Echo Path Delay Estimation for Robust Deep Echo Cancellation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing echo cancellation technologies face challenges in estimating and managing inconsistent echo path delays, which hinder the convergence of adaptive filters and degrade the performance of deep learning models, leading to suboptimal acoustic echo suppression.
Innovation Solution
A multi-task deep neural network system comprising two convolutional recurrent neural networks (CRNNs) and a fully-connected network is employed to estimate echo path delays and enhance microphone signals, using reference and microphone signals to predict and compensate for delays, thereby stabilizing echo cancellation performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing AEC or AES technologies use adaptive filters, then echo cancellation can be performed, but the convergence is slowed down due to ref-delay inconsistency
Solution Approach 1:
The system performs preliminary ref-delay estimation before echo cancellation using a dedicated ref-delay estimation module. By estimating the ref-delay in advance and compensating for it before the adaptive filter processes the signals, the system avoids the slowed convergence that would otherwise occur during the echo cancellation process itself.
2Reliability
If deep learning models are used for echo cancellation, then cancellation performance can be improved, but performance degrades due to unseen ref-delays in training data sets
Solution Approach 1:
The system segments the echo cancellation task into two independent modules: ref-delay estimation and echo cancellation. The ref-delay estimation module separately estimates the delay parameter, which is then used to compensate the reference signal before input to the echo cancellation module. This segmentation allows the echo cancellation model to focus on cancellation without being adversely affected by variations in ref-delay.
3Measurement precision
If ref-delay estimation is performed accurately, then echo cancellation performance is improved, but system complexity increases
Solution Approach 1:
The system introduces a ref-delay estimation module as an intermediary component between the reference signal and the echo cancellation module. This intermediary estimates the ref-delay parameter and applies compensation to the reference signal, thereby improving the overall system performance without requiring complete redesign of the echo cancellation architecture.
Data Source
AI summary
A method of echo path delay destination and echo cancellation is described in this disclosure. The method includes: obtaining a reference signal, a microphone signal, and a trained multi-task deep neural network, wherein the multi-task deep neural network comprises a first neural network and a second neural network; generating, using the first neural network of the multi-task deep neural network, an estimated echo path delay based on the reference signal and the microphone signal; updating the reference signal based on the estimated echo path delay; and generating, using the second neural network of the multi-task deep neural network, an enhanced microphone signal based on the microphone signal and the updated reference signal.