A computer-aided Mandarin phonemes training (CAMPT) system was developed and evaluated for training hearing-impaired students in their pronunciation of Mandarin phonemes. Deaf or hearing-impaired people have difficulty hearing their own voice, hence most of them cannot learn how to speak. Phonemes are the basis for learning to read and speak in Mandarin. Thus, the main purpose of this study is to use voice-recognition techniques to develop a Mandarin phoneme training system for the self-training of hearing-impaired people. The system analyses the spoken Mandarin phoneme of a hearing-impaired person, compares it with the phonemes database, and shows the results on the computer monitor. Our CAMPT system helps hearing-impaired people pronounce Mandarin phonemes, and the system automatically gives the learner a percentage score for each phoneme. In other words, learners can receive instant feedback from the system. Experiments conducted among high school students diagnosed with hearing impairment revealed that our system can help students learn the Mandarin phonemes with a neural net recognition algorithm for different users. After five months, most learners can pronounce 95% of the phonemes correctly.