Developmental Research of Speech Processor for Cochlear Implant transmitting high quality speech

Research Project

Project/Area Number	09650419
Research Category	Grant-in-Aid for Scientific Research (C)
Allocation Type	Single-year Grants
Section	一般
Research Field	情報通信工学
Research Institution	Kumamoto University
Principal Investigator	WATANABW Akira Faculty of Engineering, Department of Computer Science, Kumamoto University, Professor, 工学部, 教授 (50040382)
Co-Investigator(Kenkyū-buntansha)	IKEDA Takashi Kurume National College of Technology, Associate Professor, 助教授 (80222884) IKEDA Naomitsu Yatsushiro National College of Technology, Associate Professor, 助教授 (70124146) UEDA Yuichi Faculty of Engineering, Department of Computer Science, Kumamoto University, Professor, 工学部, 助教授 (00141961)
Project Period (FY)	1997 – 1999
Project Status	Completed (Fiscal Year 1999)
Budget Amount *help	¥3,400,000 (Direct Cost: ¥3,400,000) Fiscal Year 1999: ¥200,000 (Direct Cost: ¥200,000) Fiscal Year 1998: ¥1,800,000 (Direct Cost: ¥1,800,000) Fiscal Year 1997: ¥1,400,000 (Direct Cost: ¥1,400,000)
Keywords	Cochlear implant / Speech Processor / Formant estimation / Inverse filter control / hearing test / スピーチプロセッサの改良 / パラメータ推定方式 / 逆フィルタ制御法 / ホルマント推定精度 / 精度 / 主観テスト / ホルマント抽出 / 逆フィルタ法 / ホルマント分析・合成
Research Abstract	This research aims at proposing a new speech processor to be used for a cochlear implant system. We investigated possibility of adapting the formant estimation method using inverse filter control, which we developed, to the speech processor. The results are as follows (1) The formant estimation system, which we devised for the hearing aid for telephonic speech, is much superior to the proper one according to visual inspection of the formant trajectories and hearing tests of speech signals synthesized by sinusoidal waves. (2) Assuming that pitch pulse trains excite critical band filters , we conducted the hearing tests using /VCV/syllables that were synthesized by adding the response waves of the band pass filters with Q=5 to pitch pulse trains. When the center frequencies of the Butterworth filters were quantized into equidistant 22 positions on Mel-scale, articulation score of consonants was 64 [%]. (3) We have proposed a new modified system in which higher zeros within 4[kHz] are directly controlled by pitch. After setting the parameters of the new system to be best, the accuracy to estimate formant frequencies have been examined using isolated vowels uttered by 100 males and 100 females. The error rates were 4.9[%] and 8.8 [%] in male and female voice respectively. Based on the results, we conclude that the system will be useful for improvgnt of the speech processor.