Chair: Michael Larimore, AST
V. John Mathews, University of Utah (USA)
This paper presents an adaptive Volterra filter that employs a recently developed orthogonalization procedure of Gaussian signals for Volterra system identification. The algorithm is capable of handling arbitrary orders of nonlinearity P as well as arbitrary lengths of memory N for the system model. The adaptive filter consists of a linear lattice predictor of order N, a set of Gram-Schmidt orthogonalizers for N vectors of size P+1 elements each, and a joint process estimator in which each coefficient is adapted individually. The complexity of implementing this adaptive filter is comparable to the complexity of the system model when N is much larger than P, a condition that is true in many practical situations. Experimental results demonstrating the capabilities of the algorithm are also presented in the paper.
Ahmad M. Sarhan, University of Dayton
Russell C. Hardie, University of Dayton
Kenneth E. Barner, University of Delaware (USA)
In this paper we have introduced and analyzed a new class of adaptive nonlinear filters referred to as partition-based linear (Pl) filters. The operation of those filters depends on partitioning the observation space in some fashion. Specifically, we have used here scaler quantization as an example to illustrate the concept of partitioning the observation space. Each partition is then assigned an output based on a linear combination of observed samples in a moving window of finite length N. The filters are shown to exhibit appealing robustness. Simulations include a novel approach to estimating response-to-response variations in evoked potentials (EP), buried in the on-going electroencephalogram (EEG). Unlike the multi-channel filters currently used in EP estimaion, the Pl filters do not require a separate electrode to provide a reference signal. In addition, no repetition of the stimulus is needed and the time of the stimulus need not be known.
Junghsi Lee, ATC/CCL/ITRI (TAIWAN)
V. John Mathews, University of Utah (USA)
This paper introduces output-error LMS bilinear filters with stability monitoring. Bilinear filters are recursive nonlinear systems that belong to the class of polynomial systems. Because of the feedback structure, such models are able to represent many nonlinear systems efficiently. However, the usefulness of adaptive bilinear filters is greatly restricted unless they are guaranteed to perform in a stable manner. A stability monitoring scheme is proposed to overcome the stability problem. The paper concludes with simulation results that demonstrate the usefulness of our technique.
F. Capman, Matra Communication (FRANCE)
J. Boudy, Matra Communication (FRANCE)
P. Lockwood, Matra Communication (FRANCE)
High quality acoustic echo cancellation is now required by hands-free systems used in mobile radio and teleconference communications. The demand for fast convergence, good tracking capabilities, and reduced complexity cannot be met by classical adaptive filtering algorithms. In this paper, a new echo canceller using multirate systems and a Fast QR-decomposition based RLS algorithm is investigated. Simulation results demonstrate the efficiency of this new combined structure for acoustic echo cancellation, and afixed-point implementation of the proposed scheme confirms the expected numerical robustness of the Fast QR-RLS algorithm.
S. Attallah, ENSERB (FRANCE)
M. Najim, ENSERB (FRANCE)
The wavelet transform least mean squares algorithm (WTLMS) has been proposed as an alternative to the simple and transform based (DCT) LMS algorithms, because of its faster convergence. In this paper, we first show the influence of the regularity of the wavelet low-pass filter on the convergence behavior of the normalized WTLMS algorithm (NWTLMS). Then, we show that the subband decomposition of the input signal along a regular subband tree, which splits the signal frequency band uniformly, gives better results, i.e., a faster convergence rate than the dyadic subband tree, which splits the frequency band dyadically. Finally, we show that perfect reconstruction quadrature mirror filters (PR-QMFs), which are less regular, can lead to as good results while the multiplier-free PR-QMFs offer, furthermore, a very reduced computational complexity, and hence can be used as an alternative to the wavelet filters for accelerating the convergence rate of the NWTLMS algorithm.
Zoran Fejzo, Aware Inc.
Hanoch Lev-Ari, Northeastern University (USA)
In this work we develop an adaptive nonlinear estimation technique, polynomial model-based, that has guaranteed stability and makes parsimonious use of coefficients. Our approach to the development of reduced-complexity adaptive nonlinear filters is based on a combination of: (a) The Wiener model of nonlinear systems (both FIR and IIR) and its application to nonlinear estimation from white Gaussian signals; (b) Wiener's notion of fixed (Laguerre) pre-orthogonalization, which we have extended to include adaptive pre-orthogonalization with respect to arbitrary (non-white and non-Gaussian) input signals; (c) Efficient implementation of memoryless nonlinear maps for uncorrelated inputs based on (Hermite) orthogonal polynomials; (d) Application of suitably modified RLS and LMS adaptation techniques to determine the coefficients of such nonlinear maps.
J.T. Stonick, Carnegie Mellon University
V.L. Stonick, Carnegie Mellon University
J.M.F. Moura, Carnegie Mellon University
R. Sam Zborowski, Information Transmission Systems Corporation (USA)
In this paper we investigate algorithms to adaptively adjust the coefficients of memoryless polynomial structures used to precompensate for the nonlinear amplitude and phase distortion of the high-power amplifier in a terrestrial digital television transmitter. The results of the investigation are twofold. First, the phase error is a non-Euclidean measure of the absolute symbol error. For small inputs, noise causing a small Euclidean change can create a large phase error. We compensate for this heuristically by not updating the predistorter coefficients for small inputs. This thresholding is shown to decrease the residual error of the phase predistorter. Second, the pre-compensation nature of the amplitude corrector requires a modification to the traditional LMS algorithm. This modification will be seen to produce a smaller residual error than traditional LMS. We demonstrate the superior performance of our algorithms via simulations based on the measured characteristics of production high-power amplifiers.
Marc de Courville, Telecom Paris (FRANCE)
Pierre Duhamel, Telecom Paris (FRANCE)
Classical transform-domain algorithms adapt the filter coefficients (in each ``frequency bin'') by minimizing a criterion depending on a full-band time-domain error. This paper proposes an algorithm which updates each portion of the frequency response of the adaptive filter according to the error in the same subband. For this purpose, a multirate adaptive filter is used where a subband decomposition of the error is performed using critically sampled lossless perfect reconstruction filter banks. This new algorithm is based on the minimization of a weighted criterion by a stochastic gradient algorithm and leads to improvements in convergence rate compared to both LMS and classical frequency domain algorithms.
Koji Ashihara, Tokyo Metropolitan University (JAPAN)
Kiyoshi Nishikawa, Tokyo Metropolitan University (JAPAN)
Hitoshi Kiya, Tokyo Metropolitan University (JAPAN)
We propose a method for achieving both a fast convergence speed and the low orders of adaptive digital filters (ADFs) for subband ADFs. The proposed method is based on the multirate repeating method which uses wasted signals by downsampling. First we show how to apply the multirate repeating method to the standard structure of subband ADFs. Next we consider a new structure of subband ADFs for extending the multirate repeating method. Finally, we show the validity of the proposed method by computer simulations.
Hiroshi Ochi, University of Ryukyus
Yoshito Higa, Texas Instruments
Shigenori Kinjo, University of Ryukyus (JAPAN)
Conventional subband ADF's (adaptive digital filters) using filter banks have shown degradation in performance because of the non-ideal nature of filters. For this problem, we propose a new type of subband ADF incorporating two types of analysis filter bank. In this paper, we show that we can design the optimum filter bank which minimizes the LMSE (least mean squared error). In other words, we can design a subband ADF with less MSE than that of conventional subband ADF's.