Multi-Language Voice Recognition Using Context-Based Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice recognition methods for multiple languages are inefficient, requiring excessive calculations and unable to operate in real-time, especially when dealing with mixed languages, leading to difficulties in identifying non-primary languages and processing voices effectively.
Innovation Solution
An apparatus and method that segment input voices into phonemes, use a phoneme database for matching, and determine the type of non-primary language based on context information such as location, dialogue history, and user feedback, allowing for real-time recognition by applying specific language algorithms to primary and non-primary language components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If mapping is performed on all languages to recognize voice involving multiple languages, then the recognition capability is improved, but the amount of calculations rapidly increases
Solution Approach 1:
The patent segments the voice recognition process into two distinct phases: first identifying the language type using context information (location, dialogue history, user feedback), then performing phoneme matching only for the identified language type. This segmentation avoids the need to perform comprehensive mapping on all languages simultaneously, thereby reducing computational complexity while maintaining multi-language recognition capability.
Solution Approach 2:
The patent performs preliminary language type identification before executing the main voice recognition task. By using context information to determine which language type is being spoken first, the system prepares the appropriate phoneme database and recognition parameters in advance, avoiding unnecessary calculations for other language types and enabling real-time processing.
2Adaptability or versatility
If mapping is performed on all languages to recognize voice involving multiple languages, then the recognition capability is improved, but real-time processing cannot be achieved
Solution Approach 1:
The patent divides the recognition system into a language type identification module and a phoneme recognition module. The identification module uses lightweight context information processing to quickly determine language type, while the phoneme recognition module processes only the relevant language data. This segmentation enables real-time processing by avoiding the computational burden of simultaneous multi-language mapping.
Solution Approach 2:
The system performs preliminary language type identification using context information before initiating the main recognition process. This preliminary action allows the system to configure the appropriate phoneme database and recognition parameters in advance, enabling real-time voice recognition without the delay of comprehensive language mapping during the recognition phase.
3Measurement precision
If phoneme matching is performed for all languages, then recognition accuracy is improved, but computational load increases
Solution Approach 1:
The patent applies local quality by loading and performing phoneme matching only with the phoneme database corresponding to the identified language type, rather than maintaining and processing all language databases simultaneously. This approach ensures high recognition accuracy for the actual language being spoken while significantly reducing the computational load and memory requirements.
Solution Approach 2:
The system performs phoneme matching only for the partial set of phonemes relevant to the identified language type, rather than performing exhaustive matching across all languages. This partial action maintains recognition accuracy for the target language while avoiding the excessive computational burden of processing all possible language phonemes.
Data Source
Figure 1
Figure 2A~2D
Figure 3A
AI summary
A method for recognizing a voice includes receiving, as an input, a voice involving multiple languages, recognizing a first voice of the voice by using a voice recognition algorithm matched to a preset primary language, identifying the preset primary language and a non-primary language different from the preset primary language, which are included in the multiple languages, determining a type of the non-primary language based on context information, recognizing a second voice of the voice in the non-primary language by applying a voice recognition algorithm, which is matched to the non-primary language of the determined type, to the second voice, and outputting a result of recognizing the voice which is based on a result of recognizing the first voice and a result of recognizing the second voice.