Speech processing model training method, speech processing method, device and equipment

By introducing semantic and acoustic information in the training process of the speech processing model, the problem that traditional speech processing solutions cannot retain the speech sublingual information is solved, and the naturalness of synthetic speech is improved.

CN119763548AActive Publication Date: 2025-04-04BEIJING DAJIA INTERNET INFORMATION TECH CO LTD
View PDF 10 Cites 0 Cited by

Patent Information

Application Number
CN202411892310.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-20
Publication Date
2025-04-04
Estimated Expiration
2044-12-20

AI Technical Summary

Technical Problem

Traditional speech processing solutions cannot effectively retain sublingual information in speech, such as emotions and tone, resulting in low naturalness of synthetic speech.

Method used

By introducing semantic information and acoustic information during the training of the speech processing model, cleaner semantic information is learned and more language information is retained when processing speech. Specific methods include phonological encoding of sample speech signals, extracting semantic embeddings and acoustic embeddings, and training the model with phoneme embeddings.

Benefits of technology

The naturalness of synthetic pronunciations is improved, making the generated pronunciations closer to the natural language expression of human beings.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119763548A_ABST
    Figure CN119763548A_ABST
Patent Text Reader

Abstract

The invention provides a voice processing model training method, a voice processing method, a voice processing device and voice processing equipment, and belongs to the technical field of computers. The method comprises the following steps: performing voice coding on a sample voice signal to obtain a semantic embedding representation and an acoustic embedding representation of the sample voice signal; performing phoneme extraction and phoneme coding on a reference voice text of the sample voice signal to obtain a phoneme embedding representation of the reference voice text; and training a speech processing model based on the semantic embedded representation, the acoustic embedded representation and the phoneme embedded representation, the speech processing model being used for performing real-time speech synthesis on the input speech. According to the method, semantic information and acoustic information are introduced in the model training process, so that the model can learn cleaner semantic information, more language information can be reserved when the voice is processed, and the naturalness of the synthesized voice is improved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Speech synthesis model training method and device, equipment and medium

    CN113870827A

  • Speech synthesis model training method and speech synthesis method

    CN114387946A

  • Speech synthesis method and device

    CN116052640A

  • Speech synthesis model training method, speech synthesis method and related device

    CN116129863A

  • Model training method and device and voice generation method and device

    CN116129875A