Speech Recognizer Grammar Parsing and Syllable Network Minimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current local speech recognition technologies are inflexible, limited to singular vocabulary and grammar formats, and lack support for multi-grammar construction and class models, making them unsuitable for environments with limited network connectivity, such as vehicle-mounted systems where quick and adaptable speech recognition is needed.
Innovation Solution
A method and system for building a speech recognizer that involves parsing grammar files, deploying syllable networks, performing merge minimization operations, and integrating an acoustic model and language model, specifically using pre-trained DNN models and N-Gram language models to enable flexible and extensible speech recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cloud-based speech recognition is used, then recognition accuracy is improved, but network dependency increases and response time worsens in environments with limited network connectivity
Solution Approach 1:
The patent extracts the core speech recognition engine from the cloud-based system and deploys it locally on the terminal device. This allows the system to maintain high recognition accuracy while eliminating network dependency and reducing response time, as the recognition process no longer requires cloud server communication.
Solution Approach 2:
The patent introduces an intermediate local recognition model that bridges the gap between cloud-based accuracy and local speed. This local model processes speech data on-device, serving as a mediator that provides both the accuracy of sophisticated recognition and the speed of local processing, without requiring network connectivity.
2Reliability
If traditional local speech recognition is used, then network dependency is reduced, but flexibility and extensibility worsen due to singular vocabulary and grammar formats
Solution Approach 1:
The patent implements a universal grammar file format that can accommodate multiple grammar types (phrase structure grammar, context-free grammar, regular expression grammar) within a single standardized framework. This allows the local recognition system to maintain network independence while achieving high flexibility and extensibility through unified grammar management.
Solution Approach 2:
The patent introduces dynamic grammar loading and switching capabilities that allow the recognition system to adapt to different application scenarios in real-time. Grammar files can be dynamically loaded, modified, and switched without system reconfiguration, enabling the local system to maintain both reliability and adaptability.
3Adaptability or versatility
If multiple grammar files are integrated, then recognition coverage is improved, but system complexity increases
Solution Approach 1:
The patent merges multiple grammar files into a unified grammar structure through automated integration algorithms. The system automatically resolves conflicts, optimizes grammar rules, and consolidates vocabulary across multiple files, achieving comprehensive recognition coverage while maintaining manageable system complexity through unified management.
Solution Approach 2:
The patent performs preliminary grammar validation, conflict detection, and optimization during the grammar file integration process. By pre-processing and preparing grammar files before they are loaded into the recognition system, the patent reduces the complexity of real-time grammar management and ensures smooth integration of multiple grammars.
Data Source
AI summary
A method and system for building a speech recognizer, and a speech recognition method and system are proposed. The method for building a speech recognizer includes: reading and parsing each grammar file, and building a network of each grammar; reading an acoustic syllable mapping relationship table, and deploying the network of each grammar as a syllable network; performing a merge minimization operation for each syllable network to form a sound element decoding network; forming the speech recognizer by using the sound element decoding network and a language model. The technical solutions of the present disclosure may be applied to exhibit strong extensibility, support an N-Gram language model, support a class model, present flexible use, and adapt for an embedded recognizer in a vehicle-mounted environment.