Speech Recognition False Accept Reduction via Parameter Consistency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems often generate false accepts due to inconsistencies in duration, energy, and pitch parameters, which are not effectively validated across a broader context, leading to inaccurate recognition results, especially in resource-constrained handheld devices.
Innovation Solution
The method involves processing spoken utterances to determine the consistency of duration, energy, and pitch parameters across component sounds, validating recognition results based on these consistencies, and using a combined score to reject inconsistent results, thereby reducing false accepts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If speech recognition systems process spoken utterances without validating parameter consistencies, then processing speed is maintained, but false accepts increase and recognition accuracy deteriorates
Solution Approach 1:
The system performs preliminary analysis of duration, energy, and pitch parameters during the speech processing stage, before final recognition validation. By pre-computing these parameter consistencies and storing them for later validation, the system avoids re-processing during the validation phase, thus improving recognition accuracy without proportionally increasing overall processing complexity.
Solution Approach 2:
The speech recognition system is divided into distinct functional modules: one module processes and extracts duration, energy, and pitch parameters from spoken utterances, while another module validates recognition results against these pre-computed parameters. This segmentation allows each module to specialize, improving overall reliability while keeping individual module complexities manageable.
2Reliability
If speech recognizers generate output without validating against parameter consistencies, then processing speed is maintained, but false accepts occur when words or phrases are incorrectly identified
Solution Approach 1:
The system computes and stores duration, energy, and pitch parameter consistencies during the initial processing of spoken utterances, before recognition validation is needed. This preliminary computation ensures that when validation is required, the system can quickly compare recognition results against pre-computed parameters without time-consuming re-analysis, thereby reducing false accepts with minimal validation time penalty.
Solution Approach 2:
The system replaces complex real-time acoustic analysis during validation with simpler comparisons against pre-computed parameter ranges and consistency metrics. Instead of re-performing full speech analysis during validation, the system substitutes this with efficient numerical comparisons, significantly reducing validation time while maintaining false accept reduction capability.
Data Source
AI summary
Embodiments of the present invention improve methods of performing speech recognition. In one embodiment, the present invention includes a method comprising receiving a spoken utterance, processing the spoken utterance in a speech recognizer to generate a recognition result, determining consistencies of one or more parameters of component sounds of the spoken utterance, wherein the parameters are selected from the group consisting of duration, energy, and pitch, and wherein each component sound of the spoken utterance has a corresponding value of said parameter, and validating the recognition result based on the consistency of at least one of said parameters.