Personalized Phoneme Mapping for Voice Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generic speech recognition models are inaccurate for individuals with heavy accents or speech impediments, and are computationally intensive, often requiring powerful servers for voice calibration and analysis, which can be resource-intensive and inefficient.
Innovation Solution
A video game system that determines a personal phoneme mapping for each user by having them read a grammar training set, adjusting the mapping until a confidence threshold is met, and storing it for improved voice recognition, even on resource-constrained devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If generic speech recognition models are used, then device complexity is reduced, but speech recognition accuracy deteriorates for users with heavy accents or speech impediments
Solution Approach 1:
The system performs preliminary voice calibration by having the user read a grammar training set before actual voice recognition tasks. This preliminary action creates a personalized phoneme mapping that improves accuracy for subsequent recognition without adding complexity during the actual recognition process
Solution Approach 2:
The system automatically determines personalized phoneme mappings by analyzing the user's speech sample against the grammar training set and iteratively adjusting the mapping until a confidence threshold is met. This self-service approach creates custom recognition models without requiring manual configuration or complex server processing
2Reliability
If voice calibration and voice recognition are performed on powerful servers, then speech recognition accuracy is improved, but processing speed and responsiveness deteriorate
Solution Approach 1:
The computationally intensive voice calibration process is performed as a preliminary action before actual recognition tasks. The system determines personalized phoneme mappings during calibration and stores them for rapid retrieval during subsequent voice recognition, eliminating the need for complex server processing during real-time recognition
Solution Approach 2:
The system extracts and stores the personalized phoneme mapping results from the complex calibration process, separating the heavy computational work from the actual recognition operation. This allows the lightweight stored mapping to be used for rapid recognition without requiring powerful servers during real-time use
3Reliability
If iterative phoneme mapping adjustment is performed until confidence threshold is met, then speech recognition accuracy is improved, but processing time increases
Solution Approach 1:
The system implements feedback by determining a confidence score after each phoneme mapping adjustment and using this feedback to decide whether to continue iterating. When the confidence score meets the predefined threshold, the process stops, avoiding unnecessary additional processing time while ensuring adequate accuracy
Solution Approach 2:
The system performs partial action by stopping the iterative adjustment process when the confidence threshold is met, rather than continuing to refine the phoneme mapping indefinitely. This partial completion approach achieves sufficient accuracy without the excessive time cost of exhaustive optimization
Data Source
AI summary
The present disclosure provides a voice recognition system configured to generate a custom phoneme mapping for a user. The voice recognition system can analyze a user speech sample of a grammar training set in order to generate the custom phoneme mapping. The custom phoneme mapping can be used for subsequent recognition of the user's voice within an application.