In this paper, the pitch mapping method is used to synthesize the teaching voice with the same tone and the tone changed into the standard voice tone by selecting the appropriate standard voice, which reduces the information redundancy and interference caused by complex changes other than the tone information in the voice signal. Taking Chinese disyllabic tone synthesized speech as experimental material, the tone training experiment was carried out for Japanese subjects. The training results show that the synthetic speech method is superior to the standard speech method in the relative improvement rate of tone perception and output and the effect of generalized output, and is far better than the untrained control group. Most of the experimental results are statistically significant. The experimental results prove that the human brain has selective attention mechanism in phonetic learning, which provides experimental and theoretical basis for integrating synthetic speech method into computer-aided Chinese tone teaching system.