System identification: theory for the user
System identification: theory for the user
SIAM Journal on Scientific and Statistical Computing
Neural Computation
Fundamentals of statistical signal processing: estimation theory
Fundamentals of statistical signal processing: estimation theory
Adaptive filter theory (3rd ed.)
Adaptive filter theory (3rd ed.)
Rank-deficient and discrete ill-posed problems: numerical aspects of linear inversion
Rank-deficient and discrete ill-posed problems: numerical aspects of linear inversion
Optimization by Vector Space Methods
Optimization by Vector Space Methods
Neural Networks for Optimization and Signal Processing
Neural Networks for Optimization and Signal Processing
Fast projection algorithms with application to voice echo cancellation
Fast projection algorithms with application to voice echo cancellation
Dynamically regularized fast RLS with application to echo cancellation
ICASSP '96 Proceedings of the Acoustics, Speech, and Signal Processing, 1996. on Conference Proceedings., 1996 IEEE International Conference - Volume 02
A robust proportionate affine projection algorithm for network echo cancellation
ICASSP '00 Proceedings of the Acoustics, Speech, and Signal Processing, 2000. on IEEE International Conference - Volume 02
Advances in Network and Acoustic Echo Cancellation
Advances in Network and Acoustic Echo Cancellation
A leaky RLS algorithm: its optimality and implementation
IEEE Transactions on Signal Processing - Part I
Leaky LMS algorithm: MSE analysis for Gaussian data
IEEE Transactions on Signal Processing
Underdetermined-order recursive least-squares adaptive filtering: the concept and algorithms
IEEE Transactions on Signal Processing
Bayesian regularization and nonnegative deconvolution for room impulse response estimation
IEEE Transactions on Signal Processing
Double-Talk-Robust Prediction Error Identification Algorithms for Acoustic Echo Cancellation
IEEE Transactions on Signal Processing
Acoustic feedback cancellation for long acoustic paths using a nonstationary source model
IEEE Transactions on Signal Processing
| Hi-index | 0.09 |
In many room acoustic signal processing applications, a room impulse response identification is needed to eliminate undesired effects such as echo, feedback, or reverberation. This is typically done using an adaptive filter driven by a speech or audio input signal. However, such signals exhibit poor excitation properties, which cause standard adaptive filtering algorithms to be very sensitive to disturbing signals, especially in the underdetermined case. A popular remedy is regularization, which is usually implemented with a scaled identity regularization matrix. This type of regularization is governed by a single regularization parameter, the value of which is often chosen in an arbitrary way. We propose to regularize the adaptive filter using a non-identity regularization matrix, in which prior knowledge on the unknown room impulse response may be incorporated. When knowledge of the disturbing signal is also used to add prefiltering and weighting in the adaptation, a new family of regularized adaptive filtering algorithms is obtained, which is shown to be optimal in a mean square error sense. Existing regularized algorithms can then be obtained as special cases, assuming limited or no prior knowledge is available. When combined with a recently proposed method of extracting prior knowledge from the acoustic setup, our algorithms exhibit superior convergence behaviour compared to existing algorithms in different simulation scenarios, while the additional computational cost is small.