Speech Recognition Device Derived Word Rejection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition technologies face challenges in accurately distinguishing between target words and similar words due to recognition errors, particularly when different speakers pronounce the same word differently, leading to incorrect extraction of audio segments.
Innovation Solution
A speech recognition device that employs a derived word generation rule to generate and differentiate between search target words and rejecting words, using a second recognition dictionary to set threshold values based on phoneme length and syllable length, and outputs only the segments where the search target word is spoken, excluding derived word segments as rejecting objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If speech recognition uses waveform matching to extract word portions, then the extraction process is simple, but recognition accuracy deteriorates because waveforms differ for different speakers even for the same word
Solution Approach 1:
The patent segments the recognition process into multiple stages: first extracting candidate word portions using waveform matching, then refining the results by generating and comparing derived words. This segmentation allows the system to maintain the simplicity of waveform-based extraction while adding a refinement layer that addresses speaker variability through contextual word analysis.
Solution Approach 2:
The patent introduces derived words as an intermediary element between the search target word and the final recognition result. By generating derived words that capture phonetic variations and contextual relationships, the system creates a mediator that bridges the gap between simple waveform matching and accurate recognition, allowing the system to handle speaker differences without sacrificing extraction simplicity.
2Measurement precision
If the system generates multiple rejecting words to improve recognition accuracy, then recognition accuracy improves, but device complexity increases due to multiple dictionaries and word generation processes
Solution Approach 1:
The patent merges the rejecting word generation process with the main recognition dictionary structure. Instead of maintaining separate dictionaries for search target words and rejecting words, the system generates rejecting words on-demand from the search target word using derived word generation rules. This merging approach reduces structural complexity while maintaining the ability to accurately distinguish between target and similar words.
Solution Approach 2:
The patent performs preliminary action by pre-defining derived word generation rules that can be applied to any search target word. These rules are established in advance and stored in the system, allowing the rejection process to proceed efficiently without requiring complex real-time analysis. The preliminary preparation of generation rules simplifies the overall system structure while enabling accurate word distinction.
3Measurement precision
If the system uses derived word generation rules to differentiate target and rejecting words, then discrimination accuracy improves, but processing time increases due to additional word generation and comparison steps
Solution Approach 1:
The patent applies partial action by generating and comparing only the necessary derived words needed for discrimination, rather than exhaustively analyzing all possible word variations. The system generates derived words based on phonetic rules and compares them selectively against the audio waveform, achieving sufficient discrimination accuracy without the time cost of complete exhaustive analysis. This partial approach balances accuracy and processing speed.
Data Source
AI summary
A speech recognition device includes, a speech recognition section that conducts a search, by speech recognition, on audio data stored in a first memory section to extract word-spoken portions where plural words transferred are each spoken and, of the word-spoken portions extracted, rejects the word-spoken portion for the word designated as a rejecting object; an acquisition section that obtains a derived word of a designated search target word, the derived word being generated in accordance with a derived word generation rule stored in a second memory section or read out from the second memory section; a transfer section that transfers the derived word and the search target word to the speech recognition section, the derived word being set to the outputting object or the rejecting object by the acquisition section; and an output section that outputs the word-spoken portion extracted and not rejected in the search.