In general adaptive time-varying filters are used for acoustic echo cancellation (AEC) and active noise suppression. In non-linear AEC we assume there is low or no noise when cancel the echo with adaptive filters. My concern is to find algorithms or adaptive filter structures that can manage both the issues simultaneously without any performance degradation.
Kindly help me in getting some new idea in this direction.
Regards,
A.Kar
You can use two separate nonlinear models based on NARX or NARMAX for echo and noise , but the coefficients will be updated simultaneously. Because in case of AEC only echo is modeled as FIR filter with multiple taps and the filter coefficients are updated with proper algorithm.
I think it cannot be done by a single system, because the inputs to the adaptive filters for noise cancelation and echo cancelation are different. Only if you regard two s+adaptive filters as a single system when they are implemented in the same IC.
You can use Independent Component Analysis is used for for separating your echo and noise as a single system
To get some ideas you might want to look at work done on deghosting on marine seismic data. Loot at seg.org or eage.org. There are several approaches that are described. The most common is to apply an inverse filter. However, you may also find a finite difference algorithm that might be suitable. Even though the theory is simple, a practical implementation in 2 or 3-D might be rather tricky.
- Badges
- Science topic
- Similar topics
- Engineering
- Telecommunications Engineering