Adaptive Speech Recognition Hearing Threshold Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining hearing thresholds rely on pure-tone testing, which is inefficient and inaccessible, especially for consumers wanting to purchase hearing aids or personal sound amplification products (PSAPs) online, as they require clinical evaluations, increasing costs and reducing accessibility.
Innovation Solution
Development of an adaptive speech recognition test that uses multiple regression statistical methods to convert speech recognition test scores into hearing threshold values at various frequencies, allowing for self-administered testing without pure-tone evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pure-tone testing is used to determine hearing thresholds, then measurement precision is improved, but device complexity and ease of operation deteriorate due to requiring specialized equipment and clinical environments
Solution Approach 1:
The patent replaces the mechanical/acoustic testing system (pure-tone audiometer requiring sound-treated booth) with a speech recognition-based system using ordinary audio equipment. Speech recognition scores are obtained through computer-based testing and converted to hearing threshold estimates, eliminating the need for specialized acoustic equipment and controlled environments.
Solution Approach 2:
The patent creates a computational model that copies the functionality of pure-tone threshold measurement through speech recognition data. By developing regression equations that map speech recognition scores to hearing threshold values, the system replicates the clinical measurement process using alternative, less complex means.
2Measurement precision
If pure-tone testing is used to determine hearing thresholds, then measurement precision is improved, but productivity deteriorates due to time-consuming clinical procedures
Solution Approach 1:
The patent performs preliminary speech recognition testing that can be administered quickly before or instead of the full pure-tone audiometric evaluation. This preliminary assessment provides sufficient information for initial hearing aid fitting decisions, reducing the time required for complete hearing threshold determination in clinical practice.
Solution Approach 2:
The patent changes the measurement parameter from pure-tone detection threshold to speech recognition score, which can be obtained more rapidly. By converting speech recognition scores to hearing threshold estimates through regression equations, the system maintains measurement utility while improving testing efficiency and reducing clinical time requirements.
3Measurement precision
If pure-tone testing is used to determine hearing thresholds, then measurement precision is improved, but ease of operation deteriorates due to requiring trained professionals and specialized environments
Solution Approach 1:
The patent enables self-administered speech recognition testing that does not require trained audiologists to administer. Individuals can complete the speech recognition portion of the assessment independently using computer-based interfaces, with results automatically converted to hearing threshold estimates, making the service more accessible and reducing dependency on clinical personnel for the initial assessment.
4Ease of operation
If conventional speech recognition scores are used, then ease of operation is improved, but measurement precision deteriorates because scores are not highly correlated with hearing thresholds
Solution Approach 1:
The patent incorporates feedback from scaled speech recognition performance to adjust and refine hearing threshold predictions. By analyzing the difficulty of speech recognition items and the respondent's performance pattern, the system provides feedback that improves the accuracy of converting speech scores to hearing threshold estimates, addressing the low correlation problem through adaptive measurement.
Solution Approach 2:
The patent transforms conventional speech recognition scores into hearing threshold predictions through mathematically-derived regression equations. By changing the parameter representation from raw speech recognition percentages to estimated hearing threshold levels in decibels, the system creates a more precise mapping between speech performance and actual hearing sensitivity.
Data Source
AI summary
A method for determining hearing thresholds in the absence of pure-tone testing includes an adaptive speech recognition test which includes a calibrated item pool, each item having a difficulty value (di) and associated standard error (se). Test items are administered to a group of individuals having mild to moderately severe hearing loss to obtain responses. The responses are subjected to multiple regression statistical methods to develop an equation for converting an individual's test score to hearing threshold values at several frequencies. The adaptive speech recognition test can be administered to an individual and their score can be converted into hearing thresholds at each of the several frequencies by use of the equation.