The invention discloses a speaker speech adaptive training method and belongs to the technical field of speech synthesis. The method comprises giving training emotional speech data and target speakeremotional speech data; representing an acoustic parameter and estimating and modeling the state output distribution and the duration distribution of the acoustic parameter; normalizing a difference between the state output distribution of a training speech data model and the state output distribution of an average sound model to obtain the average sound model of the target speaker emotional speechdata; and subjecting the average sound model to speaker adaptive conversion to obtain a speaker-dependent adaptive model. The speaker speech adaptive training method of the present invention uses theobtained adaptive model for speech synthesis, can reduce the influence caused by the difference of speakers in a speech library, and improves the emotional similarity of the synthesized speech, and can synthesize emotional speech with good naturalness, fluency and emotional similarity with only a small amount of emotional corpus to be synthesized.