Echo Cancellation Using Sub-Band Signal Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing echo cancelation technologies face challenges in reducing complexity and eliminating echo leakage on sub-band edges, particularly due to high calculation requirements and insufficient echo attenuation in adaptive filtering.
Innovation Solution
The method involves dividing audio signals into high-band and low-band signals, using adaptive filtering for low-band echo cancelation and envelope prediction echo suppression for high-band signals, with specific processing steps including reference signal analysis, FFT transformation, and envelope prediction coefficient calculation to reduce calculation complexity and prevent echo leakage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If adaptive filtering is used for echo cancelation with high sampling rate and long reverberation time, then echo cancelation performance is improved, but calculation complexity increases significantly
Solution Approach 1:
The patent divides the full-band signal into multiple sub-bands (e.g., 8 sub-bands for 48kHz sampling rate). Each sub-band is processed independently with a lower-order adaptive filter, reducing the computational complexity from O(N) where N=14400 to O(N/M) where M is the number of sub-bands, while maintaining effective echo cancelation performance
Solution Approach 2:
The patent transforms the time-domain adaptive filtering problem into frequency-domain processing by applying FFT to divide the signal into sub-bands. This dimensional transformation allows parallel processing of multiple frequency components, reducing the order of individual filters while preserving overall cancelation effectiveness
2Device complexity
If sub-band adaptive filtering is used to reduce complexity, then calculation amount is reduced, but echo leakage occurs on sub-band edges
Solution Approach 1:
The patent applies different processing strategies to different regions of the frequency spectrum. For sub-band edges where echo leakage occurs, overlapping processing and coordinated optimization are used to enhance attenuation. For non-edge regions, standard sub-band adaptive filtering is applied. This localized quality adjustment ensures high echo attenuation at critical sub-band edges while maintaining overall low computational complexity
3Reliability
If high-order adaptive filter is used for long reverberation time, then echo cancelation completeness is improved, but device cost increases
Solution Approach 1:
The patent segments the high-order filtering task into multiple low-order parallel filters operating on different sub-bands. Instead of implementing one filter of order 14400, the system uses multiple filters of much lower order (e.g., order 1800 for 8 sub-bands), reducing memory requirements, computational load, and device cost while achieving the same overall echo cancelation completeness through parallel processing
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
An echo cancelation method is disclosed. The method includes: dividing a to-be-processed audio signal into a to-be-processed high-band audio signal and a to-be-processed low-band audio signal; performing, by using an AEC module, adaptive filtering echo cancelation processing on the to-be-processed low-band audio signal, and skipping performing adaptive filtering echo cancelation processing on the to-be-processed high-band audio signal, to generate a preliminary echo cancelation signal; performing, by using an RES module, envelope prediction echo suppression on a high-band signal in a preliminary echo cancelation frequency domain signal, and calculating and outputting a residual echo suppression coefficient; performing, by using the RES module, echo suppression on a low-band signal in a preliminary echo cancelation frequency domain signal, and outputting a processing result; and multiplying the output result and the residual echo suppression coefficient, and outputting a signal of which echoes are canceled. In addition, an echo cancelation device is provided. In the embodiments of the present invention, high-band processing and low-band processing are separately performed on a to-be-processed signal, thereby not only effectively reducing a calculation amount of echo cancelation, but also solving a problem of echo leakage existing on sub-band edges in the prior art.