Model training method, speech synthesis method, device and computer program product

Abstract

The invention relates to the field of intelligent voice, and provides a model training method, a voice synthesis method, equipment and a computer program product. According to the invention, the training efficiency of the character phoneme conversion model can be improved. The method comprises the following steps: acquiring a text sequence containing target characters with various pronunciations and an audio material of pronunciations corresponding to the text sequence from a training database, extracting a phoneme sequence corresponding to the audio material through a speech recognition model, and taking the phoneme sequence as a phoneme sequence labeling result, and inputting the text sequence into a to-be-trained character phoneme conversion model to obtain a phoneme sequence prediction result output by the to-be-trained character phoneme conversion model, then comparing the phoneme sequence labeling result with the phoneme sequence prediction result to obtain a phoneme sequence prediction deviation, and adjusting parameters of the to-be-trained character phoneme conversion model according to the phoneme sequence prediction deviation. And when a training completion condition is satisfied, obtaining a character phoneme conversion model.

Description

technical field [0001] The present application relates to the field of intelligent speech technology, in particular to a model training method, speech synthesis method, device and computer program product. Background technique [0002] Speech recognition and speech synthesis are currently widely used intelligent speech technologies. Speech recognition can convert human speech content into text, and speech synthesis is the reverse process, that is, input a piece of text to generate corresponding speech audio. [0003] For speech synthesis, when a text is entered, a speech synthesis system converts it into a sequence of phonemes and generates audio from this sequence of phonemes. Among them, for polyphonic characters, although the text-to-phoneme conversion model in the speech synthesis system will refer to word segmentation results and context information to determine the pronunciation of the text to obtain the corresponding phoneme, but the conversion error will still occur....

AI Technical Summary

Problems solved by technology

[0004] In order to find the conversion error of the text-to-phoneme conversion model and train the model to optimize it, the traditional technology mainly adopts the method of artificial exhaustion, firstly, the corresponding text is manually created according to the target text, and the text in the text is artificially marked with phonemes Obtain model training samples, so as to use these samples to train the model, but this method is prone to the problem of low model training efficiency due to the scarcity of artificial syllable-labeled training samples, which affects the accuracy of model conversion

Images