Performance analysis of the deficient length NSAF algorithm and a variable step size method for improving its performance

Yu, Yi; Zhao, Haiquan

doi:10.1016/j.dsp.2016.11.009

Cited by 22 publications

(7 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The background noise

η false(n false)

is a white Gaussian process with a low SNR of 10 dB. Here, it is assumed that the variance of the background noise,

σ_{η}^{2}

, is known for all the set‐membership algorithms, because it can be easily estimated online in practice like in [12, 31, 33]. The coloured input signal

u false(n false)

is either an autoregression signal or a speech signal, where the autoregression input, AR(1), is generated by filtering a zero‐mean white Gaussian signal through a first‐order autoregression system with a pole at 0.9.…”

Section: Simulation Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Set‐membership improved normalised subband adaptive filter algorithms for acoustic echo cancellation

Zhao

Chen

2018

IET signal process.

Self Cite

View full text Add to dashboard Cite

In order to improve the performances of recently-presented improved normalized subband adaptive filter (INSAF) and proportionate INSAF algorithms for highly noisy system, this paper proposes their set-membership versions by exploiting the theory of set-membership filtering. Apart from obtaining smaller steady-state error, the proposed algorithms significantly reduce the overall computational complexity. In addition, to further improve the steady-state performance for the algorithms, their smooth variants are developed by using the smoothed absolute subband output errors to update the step sizes. Simulation results in the context of acoustic echo cancellation have demonstrated the superiority of the proposed algorithms.Keywords. Acoustic echo cancellation, improved normalized subband adaptive filter, low signal-noise-ratio, setmembership filtering, sparse system IntroductionIn modern communication network, hands-free telephony and teleconference, the acoustic echo is a common problem that must be eliminated to improve the call quality. In recent decades, therefore, acoustic echo cancellation (AEC) based on adaptive filter has obtained significant attention [1]- [3]. Naturally, AEC is also a system identification problem (i.e., identifying the impulse response of the echo path), but it imposes several characteristics on the adaptive filter. Namely, 1) the input is the speech signal which is nonstationary and highly colored, and 2) the impulse response of the echo path is long and sparse. Owing to its simplicity and ease of implementation, the normalized least mean square (NLMS) algorithm is often used in AEC. However, the NLMS algorithm will suffer from slow convergence when the input signal is colored.To overcome this drawback, the affine projection (AP) family uses K past input signal vectors for updating the filter' coefficients, at each iteration, where K is the projection order [4], [5]. The convergence rate of AP increases as K increases, but meanwhile the steady-state error and computational cost of that become large. To reduce the computational complexity, several AP algorithms with variable projection order [6] as well as fast AP algorithms [7] were opened.

show abstract

“…The background noise

η false(n false)

is a white Gaussian process with a low SNR of 10 dB. Here, it is assumed that the variance of the background noise,

σ_{η}^{2}

, is known for all the set‐membership algorithms, because it can be easily estimated online in practice like in [12, 31, 33]. The coloured input signal

u false(n false)

Section: Simulation Resultsmentioning

confidence: 99%

“…Assumption A1: The background noise η(n) is a white process with zero-mean and variance σ η 2 . Assumption A2: The analysis filter banks H i (z), i = 0, 1, …, N − 1, are paraunitary, which was widely used in subband algorithms [32][33][34]. With this assumption, the decimated signal d i, D (k) can be expressed equivalently as:…”

Section: Appendixmentioning

confidence: 99%

Set‐membership improved normalised subband adaptive filter algorithms for acoustic echo cancellation

Zhao

Chen

2018

IET signal process.

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, any (m, j)-th entry of the matrix    , whose magnitude is less than and equal to NM so that it is much less than 1 especially for a long SAF [36], [38]. Furthermore, at each subband, the decimated subband input () i u kN is close to the white signal for large N [32]. It follows that, in contrast with (5) can be neglected.…”

Section: Proposed Algorithmsmentioning

confidence: 99%

“…The filter banks for partitioning the input signal () un and the desired signal () dn are assumed to be identical and paraunitary [32]- [36]. This assumption is to avoid the computation of the filter banks in the performance analyses.…”

Section: Assumptionmentioning

confidence: 99%

Sparsity-aware subband adaptive algorithms with adjustable penalties

Zhao

Lamare

et al. 2019

Digital Signal Processing

Self Cite

View full text Add to dashboard Cite

“…The NSAF technique includes improved convergence when distinguished with NLMS for the colored signals given at input [20][21][22]. Also, the NSAF's complication is equivalent when compared with NLMS for a long AF function [23][24][25]. Subsequently, to achieve both low steady-state error and fast convergence rate, numerous VSS-NSAF techniques were suggested [26].…”

Section: Introductionmentioning

confidence: 99%

Robust Normalized Subband Adaptive Filter For Acoustic Noise And Echo Cancellation Systems

Samuyelu¹,

Kumar²

2019

AMA_B

View full text Add to dashboard Cite

In modern decades, a major problem like echo in recent communication network can be solved by AEC. Even though AEC can be measured as a characteristic significance of identifying system, that is recognizing the path of echo; current teleconferencing and automatic telephones systems enforce numerous restrictions on traditional AFs. The initial limitation occurs when the signals given at input are the signals of speech that are frequently highly colored more willingly when distinguished with white noise; moreover, subsequent one is the impulse response in which the path of echo is sparse and extensive. It further designates that many coefficients are either zero or equivalent to zero. This paper implements a D-MVS-SNSAF technique for identifying the echo cancellation systems by adopting NSAF technique. Here, the amount of transitions in the input/output signals was measured for deriving the polynomial order from two sets of audio signals as input. Further, the proposed method is distinguished with conventional algorithms like NSAF, SS-NSAF, VS-NSAF, VS-SNSAF and MVS-SNSAF and enhancement in the implemented method is proved.

show abstract

Performance analysis of the deficient length NSAF algorithm and a variable step size method for improving its performance

Cited by 22 publications

References 25 publications

Set‐membership improved normalised subband adaptive filter algorithms for acoustic echo cancellation

Set‐membership improved normalised subband adaptive filter algorithms for acoustic echo cancellation

Sparsity-aware subband adaptive algorithms with adjustable penalties

Robust Normalized Subband Adaptive Filter For Acoustic Noise And Echo Cancellation Systems

Contact Info

Product

Resources

About