Hearing Audio Noise Prediction for Low-Latency Noise Subtraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hearing devices face challenges in providing satisfactory noise cancellation and speech enhancement, with a need for improved methods to effectively remove background noise and enhance speech.
Innovation Solution
A computer-implemented method for training a neural network to process audio data from microphones, using sequential training to predict future noise, and applying this network in hearing devices to estimate and subtract noise from input signals, combined with a hearing system comprising processors and transducers for improved noise cancellation and sound quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If traditional noise cancellation methods are used in hearing devices, then noise removal capability is provided, but speech enhancement satisfaction remains insufficient
Solution Approach 1:
The patent segments the audio signal processing into distinct components: separating the mixture signal into speech components and noise components through spectral analysis, then processing them independently through different filtering operations. This allows optimized noise cancellation without compromising speech enhancement quality.
Solution Approach 2:
The patent employs spectral processing techniques that analyze and modify frequency parameters of the audio signal. By transforming the signal into the frequency domain and applying frequency-selective filtering, the system achieves effective noise removal while preserving speech characteristics through parameter-based differentiation between noise and speech spectra.
2Object-affected harmful factors
If improved noise cancellation algorithms are implemented, then noise removal effectiveness increases, but processing time and latency increase
Solution Approach 1:
The patent performs preliminary spectral analysis and noise characterization in advance, creating noise profiles and spectral templates before the actual noise cancellation operation. This pre-processing enables faster real-time noise removal by avoiding complex calculations during critical audio processing moments.
Solution Approach 2:
The patent replaces computationally intensive traditional noise cancellation algorithms with spectral processing techniques and neural network-based models that can achieve similar or better performance with reduced computational complexity, thereby lowering processing latency while maintaining effectiveness.
3Measurement precision
If complex neural network training methods are used, then noise prediction accuracy improves, but training time and computational resources increase
Solution Approach 1:
The patent uses copy-based training strategies where multiple simplified neural network models are trained on different subsets of training data and then aggregated. This approach achieves high prediction accuracy through ensemble methods while reducing the computational burden on any single model and enabling more efficient training processes.
Solution Approach 2:
The patent employs partial training strategies where neural networks are trained on representative subsets of audio data rather than requiring complete datasets. By using partial training with carefully selected training samples, the system achieves sufficient prediction accuracy with significantly reduced training time and computational resources.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A hearing device, a hearing system, and related methods are disclosed, the hearing device comprising a set of input transducers providing a transducer input, the set of input transducers comprising a first input transducer for provision of a first transducer input signal, wherein the transducer input is based on the first transducer input signal; a processor configured to process the transducer input and provide an electrical output signal based on the transducer input; and a receiver for converting the electrical output signal to an audio output signal, wherein to process the transducer input comprises to process the transducer input for provision of a first input and a second input; apply a neural network to the second input for provision of a second output being a signal that is a prediction of future noise; provide a noise estimate based on the second output; and subtract the noise estimate from a first input magnitude of the first input for provision of a first output, wherein to provide the electrical output signal is based on the first output.