Adaptive Audio Filtering for Echo Cancellation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio processing technologies face challenges in effectively canceling echo in teleconferencing applications due to aliasing and spectral leakage between frequency subbands, which persist even after acoustic echo cancellation, leading to residual echo issues.
Innovation Solution
The proposed method involves forming composite filtered references by multiplying filtered inband references with scalar factors from other frequency subbands and adjusting these factors based on residual signals to compensate for aliasing and spectral leakage, reducing computational complexity and improving adaptation rate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If downsampling is employed in subband filtering to reduce computational complexity, then computational complexity is reduced, but aliasing occurs causing audible echo to persist in the residual signal
Solution Approach 1:
The patent introduces crossband reference signals as intermediary elements that mediate between the downsampled subband signals and the final echo cancellation output. These crossband references capture the aliasing components introduced by downsampling and provide them as additional inputs to the adaptive filters, allowing the system to compensate for the harmful effects of downsampling without increasing computational complexity proportionally to full-band processing.
Solution Approach 2:
The patent changes the parameters of the filtering system by introducing scalar factors that scale the crossband reference signals. These scalar factors are adapted to optimize the contribution of each crossband reference, allowing dynamic adjustment of the aliasing compensation based on the actual signal conditions. This parameter adaptation enables effective echo cancellation while maintaining the computational efficiency of downsampling.
2Productivity
If separate filters are employed in respective frequency subbands with different update step sizes, then adaption rate is improved, but computational complexity increases
Solution Approach 1:
The patent segments the audio signal into multiple frequency subbands and applies separate adaptive filters to each subband. This segmentation allows different update step sizes to be used in different subbands, optimizing the adaption rate for each frequency range. The computational complexity is managed by processing each subband independently at a lower sample rate due to downsampling, rather than using a single complex full-band filter.
Solution Approach 2:
The patent adds a temporal dimension to the frequency-domain segmentation by introducing time-varying scalar factors that adapt the contribution of crossband references over time. This allows the system to dynamically adjust to changing acoustic environments and speech characteristics, improving adaption rate without requiring a proportional increase in the number of filters or computational resources.
3Measurement precision
If crossband references are formed by multiplying filtered inband references with scalar factors from other frequency subbands, then aliasing compensation is achieved, but computational complexity increases
Solution Approach 1:
The patent creates copies of the filtered inband references and scales them by scalar factors to form crossband references. Instead of implementing complex full-band filtering to capture aliasing components, the system copies the existing subband-filtered signals and combines them with appropriate scaling. This approach achieves aliasing compensation with minimal additional computational overhead, as it reuses already-computed subband signals rather than requiring separate full-band processing paths.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
In an audio processing system (300), a filtering section (350, 400): receives subband signals (410, 420, 430) corresponding to audio content of a reference signal (301) in respective frequency subbands; receives subband signals (411, 421, 431) corresponding to audio content of a response signal (304) in the respective subbands; and forms filtered inband references (412, 422, 432) by applying respective filters (413, 423, 433) to the subband signals of the reference signal. For a frequency subband: filtered crossband references (424, 425) are formed by multiplying, by scalar factors (426, 427), filtered inband references of other subbands; a composite filtered reference (428) is formed by summing the filtered inband reference of the subband (422) and the filtered crossband references; a residual signal (429) is computed as a difference between the composite filtered reference and the subband signal of the response signal corresponding to the subband; and the scalar factors and the filter applied to the subband signal of the reference signal corresponding to the subband are adjusted based on the residual signal.