Method of generating fundamental frequency trajectory of speech for speech synthesis

摘要

The present invention relates to a method for generating a fundamental frequency trajectory for speech synthesis, wherein the average fundamental frequency of phonemes for each firing point position is determined by the pattern of the firing pattern, the average fundamental frequency of the morphemes by word type, the word pattern, the accent reflection degree of the speech phrase, In the method of generating a basic frequency trajectory of a speaker for converting a given sentence into a voice when the degree of reflection of the preceding phoneme in the speaker is called a prior phone coefficient, the first speaker pattern, the first word pattern, the first word pattern, Extracting the first accent coefficient and the first preceding phoneme coefficient; And a first speech pattern and a first speech pattern extracted by dividing a given sentence into units of speech phrases, inputting phonemes constituting each speech phrase, lengths of speech phrases, morpheme information of words constituting the speech phrases, and accent positions. A second step of generating a basic frequency trajectory of the whole speech phrase according to the second speech pattern, the second word pattern, the second accent coefficient, and the second leading phone coefficient determined from the first accent coefficient and the first leading phone coefficient; It includes.;According to the present invention, since the data necessary for generating a basic frequency trajectory such as a crater pattern, a word pattern, an influence factor of an accent, and an influence coefficient of a preceding phoneme are directly extracted from the voice corpus of the talker, the voice characteristics of the talker are well reflected. The tone of the talker can be maintained.