Model training methods, devices, electronic equipment, computer-readable storage media, and computer program products

By aligning audio feature sequences and phoneme feature sequences and configuring ignore labels, structured data is generated, and the model parameters are updated using data without configured ignore labels during training. This solves the problem of efficiency and accuracy in processing multidimensional feature data and improves the training efficiency and accuracy of audio synthesis models.

CN121983024BActive Publication Date: 2026-07-03TENCENT TECHNOLOGY (SHENZHEN) CO LTD
3 Cites 0 Cited by

Patent Information

Application Number
CN202610439270.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-04-03
Publication Date
2026-07-03
Estimated Expiration
2046-04-03

AI Technical Summary

Technical Problem

In model training scenarios involving large-scale data samples and multimodal joint modeling, it is challenging to balance the processing efficiency of multidimensional feature data with the accuracy of sequence feature representation. Existing technologies have relatively low training efficiency for audio synthesis models.

Method used

Audio feature sequences and phoneme feature sequences are extracted, aligned, and configured with ignore flags. Structured data is generated and stored. During training, the model parameters are updated directly using the data without ignore flags, thus decoupling the audio feature extraction and model training processes.

Benefits of technology

It improves the training efficiency and accuracy of audio synthesis models, simplifies the data processing flow during the training phase, and eliminates the interference of invalid features on model training.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

This application provides a model training method, apparatus, electronic device, computer-readable storage medium, and computer program product. The method includes: extracting audio feature sequences from raw audio data; determining a first phoneme feature sequence based on the raw audio data and text data; aligning the first phoneme feature sequence based on the audio feature sequence to obtain a second phoneme feature sequence; configuring an omission flag in the second phoneme feature sequence based on text data to obtain a third phoneme feature sequence; structurally associating the audio feature sequence and the third phoneme feature sequence to obtain structured data; storing the structured data in a target storage file; and upon receiving a training instruction, reading the structured data from the target storage file and updating the model parameters of the audio synthesis model based on the data in the structured data that does not have an omission flag configured, thereby obtaining the trained audio synthesis model. This application can improve the training efficiency and accuracy of audio synthesis models.
Need to check novelty before this filing date? Find Prior Art