A hearing aid was proposed previously using single resonant analysis for telephonic speech. First, using this hearing aid, experiments by normal-hearing persons with masking noise as a simulation of a hearing threshold level were done and showed satisfactory results. Next, experiments by two hearing-impaired persons were done to examine the effectiveness of the hearing aid. The usefulness of this hearing aid for vowels and consonants was ascertained. In some consonant samples, in spite of different consonants, the subjects felt they were the same as in the case of the linearly amplified hearing aid. In the case of a hearing aid by single resonant analysis, however, the samples tend to sound different to the subjects. Since the subjects were not used to the sounds of the hearing aid, an assistant system was necessary. An audio--visual system is proposed that can make phoneme cues for the hearing-impaired person by using vision together with hearing. It will help subjects to grasp correspondence between sound image and phonemes. The audio--visual system and training by the system will be discussed.