Voice Recognition System Using Dual Grammar Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice recognition systems face increased processing time and complexity when using a garbage model that covers a wide scope, as it enlarges the search space and requires independent score calculations for each garbage model, even when many garbage models are included in the grammar.
Innovation Solution
A voice recognition system that uses two processing units: one generates a recognition hypothesis graph based on a first grammar for keyword recognition and another unit outputs a recognition result based on a second grammar that accepts non-keyword sections as garbage, sharing the structure and score from the hypothesis graph to reduce redundant calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a garbage model using all syllable models is used to recognize optional utterance, then the covering scope of the garbage model becomes large, but the processing amount is increased because the search space is enlarged
Solution Approach 1:
The patent divides the grammar into two types: keyword grammar for recognizing specific vocabulary words and garbage grammar for recognizing optional syllable strings. This segmentation allows the system to handle different recognition requirements separately, preventing the garbage model from unnecessarily expanding the search space while maintaining its covering scope for optional utterances.
Solution Approach 2:
The patent extracts the garbage recognition function from the main keyword recognition process. By using a separate garbage grammar that operates independently, the system can accept optional syllable strings without including them in the main search hypothesis space, thus maintaining recognition speed while preserving the garbage model's versatility.
2Reliability
If the garbage model is executed in a different path independently, then the grammatical restrictions are maintained, but the overall recognition time is increased when more garbage models are included
Solution Approach 1:
The patent performs preliminary action by pre-processing the garbage grammar to create a separate recognition path. This allows the garbage models to be prepared in advance without interfering with the main keyword recognition process, maintaining grammatical restrictions while reducing the time penalty associated with executing multiple garbage models.
Solution Approach 2:
The patent introduces an intermediary structure that separates the garbage recognition path from the main keyword recognition path. This intermediary allows both paths to operate independently, maintaining the reliability of grammatical restrictions while minimizing the time loss through efficient path management and avoiding redundant processing.
Data Source
AI summary
The present invention enables the recognition process at high speed even when a lot of garbage is included in the grammar. The first voice recognition processing unit generates a recognition hypothesis graph which indicates a structure of hypothesis that is derived according to a first grammar together with a score associated with respective connections of a recognition unit by executing a voice recognition process based on the first grammar to a voice feature amount of input voice, and the second voice recognition processing unit outputs the recognition result from a total score of a hypothesis which is derived according to a second grammar after executing a voice recognition process according to the second grammar that is specified to accept a section other than keywords in input voice as the garbage section to a voice feature amount of input voice, and the second voice recognition processing unit acquires the structure and the score of the garbage section from the recognition hypothesis graph.