(CI) A cochlear implant is a device that provides a sense of sound to a person who is profoundly deaf or severely hard of hearing The implant is surgically placed under the skin behind the ear A cochlear implant has four major parts: Microphones Speech processor Transmitter and receiver/stimulator Electrodes College of Imaging Arts & Sciences. (n.d.). Retrieved March 02, 2018, from https://cias.rit.edu/faculty-staff/101/faculty/337
Introduction How cochlear implant works Normal Hearing/Hearing Aids Acoustic Electrical
Proposed Solution: Design a new stimulation strategy Problem Statement In Quiet Improved sound processing strategies have helped many cochlear implant recipients regain satisfactory speech perception in quiet during the past three decades. The average monosyllabic word score is nearly 80% In Noise Many signal processing techniques such as multiple-microphone noise-reduction and speech-enhancement methods have been developed to address this issue. In general, they increase speech distortion as well as battery power consumption with limited benefits. SOLVED Proposed Solution: Design a new stimulation strategy Srinivasan, A., Padilla, M., Shannon, R. and Landsberger, D. (2013). Improving speech perception in noise with current focusing in cochlear implant users. Hearing Research, 299, pp.29-36.
Problem Statement Desynchronize neural response to electric signal Poor speech perception Poor electrode discrimination Abnormally high synchrony in neural response Loss of inner hair cell synapse Desynchronize neural response to electric signal Limited dynamic range Auditory nerves cannot process spontaneous activity Runge-Samuelson, C. L. (2009). Effects of high-rate pulse trains on electrode discrimination in cochlear implant users. Trends in amplification, 13(2), 76-86.
Problem Statement Jay Rubinstein’s Model: It has been shown that high-rate desynchronized pulse train (DPT) has produced psychophysical dynamic range increases as large as 17dB in human subject. No results have been reported to show Rubinstein’s model improves cochlear implants receivers’ speech perception Adding continuous, appropriate-level, high-rate desynchronized pulse train (DPT) stimuli facilitates the desynchronizing of the neural response and increases the dynamic range. Hong, R. S., Rubinstein, J. T., Wehner, D., & Horn, D. (2003). Dynamic range enhancement for cochlear implants. Otology & neurotology, 24(4), 590-595.
Strategy Design Speech signal High-rate desynchronize pulses Regular Continues Interleaved Stimulation (CIS) strategy CIS strategy with DPT pulse (Jay Rubinstein’s Model) Speech signal High-rate desynchronize pulses Litvak, L., Delgutte, B., & Eddington, D. (2001). Auditory nerve fiber responses to electric stimulation: modulated and unmodulated pulse trains. The Journal of the Acoustical Society of America, 110(1), 368-379.
Strategy Design Proposed High-rate DPT Strategy Stimulation Strategy without DPT
Strategy Implementation Envelope Extraction Bandpass filters Compression
Strategy Implementation Filter Bank To separate the speech signal to different electrodes FB was individually customized based on patients’ map(corner frequency). Each band is 6th-order Butterworth bandpass filter
Strategy Implementation Envelope Extraction To determine the amplitude of biphasic pulse delivered to electrode Hilbert Transform was used to closely extract the overall shape of a sound
Strategy Implementation Compression To ensure the output to fit into the patient’s dynamic range of electrically evoked hearing 𝑌= MCL −THR × log(1+𝑦×𝑐) log(1+𝑐) +THR Y Compression output y Envelope extraction output MCL Most comfortable level THR Threshold level c Patient maplaw
Strategy Implementation Speech without DPT Speech with DPT Patient Information Future Testing
Speech Perception Testing Ten cochlear implants users participated up to now IEEE sentences were processed in both quiet and 4-talker babble background noise (SNR +10dB) Each testing compared CIS without DPT and with DPT given different level of conditioner(100, 200, 300, 400 CU) Total testing conditions consists of 26 lists, and each list has ten sentences.
Speech Perception Testing Subject # Age at Implantation (years) Age at test (years) Length of CI Use (years) Length of Deafness (years) Etiology Internal Device and Electrode Strategy 1 56 59 3.0 3 Meniere Concert FLEX28 FS4-p 2 71 76 5.0 41 Idiopathic progressive 65 69 4.0 27 4 57 62 5 Sonatati100 Standard 49 55 6.0 6 FSP 81 85 7 29 Idiopathic sudden 8 34 7.0 Sonatati100 Medium 9 64 25 Concert Standard 10 46 9.0 24 Genetic Pulsarci100 FLEX24
Speech Perception Data Analysis
Speech Perception Data Analysis Average improvement is about 30%
Speech Perception Data Analysis
Conclusion Developed a new CI stimulation strategy Implemented a GUI-based code generator to create customized CIS stimuli Conducted subject testing which showed significant improvement in speech perception in noise
Thank you!