Adaptive Hearing Filter Banks With Shared Subchannel Weights
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hearing devices face challenges in efficiently processing audio signals due to the high computational demands of adaptive filtering, which can lead to battery life issues and increased power consumption, especially when dividing audio signals into sub-band and sub-sub-band signals for noise suppression and amplification.
Innovation Solution
The method employs a multi-stage filter bank arrangement where the audio signal is transformed into sub-band signals and further divided into sub-channels, with adaptive filtering using a second transformation stage to determine weighting factors. These weighting factors are calculated for one sub-channel and shared among others, reducing the number of calculations required by up to 50% through cross-correlation analysis and efficient transmission patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If adaptive filtering is applied to each sub-sub-band signal individually, then noise suppression effectiveness is improved, but computational load increases significantly
Solution Approach 1:
The audio signal is divided into multiple sub-band signals through first-stage transformation, and each sub-band signal is further divided into sub-sub-band signals through second-stage transformation. This segmentation allows adaptive filtering to be applied selectively to specific sub-sub-band signals containing noise, rather than processing all signals equally, thus reducing overall computational load while maintaining noise suppression effectiveness.
Solution Approach 2:
Different processing strategies are applied to different sub-sub-band signals based on their specific characteristics. The system identifies which sub-sub-band signals contain noise and applies adaptive filtering only to those specific signals, while leaving other signals unprocessed or processed with simpler methods. This localized approach ensures noise suppression where needed while minimizing unnecessary computational operations.
2Reliability
If the audio signal is divided into multiple sub-band and sub-sub-band signals, then noise suppression capability is improved, but power consumption increases
Solution Approach 1:
Instead of applying full adaptive filtering processing to all sub-sub-band signals, the system performs partial processing only on those signals identified as containing noise. The processing depth is adjusted based on the actual noise content in each sub-sub-band, avoiding excessive computation on clean signals and thereby reducing overall power consumption while maintaining effective noise suppression capability.
3Reliability
If comprehensive adaptive filtering is applied to all sub-channels, then audio quality is improved, but processing time increases
Solution Approach 1:
The system performs preliminary analysis to identify which sub-sub-band signals contain noise before applying adaptive filtering. By pre-identifying the signals that require processing, the system avoids wasting time on comprehensive filtering of all signals. This preliminary classification step enables selective processing that maintains audio quality for signals needing improvement while significantly reducing total processing time.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An object of the present invention is to enable efficient operation of a hearing device that processes an audio signal by means of an adaptive filter. The hearing device (10) according to the invention comprises a first-stage transformation unit (14) and at least two second-stage transformation units (16), each second-stage transformation unit (16) being designed to divide the signal of one of the first-stage channels into subchannels of the channel by transformation. A filter unit (18) in each channel, which has a second-stage transformation unit (16), filters the signal of the channel depending on weighting factors determined for each of the subchannels. The filter units (18) are designed to exchange weighting factors with each other, so that only some of the weighting factors actually need to be calculated.