This paper presents a new multi-clustering network for the purpose of intelligent data classification. In this network, the first layer is a self-organized clustering layer and the second layer is a restricted clustering layer with a neighborhood mechanism. A new clustering algorithm is developed in this system for the efficiently use of parallel processors. This parallel algorithm enables the nodes of this network to be independently processed in order to minimize data communication load among processors. Using the parallel processors, the quite low calculation cost can be realized among the conventional networks. For example, a 4-processor parallel computing system has shown its ability to reduce the time taken for data classification to 26.75% of a single processor system without declining its performance.

This paper introduces some modelling methods of time-varying stochastic process and its linear/nonlinear adaptive identification. Time-varying models are often identified by using a least square criterion. However the criterion should assume a time invariant stochastic model and infinite observed data. In order to adjust these serious different assumptions, some windowing techniques are introduced. Although the windows are usually applied to a batch processing of parameter estimates, all adaptive methods should also consider them at difference point of view. In this paper, two typical windowing techniques are explained into adaptive processing. In addition to the use of windows, time-varying stochastic ARMA models are built with these criterions and windows. By using these criterions and models, this paper explains nonlinear parameter estimation and the property of estimation convergence. On these discussions, some approaches are introduced, i.e., sophisticated stochastic modelling and multi-rate processing.

In recent years, fractal processes have played important roles in various application fields. Since a 1/f process possesses the statistical self-similarity, it is considered sa a main part of fractal signal modeling. On the other hand, noise reduction is often needed in real-world signal processing. Hence, we propose an enhancement algorithm for 1/f signal disturbed by white noise. The algorithm is based on constrained minimization in a wavelet domain: the power of 1/f signal distortion in the wavelet domain is minimized under a constraint that the power of residual noise in the wavelet domain is smaller than a threshold level. We solve this constrained minimization problem using a Lagrangian equation. We also consider a setting method of the Lagrange multiplier in the proposed algorithm. In addition, we will confirm that the proposed algorithm with this Lagrange multiplier setting method obtains better enhancement results than the conventional algorithm through computer simulations.

This paper proposes a nonlinear signal processing by using a three layered network which is trained with self-organized clustering and supervised learning. The network consists of three layers, i.e., self-organized layer, an evaluation layer and an output layer. Since the evaluation layer is designed as a simple perceptron network and the output layer is designed as a fixed weight linear node, the training complexity is the same as a conventional one consisting of self-organized clustering and a simple perceptron network. In other words, quite high speed training can be realized. Generally speaking, since the data range is arbitrary large in signal procession, the network shoulk cover this range and output a value as accurately as possible. However, it may be hard for only a node in the network to output these data. Instead of this mechanism, if this dynamic range is covered by using several nodes, the complexity of each node is reduced and the associated range is also limited. This results on the higher performance of the network than conventional RBFs. This paper introduces a new non-linear spectrum estimation which consists of LPC analysis and RBF network. It is shown that accuracy spectrum envelopes can be obtained since a new RBF network can estimate some nonlinearities in a speech production.

In many actual applications of the adaptive filtering, input signals as well as output signals often contain observation noises. Hence, it is necessary to develop an adaptive filtering algorithm to such an errors-in-variables (EIV) model. One solution for identifying the EIV model is a total least squares (TLS) algorithm based on a singular value decomposition of an off-line processing. However, it has not been considered to identify the EIV IIR system using an adaptive TLS algorithm of which stability has been guaranteed during adaptation process. Hence we propose a normalized lattice IIR adaptive filtering algorithm for the TLS parameter estimation. We also show the effectiveness of the proposed algorithm under noisy circumstances through simulations.

This paper proposes a multi-layer cellular network in which a self-organizing method is implemented. The network is developed for the purpose of data clustering and recognition. A multi-layer structure is presented to realize the sophisticated combination of several sub-spaces which are spanned by given input characteristic data. A self-organizing method is useful for evaluating the set of clusters for input data without a supervisor. Thus, using these techniques this network can provide good clustering ability as an example for image/pattern data which have complicated and structured characteristics. In addition to the development of this algorithm, this paper also presents a parallel VLSI architecture for realizing the mechanism with high efficiency. Since the locality can be kept among all processing elements on every layer, the system is easily designed without large global data communication.

Polyvinyl alcohol gel (PVA gel) has been applied in various fields as a biomedical material in Japan since the mechanical and the electrical characteristics are very similar to the human body. In this paper, the electrical characteristics of the electrode coated with PVA gel containing saline solution are described. The electrode has been applied to a stimulation electrode in an extracochlear prosthesis. It is expected that the coated electrode can stick to the round window membrane without scarring it. From the experimental results, is was found that the electrical impedance of a stimulation electrode coated with PVA gel exhibited good stability and the effective double-layer capacitance of the coated electrode was less dependent on the current than the capacitance of the Pt-Ir electrode. The electrical characteristics of the PVA electrode were proved to be very stable for long term use in-vivo. The coated electrode was ascertained to be able stimulate reliably the auditory nerves of a guinea pig.

The throughput of a parallel execution of a Digital Signal Processing (DSP) algorithm is limited by the iteration bound, which is the minimum period between the start of consecutive iterations. It is given by T=max (Ti/Di), where Ti and Di are the total time of operations and the number of delays in loop i, respectively. A schedule is said rate-optimal if its iteration period is T. The throughput of a DSP algorithm execution can be increased by reducing the Ti's, which can be done by taking as many operations as possible out of loops without changing the semantic of the calculation. This paper presents an optimization technique, called Loop Shrinking, which reduces the iteration bound this way by using commutativity, associativity and distributivity. Also, this paper presents a scheduling method, called Period-Driven Scheduling, which gives rate-optimal schedules more efficiently than existing approaches. An implementation of both is then presented for a system in development by the authors. The system shows reduction in the iteration bound near or equal to careful hand-tunning, and hardware-optimal designs in most of the cases.

An adaptive method analyzing analytic speech signals is proposed in this paper. The method decreases the errors of finite precision on calculation in a method with real coefficients. It is shown from the results of experiments that the proposed method is more useful than adaptive methods with real coefficients.