This paper addresses the parameter estimation problem of an interval-based hybrid dynamical system (interval system). The interval system has a two-layer architecture that comprises a finite state automaton and multiple linear dynamical systems. The automaton controls the activation timing of the dynamical systems based on a stochastic transition model between intervals. Thus, the interval system can generate and analyze complex multivariate sequences that consist of temporal regimes of dynamic primitives. Although the interval system is a powerful model to represent human behaviors such as gestures and facial expressions, the learning process has a paradoxical nature: temporal segmentation of primitives and identification of constituent dynamical systems need to be solved simultaneously. To overcome this problem, we propose a multiphase parameter estimation method that consists of a bottom-up clustering phase of linear dynamical systems and a refinement phase of all the system parameters. Experimental results show the method can organize hidden dynamical systems behind the training data and refine the system parameters successfully.

This paper presents new encoding methods for the binary genetic algorithm (BGA) and new converting methods for the real-coded genetic algorithm (RCGA). These methods are developed for the specific case in which some parameters have to be searched in wide ranges since their actual values are not known. The oversampling effect which occurs at large values in the wide range search are reduced by adjustment of resolutions in mantissa and exponent of real numbers mapped by BGA. Owing to an intrinsic similarity in chromosomal operations, the proposed encoding methods are also applied to RCGA with remapping (converting as named above) from real numbers generated in RCGA. A simple probabilistic analysis and benchmark with two ill-scaled test functions are carried out. System identification of a simple electrical circuit is also undertaken to testify effectiveness of the proposed methods to real world problems. All the optimization results show that the proposed encoding/converting methods are more suitable for problems with ill-scaled parameters or wide parameter ranges for searching.

A noise reduction technique to reduce background noise in noisy speech is proposed. We have proposed the noise reduction method which uses a noise reconstruction system. However, since a residual speech signal is included in the input signal of a noise reconstruction filter (NRF) used for reconstructing the background noise, the long time average value of error signal for estimating the background noise is needed not to estimate the speech signal. Therefore, the ability of tracking the non-stationary noise is decreased. In order to solve this problem, we propose the noise reconstruction system with adaptive line enhancer (ALE). Since ALE works to obtain the signal occupied by noise components, the input signal of the NRF includes only a few speech components. Therefore, we can give the high tracking ability to NRF.

Coefficient-based test (CBT) is introduced for detecting parametric faults in analog circuits. The method uses pseudo Monte-Carlo simulation and system identification tools to determine whether a given circuit under test (CUT) is faulty.

The paper is concerned with an identification-based predistortion scheme for compensating nonlinearity of high power amplifiers (HPA). The identification algorithms for the Wiener-Hammerstein nonlinear model are developed in the frequency domain. By approximately modeling the nonlinear distortion part in HPA by polynomial or spline functions, and introducing linear distortion parts in the input and output of the nonlinear element, the iterative identification schemes are proposed to estimate all the uncertain parameters and to construct an inverse system for the predistortion.

A new blind identification method of transfer functions between variables in feedback systems is introduced for single sweep type of MEG data. The method is based on the viewpoint of stochastic/statistical inverse problems. The required conditions of the model are stationary and linear Gaussian processes. Raw MEG data of the brain activities are heavily contaminated with several noises and artifacts. The elimination of them is a crucial problem especially for the method. Usually, these noises and artifacts are removed by notch and high-pass filters which are preset automatically. In the present paper, we will try two types of more careful preprocessing procedures for the identification method to obtain impulse functions. One is a careful notch filtering and the other is a blind source separation method based on temporal structure. As results, identifiably of transfer functions and their impulse responses are improved in both cases. Transfer functions and impulse responses identified between MEG sensors are obtained by using the method in Appendix A, when eyes are closed with rest state. Some advantages of the blind source separation method are discussed.

This paper investigates the modeling of non-linearity on the generation of the single trial evoked potential signal (s-EP) by means of using a mixed radial basis function neural network (M-RBFN). The more emphasis is put on the contribution of spontaneous EEG term to s-EP signal. The method is based on a nonlinear M-RBFN neural network model that is trained simultaneously with the different segments of EEG/EP data. Then, the output of the trained model (estimator) is a both fitted and reduced (optimized) nonlinear model and then provide a global representation of the passage dynamics between spontaneous brain activity and poststimulus periods. The performance of the proposed neural network method is evaluated using a realistic simulation and applied to a real EEG/EP measurement.

In this paper, a modified neural network approach called the Guided Neural Network is proposed for the longitudinal dynamics identification of a vehicle using the well-known gradient descent algorithm. The main contribution of this paper is to take account of the known information about the system in identification and to enhance the convergence of the identification errors. In this approach, the identification is performed in two stages. First, the Guiding Network is utilized to obtain an approximate dynamic characteristics from the known information such as nonlinear models or expert's experiences. Then the errors between the plant and Guiding Network are compensated using the Compensating Network with the gradient descent algorithm. With this approach, the convergence speed of the identification error can be enhanced and more accurate dynamic model can be obtained. The proposed approach is applied to the longitudinal dynamics identification of a vehicle and the resultant performance enhancement is given.

Methods to discover laws are reviewed from among both statistical approach and artificial intelligence approach with more emphasis placed on the latter. Dimensions discussed are variable dependency checking, passive or active data gathering, single or multiple laws discovery, static (equilibrium) or dynamic (transient) behavior, quantitative (numeric) or qualitative or structural law discovery, and use of domain-general knowledge. Some of the representative discovery systems are also briefly discussed in conjunction with the methods used in the above dimensions.

In some applications, such as the echo cancellation problem of satellite-linked communication channels, there occurs a problem of estimation of a long impulse response, which consists of a long flat delay and a short dispersive response region. In this paper, it is shown that the use of the adaptive algorithm based on the frequency domain sampling theorem enables efficient identification of the long impulse response. The use of the proposed technique can lead to the reduction of both the number of adaptive weights and the complexity of flat delay estimation.

In this paper, we propose a new structure of blind equalizer and its cost function. The proposed cost function is a quadratic form and has the unique solution. In addition, the proposed scheme can employ iterative algorithms which achieve less computational complexity and can be easily realized in real time processing. In order to verify the effectiveness of the proposed schemes, several computer simulations including a 64-QAM signal equalization have been shown.

This paper is concerned with blind identification of a nonminimum phase transfer function model. By over-sampling the output at a higher rate than the input, it is shown that its input-output relation can be described by a single input multiple output model (SIMO) with a common denominator polynomial. Based on the model expression, we present an algorithm to estimate numerator polynomials and common denominator polynomial in a blind manner. Furthermore, identifiability of the proposed scheme is clarified, and some numerical results are given for demonstrating its effectiveness.

Newton based adaptive algorithms are among the algorithms which are known to exhibit a higher convergence speed in comparison to the least mean square (LMS) algorithms. In this paper we propose a simplified Newton based adaptive algorithm for an adaptive infinite impulse response (IIR) filter implemented using cascades of second order allpass filters and a finite impulse response (FIR) filter. The proposed Newton based algorithm avoids the complexity that may arise in the direct differentiation of the mean square error. The analysis and simulation results presented for the algorithm, show that the property of convergence of the poles of the IIR ADF to those of the unknown system will be maintained for both white and colored input signal. Computer simulation results confirm an increase in convergence speed in comparison to the LMS algorithm.

An efficient active noise control algorithm based on the lattice-transversal joint (LTJ) filter structure is presented, and applied to the active control of broadband noise in a 3-dimensional enclosure. The presented algorithm implements the filtered-x LMS within the LTJ structure obtained by cascading the lattice and transversal structures. Simulation results show that the LTJ-based noise control algorithm has fast convergence speed that is comparable to the lattice-based algorithm while its computational complexity is less demanding.

This paper presents a new algorithm that can solve the problem of selecting appropriate update step size in the LMS algorithm. The proposed algorithm, called a Complementary Pair LMS (CP-LMS) algorithm, consists of two adaptive filters with different update step sizes operating in parallel, one filter re-initializing the other with the better coefficient estimates whenever possible. This new algorithm provides the faster convergence speed and the smaller steady-state error than those of a single filter with a fixed or variable step size.

In this paper, we compare the performance of evolutionary digital filters (EDFs) for IIR adaptive digital filters (ADFs) in terms of convergence behavior and stability, and discuss their advantages. The authors have already proposed the EDF which is controlled by adaptive algorithm based on the evolutionary strategies of living things. This adaptive algorithm of the EDF controls and changes the coefficients of inner digital filters using the cloning method or the mating method. Thus, the adaptive algorithm of the EDF is of a non-gradient and multi-point search type. Numerical examples are given to demonstrate the effectiveness and features of the EDF such that (1) they can work as adaptive filters as expected, (2) they can adopt various error functions such as the mean square error, the absolute sum error, and the maximum error functions, and (3) the EDF using IIR filters (IIR-EDF) has a higher convergence rate and smaller adaptation noise than the LMS adaptive digital filter (LMS-ADF) and the adaptive digital filter based on the simple genetic algorithm (SGA-ADF) on a multiple-peak surface.

An adaptive infinite impulse response (IIR) filter implemented using an allpass and a minimum phase system has an advantage of its poles converging to the poles of the unknown system when the input is a white signal. However, when the input signal is colored, convergence speed deteriorates considerably, even to the point of lack of convergence for certain colored signals. Furthermore with a colored input signal, there is no guarantee that the poles of the adaptive digital filter (ADF) will converge to the poles of the unknown system. In this paper we propose a method which uses a linear predictor filter to whiten the input signal so as to improve the convergence characteristic. Computer simulation results confirm the increase in convergence speed and the convergence of the poles of the ADF to the poles of the unknown system even when the input is a colored signal.

A new cost function based on multi-band decomposition of the estimation error and application of a different step-size for each band is used in connection with the least-mean-square criterion to improve the fidelity of estimates as compared to those obtained with conventional least-mean-square adaptive algorithms. The basic new idea is to trade off time and frequency resolutions of the adaptive algorithm along the frequency domain by using different step-sizes in the analysis of distinct frequencies in accordance with the frequency-localized statistical behavior of the imput signal. The mathematical background for a stochatic approach to the multi-band decomposition-based scheme is presented and algorithms with fixed and variable step-sizes are derived. Computer experiments compare the performance of multiband and conventional least-mean-square methods when applied to system identification.

The TLS method is an unbiased estimator for solving the overdetermined set of linear equations when errors occur in all data. However it doesn't show robustness while the errors have a heavy tailed pdf. In this letter we derive a robust method of TLS (ROTLS) based on the characteristics of TLS solution, where the performance of ROTLS is verified by applying it to the system identification problem.

This letter proposes evolutionary digital filters (EDFs) as new adaptive digital filters. The EDF is an adaptive filter which is controlled by adaptive algorithm based on the evolutionary strategies of living things. It consists of many linear/time-variant inner digital filters which correspond to individuals. The adaptive algorithm of the EDF controls and changes the coefficients of inner filters using the cloning method (the asexual reproduction method) or the mating method (the sexual reproduction method). Thus, the search algorithm of the EDF is a non-gradient and multi-point search algorithm. Numerical examples are given to show the effectiveness and features of the EDF such that they are not susceptible to local minimum in the multiple-peak performance surface.