Hearing Device Neural Network Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hearing devices struggle to provide personalized sound processing that effectively adapts to individual users' acoustic environments and speech patterns, leading to suboptimal hearing experiences.
Innovation Solution
A hearing device equipped with a neural network processor that implements a detector using a neural network with optimized node parameters, allowing for personalized sound processing by selecting the best performing neural network from a set of pre-trained candidates based on user-specific data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional hearing devices use fixed sound processing algorithms, then device complexity is reduced, but adaptability to individual users' acoustic environments and speech patterns deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-training multiple neural network models with different node parameter sets before deployment in the hearing device. During a fitting session, these pre-trained models are evaluated against user-specific data (recorded speech and acoustic environments), and the best-performing model is selected and installed. This approach allows the device to achieve high adaptability without requiring complex real-time training capabilities, as the selection process uses simplified performance metrics rather than full retraining.
2Measurement precision
If hearing devices implement personalized neural networks, then speech recognition accuracy is improved, but computational resources and processing time increase
Solution Approach 1:
The patent performs the computationally intensive neural network training in advance, creating multiple pre-trained models with different node parameter configurations. During the fitting session, instead of training from scratch, the system evaluates these pre-trained models against user-specific data and selects the best performer. This dramatically reduces processing time while maintaining high speech recognition accuracy, as the selection process requires only forward propagation through existing networks rather than full training iterations.
3Reliability
If multiple pre-trained neural network candidates are evaluated, then personalized hearing performance is improved, but device complexity and memory requirements increase
Solution Approach 1:
The patent segments the neural network parameter space by creating multiple candidate models, each with different node parameter sets trained on different subsets of data or with different architectural configurations. This segmentation allows the system to evaluate diverse approaches and select the best one for each user, improving personalized performance while keeping individual model sizes manageable. The fitting software organizes these candidates in a structured manner, evaluating them systematically against user-specific criteria.
Data Source
AI summary
A hearing device comprises an input transducer comprising a microphone for providing an electric input signal representative of sound in the environment of the hearing device, a pre-processor for processing electric input signal and providing a multitude of feature vectors, each being representative of a time segment thereof, a neural network processor adapted to implement a neural network for implementing a detector configured to provide an output indicative of a characteristic property of the at least one electric input signal, the neural network being configured to receive said multitude of feature vectors as input vectors and to provide corresponding output vectors representative of said output of said detector in dependence of said input vectors. The hearing device further comprises a transceiver comprising a transmitter and a receiver for establishing a communication link to another part or device or server, at least in a particular adaptation-mode of operation, and a selector for—in said particular adaptation-mode of operation—routing said feature vectors to said transmitter for transmission to said another part or device or server, and—in a normal mode of operation-to route said feature vectors to said neural network processor for use as inputs to said neural network, a neural network controller connected to said neural network processor for—in said particular adaptation—mode of operation-receiving optimized node parameters, and to apply said optimized node parameters to said nodes of said neural network to thereby implement an optimized neural network in said neural network processor, wherein the optimized node parameters have been selected among a multitude of sets of node parameters for respective candidate neural networks according to a predefined criterion in dependence of said feature vectors. A method of selecting optimized parameters for a neural network for use in a portable hearing device is further disclosed. The invention may e.g. be used in hearing aids or headsets, or similar, e.g. wearable, devices.