Hearing Aid Noise Reduction via Neural Network Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current hearing assistance devices often fail to provide reliable noise reduction and improved sound quality due to limitations in using limited microphone information and ineffective binary classification in time/frequency domains, leading to poor speech intelligibility in varying sound environments.
Innovation Solution
A system that records background noise and speech using a hearing assistance device or external device, performs supervised training on a binary classifier with programmable feature extraction, and processes sound in real-time to adapt to changing environments, allowing for improved speech intelligibility and quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional binary classification in time/frequency domains is used for noise reduction, then device complexity is reduced, but speech intelligibility and noise reduction effectiveness deteriorate
Solution Approach 1:
The patent replaces traditional mechanical signal processing methods (binary classification in time/frequency domains) with a neural network-based machine learning system. The neural network automatically learns optimal noise reduction strategies from training data, substituting manual feature engineering and classification rules with adaptive, data-driven processing that achieves superior speech intelligibility while managing computational complexity through efficient network architectures.
2Adaptability or versatility
If hearing aids are programmed for individual use in specific environments, then adaptability to that environment improves, but reliability in changing environments deteriorates
Solution Approach 1:
The patent implements dynamic adaptability through a neural network that can continuously learn and update its parameters in response to changing environmental conditions. The system transitions from static, pre-programmed settings to dynamic, real-time adaptation by processing incoming audio data through the trained neural network, allowing the hearing aid to automatically adjust to new sound environments while maintaining optimal performance across varying conditions.
Solution Approach 2:
The patent applies preliminary action by pre-training the neural network offline using extensive training datasets that represent various sound environments. This offline training phase prepares the network with prior knowledge of different acoustic conditions, enabling the hearing aid to quickly adapt to new environments during online operation without requiring extensive real-time learning, thus ensuring reliable performance across changing conditions.
3Measurement precision
If more processing is applied to improve noise reduction, then speech intelligibility improves, but processing delay increases
Solution Approach 1:
The patent applies partial action by implementing noise reduction processing selectively based on the acoustic environment and signal characteristics. The neural network identifies situations where noise reduction is most beneficial and applies processing accordingly, rather than uniformly processing all input signals. This selective approach maintains speech intelligibility in noisy conditions while minimizing processing delay in quiet environments or when noise reduction is less critical.
Data Source
AI summary
A system is provided for training and improvement of noise reduction in hearing assistance devices. In various embodiments the system includes a hearing assistance device having a microphone configured to detect sound. A memory is configured to store background noise detected by the microphone and configured to store a previous recording of speech. A processor includes a training module coupled to the memory and configured to perform training on a binary classifier using programmable feature extraction applied to a sum of the speech and the noise. The processor is configured to process the sound using an output of the binary classifier.