Language-Segmented Speech Synthesis for Multilingual Pronunciation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current speech synthesis systems primarily support unilingual pronunciation, failing to achieve smooth multilingual pronunciation.

Innovation Solution

A method that segments text into different language categories, converts each segment into phonemes based on language-specific rules, and uses a speech synthesis model to generate vocoder parameters for accurate, natural speech synthesis, incorporating prosodic structures and tones.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a speech synthesis system supports multiple languages, then the versatility and adaptability of the system improve, but the complexity of the system increases due to the need for language-specific processing rules and models

Engineering Contradiction:
Improvemultilingual pronunciation supportVSAvoidsystem complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the text input into different language categories first, then processes each segment separately through language-specific phoneme conversion rules. This segmentation approach allows the system to handle multiple languages by dividing the complex task into manageable language-specific sub-tasks, reducing the overall system complexity while maintaining multilingual versatility.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements a universal speech synthesis framework that can handle multiple languages through a common architecture. The system uses a unified text segmentation module and a shared phoneme conversion model that adapts to different languages, allowing one system to serve multiple language requirements without requiring completely separate processing systems for each language.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Measurement precision

If the system processes text segment by segment according to language categories, then the accuracy of phoneme conversion improves, but the processing time increases

Engineering Contradiction:
Improvephoneme conversion accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary text segmentation into language categories before phoneme conversion. By pre-identifying the language type of each text segment, the system can prepare language-specific conversion rules and parameters in advance, which then enables more accurate and efficient phoneme conversion without requiring time-consuming language detection during the conversion process itself.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3937165B1Speech synthesis method and apparatus, and computer-readable storage medium
Publication Date: 2025.10.22 BEIJING JINGDONG SHANGKE INFORMATION TECH CO LTD
  • EP3937165B1 patent drawingFigure 1
  • EP3937165B1 patent drawingFigure 2A~2B
  • EP3937165B1 patent drawingFigure 3

AI summary

The present disclosure relates to a speech synthesis method and device, and a computer-readable storage medium, and relates to the field of computer technology. The method of the present disclosure includes: dividing a text into a plurality of segments according to a language category to which each of the segments belongs; converting each of the segments into a phoneme corresponding to the segment to generate a phoneme sequence of the text according to the language category to which each of the segments belongs; inputting the phoneme sequence into a speech synthesis model trained in advance and converting the phoneme sequence into a vocoder characteristic parameter; and inputting the vocoder characteristic parameter into a vocoder to generate a speech.