Symmetric Adaptive Decorrelation Research Articles

Acoustic noise reduction by new sub-band forward symmetric adaptive decorrelation algorithms

The two-channel forward adaptive filtering algorithms show a good performance in terms of speech-noise separation. The symmetric adaptive decorrelation (SAD) algorithm represents an efficient solution to resolve the acoustic noise reduction problem. The drawback of the classical full-band forward SAD algorithm is its poor performance in term of convergence rate (low) and speech quality (less). In this paper, we propose two contributions. (i) Firstly, a new sub-band configuration of forward decorrelation algorithm (SFD) is proposed to improve the convergence rate. (ii) Secondly, we propose an extended version of the first proposed SFD algorithm based on normalized step-sizes (SNFD). To validate the good performances of two proposed SFD and SNFD algorithms, intensive experiments are done. The obtained results show good performances of these algorithms in comparison with the basic full-band version.

Acoustic noise reduction by new two-channel proportionate forward symmetric adaptive decorrelating algorithms in sparse systems

In several previous works, the two-channel adaptive filtering algorithms have been proposed and combined with forward-and-backward blind source separation structures for acoustic noise reduction and speech enhancement when the impulse responses are dispersive. In this paper, firstly we present all mathematical formulations of forward symmetric adaptive decorrelation algorithm based on normalized step-sizes control. Secondly, we propose three new proportionate algorithms that improve the convergence rate of cross-adaptive filters when the impulse responses of convolutive mixing system are sparse. To validate the good performance of proposed algorithms in term of noise reduction and speech enhancement properties, we do intensive experiments based on several criteria. To evaluate exactly the convergence speed property, we use the system mismatch and segmental mean square error criteria. We use also the cepstral distance and segmental signal-to-noise ratio to evaluate the quality of enhanced speech output. The obtained results have shown good performances of the proposed proportionate algorithms in comparison with the two-channel normalized decorrelation and forward NLMS algorithms with sparse systems.

A hybrid frequency–time domain symmetric adaptive decorrelator

Symmetric adaptive decorrelation (SAD) is a semi-blind method of separating convolutely mixed signals. While it has restrictions on the physical layout of the demixing equipment, it is better suited for some applications (e.g., live sound mixing) as no post-processing is required to ascertain which output corresponds with which source. Since SAD is based on the least mean squares algorithm, it can be modified to perform the bulk of the processing in the frequency domain. This makes it more efficient for larger filter sizes and/or larger number of sources but renders it unsuitable for real-time applications as there is a lag between the output and the input. In this paper, we propose a hybrid approach that does not suffer from the lag of the frequency domain approach. While the proposed algorithm is slightly less computationally efferent than the pure frequency domain algorithm, it is significantly more efficient than the time domain approach. A comparison of the frequency domain and hybrid algorithms shows that both achieve separation equivalent to the time domain algorithm in a real-world environment.

Symmetric adaptive maximum likelihood estimation for noise cancellation and signal separation

The symmetric adaptive maximum likelihood estimation (SAMLE) algorithm is proposed. It is shown to be a generalisation of the symmetric adaptive decorrelation (SAD) algorithm for noise cancellation and signal separation. Both the SAD and SAMLE algorithms are applied to the separation of a mixture of natural speech recorded in a realistic acoustic environment. A comparative simulation confirms that the SAMLE algorithm provides superior separation capabilities.

Signal separation by symmetric adaptive decorrelation: stability, convergence, and uniqueness

The performance of signal enhancement systems based on adaptive filtering is highly dependent on the quality of the noise reference. In the LMS algorithm, signal leakage into the noise reference leads to signal distortion and poor noise cancellation. The origin of the problem lies in the fact that LMS decorrelates the signal estimate with the noise reference, which, in the case of signal leakage, makes little sense. An algorithm is proposed that decorrelates the signal estimate with a signal-free noise estimate, obtained by adding a symmetric filter to the classical structure. The symmetric adaptive decorrelation (SAD) algorithm no longer makes a distinction between signal and noise and is therefore a signal separator rather than a noise canceler. Stability and convergence are of the utmost importance in adaptive algorithms and hence are carefully studied. Apart from limitations on the adaptation constants, stability around the desired solution can only be guaranteed for a subclass of signal mixtures. Furthermore, the decorrelation criterion does not yield a unique solution, and expressions for the phantom solutions are derived. Simulations with short FIR filters confirm the predicted behavior. >