Earphone ANC With Adaptive ESN Filtering for Leakage Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ANC earphones face limitations in noise cancellation bandwidth and amount due to non-ideal linear filters, requiring extensive testing and resource-intensive RNNs, which affect user experience and music quality, especially in dual-driver designs.
Innovation Solution
An ANC system using a system-on-chip with a processor configured to filter noisy signals with a trained echo state network (ESN) and retrain when conditions change, enabling adaptive filtering without excessive resource consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a linear filter is used to fit the primary path system function, then the filtering process is simple, but the noise cancellation bandwidth and noise cancellation amount are limited due to non-linear effects of physical devices
Solution Approach 1:
The patent transitions from linear filter parameters to neural network parameters, changing the mathematical model from linear to non-linear to better fit the physical device characteristics and improve noise cancellation performance
Solution Approach 2:
The patent replaces the traditional linear filtering mechanism with a neural network-based filtering mechanism, substituting a simple mathematical operation with a more complex adaptive system that can capture non-linear relationships
2Reliability
If Recurrent Neural Networks (RNNs) are used for filtering, then non-linear effects can be fitted, but the network scale is very large, occupying excessive resources and consuming high power
Solution Approach 1:
The patent extracts and retains only the essential reservoir computing components while removing the complex learning mechanisms of full RNNs, keeping the beneficial non-linear fitting capability while reducing network scale and resource consumption
Solution Approach 2:
The patent uses a simplified ESN model that requires fewer computational resources and can be trained more efficiently, making it suitable for resource-constrained earphone devices
3Manufacturing precision
If extensive testing and experiments are conducted to optimize the earphone cavity, then the nonlinearity and abrupt changes in spectrum and phase can be minimized, but the optimization process becomes extremely difficult and time-consuming
Solution Approach 1:
The patent enables the system to automatically adapt to changes in wearing conditions through online retraining of the ESN model, eliminating the need for extensive manual testing and optimization of the earphone cavity
Solution Approach 2:
The patent makes the filtering system dynamic and adaptive through online retraining capabilities, allowing the model to adjust to changing conditions without requiring physical re-optimization of the device
4Measurement precision
If the data sampling rate during ANC operation is increased, then the noise cancellation accuracy improves, but the RNN-based approach consumes excessive power
Solution Approach 1:
The patent changes the computational parameters by using ESN instead of full RNN, which reduces the computational load per sample and allows for higher sampling rates without excessive power consumption
Data Source
AI summary
The present disclosure relates to an ANC system for an earphone, a noise cancellation method, and a storage medium. The ANC system comprises a system-on-chip comprising a processor. The processor is configured to: obtain a noisy signal collected by a feedforward microphone; filter the noisy signal using a trained neural network filter containing an ESN to output a filtered first signal, and retrain the ESN when a first preset condition is met, wherein the first preset condition at least includes a change in a wearing manner or a leakage amount of the earphone during use; and generate a noise cancellation signal based on the first signal, the noise cancellation signal is adapted to be played via a speaker to cancel residual noise of the earphone in an ear.