Hearing Instrument Fitting via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The initial fitting process for hearing instruments often results in unsatisfactory settings due to its reliance on audiogram-based gain prescriptions, leading to time-consuming and error-prone iterative refinements, and may not account for the user's specific acoustic environment, resulting in user dissatisfaction and potential returns.
Innovation Solution
A processing system generates training data from post-fitting adjustments and user profiles to train a machine learning model, which suggests initial fitting settings that better accommodate individual hearing needs and environmental factors, reducing the need for subsequent adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If audiogram-based gain prescriptions are used for initial fitting, then the fitting process is simple and quick, but the accuracy and user satisfaction are insufficient
Solution Approach 1:
The system performs preliminary actions by collecting post-fitting adjustment data and user profile information before the initial fitting process. This data is stored and processed to train machine learning models that will guide future initial fittings, allowing the system to learn from past adjustments and improve initial settings accuracy without adding time to the actual fitting process.
Solution Approach 2:
The system implements feedback by systematically collecting data from post-fitting adjustments made by professionals and user satisfaction outcomes. This feedback loop trains machine learning models that continuously improve the accuracy of initial fitting predictions, enabling the system to incorporate lessons from past fittings into future recommendations.
2Device complexity
If traditional initial fitting methods are used, then the process is straightforward, but multiple post-fitting adjustments are required
Solution Approach 1:
The system replaces the manual, experience-based mechanical process of traditional fitting with an automated machine learning system. The ML model processes user profiles and acoustic environment data to generate initial fitting settings, substituting the manual adjustment process with an intelligent algorithm that learns from historical data to improve reliability.
Solution Approach 2:
The system changes the parameters used for fitting by incorporating acoustic environment characteristics and user profile data alongside traditional audiogram information. This multi-parameter approach allows the ML model to generate more reliable initial settings that account for the specific conditions in which the user will actually use the hearing instrument.
3Reliability
If settings are optimized for one acoustic environment, then performance is good in that environment, but performance deteriorates in other acoustic environments
Solution Approach 1:
The system achieves universality by training the machine learning model to handle multiple acoustic environments simultaneously. The model processes user profiles that include information about various acoustic environments the user encounters, enabling it to generate initial fitting settings that are optimized for multiple conditions rather than a single environment, thus improving both reliability and adaptability.
Solution Approach 2:
The system adds another dimension to the fitting process by incorporating acoustic environment characteristics as additional input parameters. This transforms the fitting problem from a single-point optimization to a multi-dimensional optimization that considers various acoustic conditions, allowing the ML model to generate settings that perform well across different environments.
Data Source
AI summary
A method for fitting a hearing instrument comprises generating training data based on post-fitting adjustments made to settings of a plurality of hearing instruments and based on profiles of users of the plurality of hearing instruments, wherein the post-fitting adjustments are made to the settings of the plurality of hearing instruments after initial uses of the plurality of hearing instruments. The method further comprises training a machine learning (ML) model based on the training data to generate initial fitting suggestions. The method also comprises, prior to an initial use of a current hearing instrument by a current user, generating an initial fitting suggestion for the hearing instrument of the current user by applying the ML model to input that includes a profile of the current user.