Simplified Newton-type adaptive estimation algorithms

Mavridis, Pavlos; Moustakides, George V.

doi:10.1109/78.533714

Cited by 12 publications

(11 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We observe that the two vectors A LC1;t and B LC1;t affect the first and the last parts of the identifier impulse response, respectively. From this observation, several papers on the fast LS algorithm have highlighted and proven two important assumptions: (i) in the first assumption, the authors of [36] prove that the last real identifier impulse responses parts by LS algorithms are redundant (small coefficients) and the variance of these impulses responses is concentrated in their first part; (ii) in the second assumption [36][37][38][39][40][41][42], the authors proved that some inputs can be modeled by an autoregressive model. This assumption leads to further simplification possibilities of these LS algorithms (this assumption will be used in the second proposed Algorithm 2).…”

Section: The Proposed Algorithmmentioning

confidence: 99%

“…Algorithm based on only forward predictor. In [1], the authors used the assumption that the physical transversal impulse responses are decreasing with the filter order [36][37][38] to propose this algorithm. In AEC applications, the forward and backward predictions are carried out with a small number of coefficients [39][40][41][42].…”

Section: The Mono-channel Ftf Adaptive Algorithmmentioning

confidence: 99%

“…In [1], the authors used the specific approach of the FNTF algorithm [37][38][39][40] to propose a second algorithm. The dual Kalman is given by…”

Section: Algorithm Based On Ar Modelization Of the Inputmentioning

confidence: 99%

See 2 more Smart Citations

New efficient adaptive fast transversal filtering (FTF)‐type algorithms for mono and stereophonic acoustic echo cancelation

Djendi

2014

Adaptive Control & Signal

View full text Add to dashboard Cite

This paper addresses the field of stereophonic acoustic echo cancelation (SAEC) by adaptive filtering algorithms. Recently, simplified versions of the fast transversal filter (SFTF)-type algorithm has been proposed. In this paper, we propose two major contributions. In the first contribution, we propose two new FTF-type algorithms with low complexity and good convergence speed characteristics. These two proposed algorithms are mainly on the basis of a forward prediction scheme to estimate the so called dual Kalman gain, which is inherent in the filtering part update. This computation complexity is achieved by the introduction of new relations for the computation of the likelihood variables that are simple and lead to further simplifications on the prediction part of the two proposed algorithms. In the second contribution, we propose to adapt then apply these four new SFTF-type algorithms, (the two proposed algorithms in this paper and their original versions) in the SAEC applications. A fair comparison of the proposed algorithms with the original SFTF and the normalized least mean square algorithms, in mono and SAEC applications, is presented. Copyright 274 M. DJENDI Recently, a different approach for the SAEC problems (inter-channel coherence, adaptive filter lengths, etc.) was proposed in [18,19] by using the widely linear model, which is assigned to the processing of noncircular complex valued signals. A new hybrid solution to the same problem that uses frequency shifts to improve the performance of the state-of-art solution based on addition of half-wave rectified signals is presented in [20]. The authors in [21] proposed a windowing frequency domain adaptive filter, and an up-sampling block transforms preprocessing, to solve these constraints. The proposed adaptive filter uses windowing functions with smooth cutoff property to reduce the spectral leakage during filter updating, so the utilization of the independent noise introduced by preprocessing in SAEC can be increased. The proposed preprocessing is operated in short blocks with low processing delay, and it uses frequency-domain up-sampling to meet the minimal block length requirement given by the band limit of simultaneous masking. Another very recent and novel approach to SAEC is proposed in [22] and based on a combination of psychoacoustic effects. The principle of the proposed structure is to use an approach based on the 'missing-fundamental' phenomenon at low frequencies (below 500 Hz) and a modified phase modulation approach at higher frequencies.A basic scheme for SAEC is given in Figure 1, where we illustrate the concept with a transmission room on the left and a receiving room on the right [16,17,23,24]. In Figure 1, when we generate a signal in the transmission room, the two microphones mic 1 and mic 2 receive two sorts of signals: (i) the direct source signal modified by the path g 1 (then received by mic 1 / and g 2 (then received by mic 2 /, respectively, and (ii) the existing diffuse noises n 1 and n 2 in the transmission room and received ...

show abstract

Section: The Proposed Algorithmmentioning

confidence: 99%

Section: The Mono-channel Ftf Adaptive Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

New efficient adaptive fast transversal filtering (FTF)‐type algorithms for mono and stereophonic acoustic echo cancelation

Djendi

2014

Adaptive Control & Signal

View full text Add to dashboard Cite

show abstract

“…Reduced size predictors in the FTF algorithms have been successfully used in the FNTF algothms [7,8]. The SMFTF algorithm can easily used with reduced size prediction part (table 2).…”

Section: Simplified Ftf-type Algorithmmentioning

confidence: 99%

“…In this application, predictor sizes are much smaller than the size of the transversal filter for speech signal. This propriety was used to develop a class of algorithms called Fast Newton transversal filter algorithm [7,8] where the input signal is modelised by an AR model with 10 to 20 coefficients. From relation (6), we can see that the most significant components, the last ones, of the backward predictor affect the last terms of the Kalman Gain and this contribution is not In the proposed algorithm, we discard all backward prediction variables from (6) and use only the forward variables to compute the dual Kalman gain :…”

Section: Simplified Ftf-type Algorithmmentioning

confidence: 99%