Shim,YJ, Kim HN, Chung MH, Kim SG (2002) A new method of consonant classification and a new strategy of speech correction for congenitally deaf implantees. In: Kubo T, Takahashi Y, Iwaki T (eds) Cochlear Implants: an update, pp. 421–426. The Hague, The Netherlands: Kugler Publications. Address correspondence to: Yoon-Joo Shim, Department of Otolaryngology, Yonsei University, College of Medicine, 134 Shinchon-dong, Seodaemun-gu, Seoul 120-752, Korea. Tel: +82 2 361 8597. Fax: +82 2 393 0580. Email: mtyjs@yumc.yonsei.ac.kr CAI system for testing and training discrimination ability of change in voice pitch through the cochlear implant SHIZUO HIKI, KAZUYA IMAIZUMI, Graduate School of Human Sciences, Waseda University, Japan LIMING YANG, Institute of Language Teaching, Waseda University, Japan MASAE SHIROMA, International University of Health and Welfare, Japan YUMIKO FUKUDA, National Rehabilitation Center for the Disabled, Japan Objective In the cochlear implant, change in voice pitch is transmitted through the periodicity of change in density of neural pulse trains excited by the electrical signal. As the trans- mission quality of the voice pitch depends on the setting of the processor of each device, it is necessary to measure precisely the discrimination ability of change in voice pitch for each user, and to make maximum use of it. For this purpose, a CAI (Computer Assisted Instruction) system for testing and training the discrimination was developed. Structure of system In this system, stimuli having an arbitrary degree of contrast between each pair of original speech samples to be discriminated were prepared, by interpolating between the acoustical properties of the pair in the speech analysis-synthesis program STRAIGHT (Kawahara, 1999). Subjects, while listening to the stimuli, can watch their sound waveforms and/or pitch patterns on the display screen in training mode. In testing mode, response can be reported by pointing at the candidate pitch patterns, pictures or letters shown on the touch screen. The CAI algorithm was designed to minimize the number of steps of presenting stimuli before obtaining a reliable result of testing or achieving the intended goal of training, by referring to the history of response by the subject and choosing the stimulus having the most suitable degree of difficulty in discrimination for the next stimulus. Stimulus words Currently, groups of homophonous words with different Chinese syllabic tones or Japanese word accent types are installed as the stimuli in this system. The change in voice pitch plays most important role for discriminating those homophonous words. The word lists were edited systematically by considering their occurrence frequencies and difficulties of discrimination, especially for child subjects. 140 4th Congress of Asia Pacific Symposium on Cochlear Implant