Multikernel adaptive filtering is an attractive nonlinear approach to online estimation/tracking tasks. Despite its potential advantages over its single-kernel counterpart, a use of inappropriately weighted kernels may result in a negligible performance gain. In this paper, we propose an efficient recursive kernel weighting technique for multikernel adaptive filtering to activate all the kernels. The proposed weights equalize the convergence rates of all the corresponding partial coefficient errors. The proposed weights are implemented via a certain metric design based on the weighting matrix. Numerical examples show, for synthetic and multiple real datasets, that the proposed technique exhibits a better performance than the manually-tuned kernel weights, and that it significantly outperforms the online multiple kernel regression algorithm.

In this paper, online sparse Volterra system identification is proposed. For that purpose, the conventional adaptive projection-based algorithm with weighted l1 balls (APWL1) is revisited for nonlinear system identification, whereby the linear-in-parameters nature of Volterra systems is utilized. Compared with sparsity-aware recursive least squares (RLS) based algorithms, requiring higher computational complexity and showing faster convergence and lower steady-state error due to their long memory in time-invariant cases, the proposed approach yields better tracking capability in time-varying cases due to short-term data dependence in updating the weight. Also, when N is the number of sparse Volterra kernels and q is the number of input vectors involved to update the weight, the proposed algorithm requires O(qN) multiplication complexity and O(Nlog 2N) sorting-operation complexity. Furthermore, sparsity-aware least mean-squares and affine projection based algorithms are also tested.

In this paper, analysis of average bit error ratio (BER) performance of a quadriphase shift keying (QPSK) direct-sequence code-division multiple-access (DS-CDMA) system with narrow-band interference (NBI) suppression complex adaptive infinite-impulse response (IIR) notch filter is presented. QPSK DS-CDMA signal is transmitted over a Rayleigh frequency-nonselective fading channel and the NBI has a randomly-varying frequency. A closed-form expression that relates BER with complex coefficient IIR notch filter parameters, received signal-to-noise ratio (SNR), number of DS-CDMA active users and processing gain is derived. The derivation is based on the Standard Gaussian Approximation (SGA) method. Accuracy of the BER expression is confirmed by computer simulation results.

A new adaptive algorithm is proposed by introducing some modifications to the recursive least squares (RLS) algorithm. Except for the noise variance, the proposed algorithm does not require any statistics or knowledge of the desired signal, thus, it is suitable for adaptive filtering for channel estimation in code division multiple access (CDMA) systems in cases where the standard RLS approach cannot be used. A theoretical analysis demonstrates the convergence of the proposed algorithm, and simulation results for CDMA channel estimation show that the proposed algorithm outperforms existing channel estimation schemes.

The proportionate normalized least mean square algorithm (PNLMS) greatly improves the convergence of the sparse impulse response. It exploits the shape of the impulse response to decide the proportionate step gain for each coefficient. This is not always suitable. Actually, the proportionate step gain should be determined according to the difference between the current estimate of the coefficient and its optimal value. Based on this idea, an approach is proposed to determine the proportionate step gain. The proposed approach can improve the convergence of proportionate adaptive algorithms after a fast initial period. It even behaves well for the non-sparse impulse response. Simulations verify the effectiveness of the proposed approach.

In this paper, we propose a novel stochastic gradient algorithm for efficient adaptive filtering. The basic idea is to sparsify the initial error vector and maximize the benefits from the sparsification under computational constraints. To this end, we formulate the task of algorithm-design as a constrained optimization problem and derive its (non-trivial) closed-form solution. The computational constraints are formed by focusing on the fact that the energy of the sparsified error vector concentrates at the first few components. The numerical examples demonstrate that the proposed algorithm achieves the convergence as fast as the computationally expensive method based on the optimization without the computational constraints.

This paper presents a closed form solution to a problem of constructing the best lower bound of a convex function under certain conditions. The function is assumed (I) bounded below by -ρ, and (II) differentiable and its derivative is Lipschitz continuous with Lipschitz constant L. To construct the lower bound, it is also assumed that we can use the values ρ and L together with the values of the function and its derivative at one specified point. By using the proposed lower bound, we derive a computationally efficient deep monotone approximation operator to the level set of the function. This operator realizes better approximation than subgradient projection which has been utilized, as a monotone approximation operator to level sets of differentiable convex functions as well as nonsmooth convex functions. Therefore, by using the proposed operator, we can improve many signal processing algorithms essentially based on the subgradient projection.

The article describes recent adaptive estimation algorithms over distributed networks. The algorithms rely on local collaborations and exploit the space-time structure of the data. Each node is allowed to communicate with its neighbors in order to exploit the spatial dimension, while it also evolves locally to account for the time dimension. Algorithms of the least-mean-squares and least-squares types are described. Both incremental and diffusion strategies are considered.

An adaptive array code acquisition for direct-sequence/code-division multiple access (DS/CDMA) systems was recently proposed to enhance the performance of the conventional correlator-based method. The scheme consists of an adaptive spatial and an adaptive temporal filter, and can simultaneously perform beamforming and code-delay estimation. Unfortunately, the scheme uses a least-mean-square (LMS) adaptive algorithm, and its convergence is slow. Although the recursive-least-squares (RLS) algorithm can be applied, the computational complexity will greatly increase. In this paper, we solve the dilemma with a low-complexity conjugate gradient (LCG) algorithm, which can be considered as a special case of a modified conjugate gradient (MCG) algorithm. Unlike the original conjugate gradient (CG) algorithm developed for adaptive applications, the proposed method, exploiting the special structure inherent in the input correlation matrix, requires a low computational-complexity. It can be shown that the computational complexity of the proposed method is on the same order of the LMS algorithm. However, the convergence rate is improved significantly. Simulation results show that the performance of adaptive array code acquisition with the proposed CG algorithm is comparable to that with the original CG algorithm.

Affine projection algorithms perform well for acoustic echo cancellation and adaptive equalization. Although these algorithms typically provide fast convergence, they are unduly complex when updating the weights of the associated adaptive filter. In this paper, we propose a new subband affine projection (SAP) algorithm and a facile method for its implementation. The SAP algorithm is derived by combining the affine projection algorithm and the subband adaptive structure with the maximal decimation. In the proposed SAP algorithm, the derived weight-updating formula for the subband adaptive filter has a simple form as compared with the normalized least mean square (NLMS) algorithm. The algorithm gives improved convergence and reduced computational complexity. The efficiency of the proposed algorithm for a colored input signal is evaluated experimentally.

This paper considers the problem of recursive filtering for linear discrete-time systems with uncertain observation. A new approximate adaptive filter with a parallel structure is herein proposed. It is based on the optimal mean square combination of arbitrary number of correlated estimates which is also derived. The equation for error covariance characterizing the mean-square accuracy of the new filter is derived. In consequence of parallel structure of the filtering equations the parallel computers can be used for their design. It is shown that this filter is very effective for multisensor systems containing different types of sensors. A practical implementation issue to consider this filter is also addressed. Example demonstrates the accuracy and efficiency of the proposed filter.

An adaptive interpolated FIR (IFIR) echo canceller was recently proposed for xDSL applications. This canceller consists of an FIR filter, an IFIR filter, and a tap-weight overlapping and nulling scheme. The FIR filter is used to cancel the short and rapidly changing head echo while the IFIR filter is used to cancel the long and slowly decaying tail echo. This echo canceller, which inherits the stable characteristics of the conventional FIR filter, requires low computational complexity. It is well known that the interpolation filter for an IFIR filter has great influence on the interpolated result. In this paper, a least-squares method is proposed to obtain optimal interpolation filters such that the performance of the IFIR echo canceller can be further improved. Simulations with a wide variety of loop topologies show that the optimal IFIR echo canceller can effectively cancel the echo up to 73.0 dB (for an SHDSL system). About 57% complexity reduction can be achieved compared to a conventional FIR filter.

In this paper, we propose two adaptive filtering schemes for Stereophonic Acoustic Echo Cancellation (SAEC), which are based on the adaptive projected subgradient method (Yamada et al., 2003). To overcome the so-called non-uniqueness problem, the schemes utilize a certain preprocessing technique which generates two different states of input signals. The first one simultaneously uses, for fast convergence, data from two states of inputs, meanwhile the other selects, for stability, data based on a simple min-max criteria. In addition to the above difference, the proposed schemes commonly enjoy (i) robustness against noise by introducing the stochastic property sets, and (ii) only linear computational complexity, since it is free from solving systems of linear equations. Numerical examples demonstrate that the proposed schemes achieve, even in noisy situations, compared with the conventional technique, (i) much faster and more stable convergence in the learning process as well as (ii) lower level mis-identification of echo paths and higher level Echo Return Loss Enhancement (ERLE) around the steady state.

A robust adaptive filtering algorithm was established recently (I. Yamada, K. Slavakis, K. Yamada 2002) based on the interactive use of statistical noise information and the ideas developed originally for efficient algorithmic solutions to the convex feasibility problems. The algorithm is computationally efficient and robust to noise because it requires only an iterative parallel projection onto a series of closed half spaces highly expected to contain the unknown system to be identified and is free from the computational load of solving a system of linear equations. In this letter, we show the potential applicability of the adaptive algorithm to the identification problem for the second order Volterra systems. The numerical examples demonstrate that a straightforward application of the algorithm to the problem soundly realizes fast and stable convergence for highly colored excited speech like input signals in possibly noisy environments.

A 3D display using super high-density multi-view images should enable reproduction of natural stereoscopic views. In the super multi-view display system, viewpoints are sampled at an interval narrower than the diameter of the pupil of a person's eye. With the parallax produced by a single eye, this system can pull out the accommodation of an eye to an object image. We are now working on a real-time view-interpolation system for the super multi-view 3D display. A multi-view camera using convergence capturing to prevent resolution degradation captures multi-view images of an object. Most of the data processing is used for view interpolation and rectification. View interpolation is done using a high-speed image-processing board with digital-signal-processor (DSP) chips or single instruction stream and multiple data streams (SIMD) parallel processor chips. Adaptive filtering of the epipolar plane images (EPIs) is used for the view-interpolation algorithm. The multi-view images are adaptively interpolated using the most suitable filters for the EPIs. Rectification, a preprocess, converts the multi-view images in convergence capturing into the ones in parallel capturing. The use of rectified multi-view images improves the processing speed by limiting the interpolation processing in EPI.

A postprocessing algorithm is presented for blocking artifact reduction in block-coded images using the adaptive filters along the pattern of neighborhood blocks. Blocking artifacts appear as irregular high-frequency components at block boundaries, thereby reducing the noncorrelation between blocks due to the independent quantization process of each block. Accordingly, block-adaptive filtering is proposed to remove such components and enable similar frequency distributions within two neighborhood blocks and a high correlation between blocks. This type of filtering consists of inter-block filtering to remove blocking artifacts at the block boundaries and intra-block filtering to remove ringing noises within a block. First, each block is classified into one of seven classes based on the characteristics of the DCT coefficient and MV (motion vector) received in the decoder. Thereafter, adaptive intra-block filters, approximated to the normalized frequency distributions of each class, are applied adaptively according to the various patterns and frequency distributions of each block as well as the filtering directions in order to reduce the blocking artifacts. Finally, intra-block filtering is performed on those blocks classified as complex to reduce any ringing noise without blurring the edges. Experimental tests confirmed the effectiveness of the proposed algorithm.

A sound localization method based on the adaptive estimation of inverse Ear Canal Transfer Functions (ECTFs) using a stereo earphone-microphone combination is proposed. This method can adaptively obtain the individual's transfer functions to fit the listener in real-time. We evaluate our sound localization method by studying the relationship between the estimation error of inverse ECTFs and the auditory sound localization score perceived by several listener. As a result, we clarified that the estimation error required of inverse ECTFs are less than -10 dB. In addition, we describe two adaptive inverse filtering methods in order to realize real-time signal processing implementation using affine projection algorithm and discusses the convergence time of an adaptive inverse filter to determine the initial value. It is clarified that method 2 based on copy weights with initial value is more effective than method 1 with filtered-x algorithm, in terms of convergence, if the initial value is the average of many listeners' impulse responses for our sound localization method.

A new blocking artifact reduction algorithm is proposed that uses block classification and feedforward neural network filters in the spatial domain. At first, the existence of blocking artifact is determined using statistical characteristics of neighborhood block, which is then used to classify the block boundaries into one of four classes. Thereafter, adaptive inter-block filtering is only performed in two classes of block boundaries that include blocking artifact. That is, in smooth regions with blocking artifact, a two-layer feedforward neural network filters trained by an error back-propagation algorithm is used, while in complex regions with blocking artifact, a linear interpolation method is used to preserve the image details. Experimental results show that the proposed algorithm produces better results than the conventional algorithms.

An Acoustic echo canceller has problems adaptating under noisy or double-talk conditions. The adaptation process requires a precise identification of the temporarily changed room impulse response. To do this, both minimizing the step size parameter of the Least Mean Square (LMS) method to be as small as possible and giving up on updating the adaptive filter coefficients have been considered. This paper describes an adaptive cross-spectral technique that is robust to adaptive filtering under noisy or double-talk conditions and for colored signals such a speech signal. The cross-spectral technique was originally developed to measure the impulse response in a linear system. Here we apply in the adaptive cross-spectral technique to solve the acoustic echo cancelling problem. This cross-spectral technique takes the ensemble average of the cross spectrum between input and error signals and the averaged cross spectrum is divided by the averaged power spectrum of the input signal to update the filter coefficients. We have confirmed that the echo signal is suppressed by about 15 dB even under double-talk conditions. We also explain that this method has a systematic error due to using a short time block for estimating the room impulse response. Then we investigate overlapping every last half block by the following first half block in order to reduce the effect of the systematic error. Finally, we compare our method with the Frequency-domain Block LMS (FBLMS) method because both methods are implemented in the frequency domain using a short time block.

We present a receiver structure with joint blind equalization, carrier recovery, and timing recovery. The blind equalizer employs a decomposition transversal filtering technique which can reduce the complexity of convolution to about a half. We analyze the performance surface of the equalizer cost function and show that the global minima correspond to perfect equalization. We also derive proper initial tap settings of the equalizer for convergence to the global minima. We describe the timing recovery and the carrier recovery methods employed. And we describe a startup sequence to bring the receiver into full operation. The adaptation algorithms for equalization, carrier recovery, and timing recovery are relatively independent, resulting in good operational stability of the overall receiver. Some simulation results for cable-modem type of transmission are presented.