Chinese Speech Pitch Recognition Of Cochlear Implant In Noise

Cochlear implant (CI) is the only medical prosthesis which can help severedeaf people to regain the hearing currently. With the rapid development ofintegrated circuit and speech signal processing technology, the performance of CIwas also gradually improved and perfected. More and more hearing loss patients getbetter auditory quality with the help of CI. The CI is an equipment that replaced theimpaired sensory hair cells. The CI processes the speech signal first, then theelectrical stimulations are sent to the implanted electrodes. In recent years Chinesehearing-impaired people also began to benefit from cochlear along with theeconomic development of China. Although the western countries had a maturetheoretical basis and clinical experience in the cochlear implant research already,it’s not suitable for our country. Because Mandarin is tonal language, the pitch has avery important influence for the semantic. So it is more difficult to process andcoding the speech signal in China then western developed countries. Anotherunavoidable interference for the CI patients’ speech perception in daily life is thenoise. Above all, study in the influence factors of the mandarin speech intelligibilityin CI and the ability of the patients’ production is very important for the advantageand improvement of CI in our country.This paper analysed the Mandarin recognition capability and the productioncapability of CI users. This has reference value for the improvement of the CIperformance and CI users’ lives quality. This paper analysed the influence factors ofmand arin speech sentence and pitch recognition, so as the pitch detectionalgorithm. The pitch information is a necessary reference for the tone analyses.The present study discussed the influencing factors of cochlear speechintelligibility in noisy environments by a simulation experiment. Mandarin sentenceand tone stimuli were processed and encoded by simulated processer of CI. Thenthe stimuli was synthetized by noise-vocoder, and identified by the normal hearinglisteners. The effects of excitation spread on the mandarin speech intelligibility ofCI in the noise was analysed in this experiment by changed the excitation spreadand the signal-noise ratio.As the primary means of tone characteristic discriminant, the fundamentalfrequency detection algorithm was improved. This paper presented two improvedpitch detection method based on classic fundamental frequency extraction method.The first method combined the average magnitude difference function with the autocorrelation function so that the peak point at the fundamental frequency cycleturned more prominent and sharper to improve the noise immunity. The second method used an inverse filter based on linear prediction coefficient to filtering thechannel response and noise in the speech signal. And then the cepstrum was used tocalculate fundamental frequency cycle of the filtered signal.Two objective assessment approaches were used to assess the tone productionability of cochlear implant children. The perceptual evaluation of speechquality score was used to assess the four tones’ speech quality, and the pitch track ofthe recorded speech material was used to analyzed and summarized thecommon errors of tone production.