Adaptive Acoustic Echo Cancellation (AEC) is a modern solution to overcoming non-linearities in acoustic echo cancellation. Our enhanced Kernel Based Adaptive Acoustic Echo Cancellation package provides computationally efficient non-linear echo cancellation for low cost and low power devices while greatly simplifying component selection. Contact us to discuss your adaptive AEC requirements for low power applications.
VOCAL’s Kernel Based Adaptive Acoustic Echo Cancellation software library can be ported to run on a wide variety of DSPs and conventional microprocessors from TI, ADI, Intel, ARM and other vendors. Custom solutions are also available.
Non-Linear Acoustic Echo
More than 30 years ago, Volterra filters were used to handle acoustic echo non-linearities. Despite their performance gains and ability to handle non-linear echo paths, the exponential increase in complexity they demand per tap has limited their utility in practical performance.
In general, non-linearities can be caused by any of the following reasons: vocoders, amplifiers having non-ideal gain curves, a mismatch between the loudspeaker’s time constant and the sampling rate, analog-to-digital conversion clipping the converted audio, resonance in product housing or in the acoustic echo path, or the presence of correlated interfering signals.
The traditional solution is to place what is termed a Non-Linear Processor (NLP) at the end of the acoustic echo cancellation processing chain to remove any residual echo that persists. While improved performance certainly results, the underlying assumption is still a linear echo path, so this solution is actually not appropriate for non-linear echo paths.
Kernel Based Adaptive Acoustic Echo Canceller
VOCAL’s Kernel Based Adaptive Acoustic Echo Canceller explicitly handles all non-linearities by automatically mapping a low dimensional non-linear problem into a high dimensional linear problem using powerful state of the art machine learning techniques. It is in this higher linear space where adaptive AEC can be effectively performed.
Our carefully designed kernels offer superior adaptive AEC performance in non-linear environments with little additional computational cost. Due to the bounded range of digital audio, a mapping can be efficiently represented as a simple pre-calculated look-up table such that the only extra cost incurred is the table search.
VOCAL’s optimized software is available for the following platforms. Please contact us for specific Kernel Based Adaptive Acoustic Echo Canceller supported platforms and associated performance data.