Neural Speech Enhancement With Closed-Loop Feedback Cancellation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hearing aid technologies struggle to simultaneously optimize speech enhancement and acoustic feedback cancellation due to independent calibration of these modules, leading to artifacts like chirping and reduced sound quality.
Innovation Solution
A deep neural network is trained to jointly optimize speech enhancement and feedback cancellation by simulating the interactions between input and output modules, using a closed-loop simulation to adapt the feedback canceller step-size and decorrelate signals, thereby improving sound quality and speech intelligibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If independent calibration of speech enhancement and feedback cancellation modules is used, then each module can be optimized separately, but artifacts like chirping occur and sound quality deteriorates
Solution Approach 1:
The patent merges the speech enhancement module and feedback cancellation module into a unified neural network system. The neural network simultaneously processes both speech enhancement and feedback cancellation tasks, eliminating the artifacts caused by independent calibration while maintaining optimization for both functions.
Solution Approach 2:
The neural network is designed to perform multiple functions simultaneously: speech enhancement, feedback cancellation, and adaptation to environmental conditions. This multi-functional approach allows the system to optimize both speech quality and feedback suppression without the trade-offs inherent in separate module calibration.
2Reliability
If a unified neural network is used to jointly optimize speech enhancement and feedback cancellation, then sound quality and speech intelligibility are improved, but device complexity increases
Solution Approach 1:
The neural network is segmented into distinct functional components: an encoder for feature extraction, a recurrent neural network for temporal processing and feedback cancellation, and a decoder for speech enhancement. This segmentation allows complex functions to be distributed across manageable modules while maintaining overall system integration.
Solution Approach 2:
The recurrent neural network acts as an intermediary that processes temporal dependencies and feedback signals between the encoder and decoder. This intermediary component enables the system to handle complex feedback cancellation while maintaining speech enhancement performance, distributing computational complexity across specialized sub-components.
Data Source
AI summary
A hearing assistance device has an input processing path that receives an audio input signal from a microphone and an output processing path that provides an audio output signal to a receiver. The hearing assistance device further includes a speech enhancement module includes a first neural network trained to enhance speech in the audio input signal and a feedback cancellation module coupled to the speech enhancement module. The feedback cancellation module includes a second neural network trained to represent an acoustic path between the receiver and the microphone. The feedback cancellation module provides a feedback cancellation output that is subtracted from the audio output signal.