Attention-Based Voice Signal Estimation for Echo Cancellation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional voice signal estimation methods fail to accurately distinguish and remove echo signals in voice communication, leading to deteriorated echo cancellation performance due to phase mismatch in frequency domain processing.
Innovation Solution
A voice signal estimation apparatus using an attention mechanism that applies an attention mechanism to input signals from a microphone encoder and a far-end signal encoder, employing a pre-learned artificial neural network to estimate mask information and extract the user's voice signal by analyzing correlations and weighting echo signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional frequency domain processing is used for echo cancellation, then echo removal is performed, but phase mismatch causes deteriorated cancellation performance
Solution Approach 1:
The patent replaces conventional frequency domain processing methods with a time domain artificial neural network-based approach. The ANNET processes signals directly in the time domain, eliminating the phase mismatch problem inherent in frequency domain methods while achieving effective echo cancellation through learned temporal patterns and attention mechanisms.
2Measurement precision
If attention mechanism is applied to enhance voice signal estimation, then voice quality improves, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by using the attention mechanism to pre-identify and weight important features in the input signals before they are processed by the subsequent neural network layers. This allows the model to focus computational resources on the most relevant portions of the signal, improving estimation accuracy while managing complexity through selective processing.
Data Source
AI summary
A voice signal estimation apparatus includes: a microphone encoder that receives a microphone input signal including an echo signal and a user's voice signal, converts it into first input information, and outputs the information; a far-end signal encoder that receives a far-end signal, converts it into second input information, and outputs the information; and an attention unit outputting weight information by applying an attention mechanism to the first and second input information. The apparatus further includes a pre-learned first artificial neural network receiving third input information, which is the sum of the weight information and the second input information, and outputting first output information including mask information for estimating the voice signal from the second input information. A voice signal estimator outputs an estimated voice signal based on the first output information and the second input information.