Scale Space Filtering (SSF) is a kind of Multi-Scale Descriptions, and it has a problem that the time complexity proportionally increases with the number of the scale parameters. It is difficult to realize the real-time processing of SSF. The authors noticed SSF has a property that the calculation of one scale parameter does not interfere with the other scale parameter's and the calculation of each scale parameter can be executed in parallel. In this paper, the parallel algorithm called “Scale-Axis Division Method” is proposed for speeding up SSF, and it is inspected by the execution results using MIMD parallel machine. Then, the limitation of this method are estimated. Also, the performance of processor for the real-time processing of the underwater data transmission is considered in this algorithm.
In this paper, we provide a model reduction method for stable MIMO linear discrete-time systems such that the reduced-order model interpolates a finite series of 1st-order information (information associated with the transfer function) and 2nd-order information (information associated with the power spectrum function) of the original system with respect to a direction. The directional interpolation approach to the model reduction guarantees the stability of the reduced-order model and can suppress the unnecessary increase of its order involved by the increase of interpolated information. We also propose an algorithm of the output frequency-weighted model reduction as an effective application of the model reduction via directional interpolation.
This paper discusses a method for parameter estimation in the NARMAX (non-linear autoregressive moving average with exogenous inputs) model using neural computation. The primary aim of the method is to examine the structure of biological systems, utilizing the fact that the NARMAX model contains a relatively few parameters. A three-layered feedforward neural network is trained to describe a system. The actual input of the system and the computed output of the network are used as the input data set of the network for training, and values of weights and thresholds in the network are determined to minimize the prediction error between the actual and computed output. Parameters in the NARMAX model are calculated from the values of weights by expanding the sigmoid functions in neural units using Maclaurin's formula. The structure of the NARMAX model is finally determined by information criteria. The proposed method, therefore, requires no prior knowledge of the structure of the NARMAX model. Some numerical examples are presented to illustrate that the propised method can work well for noisy data.
This paper deals with a problem of computing the Hankel singular values of systems with two-point non-commensurate delays in control. It is explicitly shown that the Gramian can be described as an integral operator with semi-separable kernel function, that is, the sum of a Volterra operator and a finite dimensional one. As a result, the Hankel singular values are computed by solving a transcendental equation.
The mechanical behavior of RCC devices depends on the number of rubber-metal layers in elastomer shear pads, thickness and radius of the rubber and metal, axial and lateral stiffness of the elastomer, which have to be considered for complete analysis or design of the RCC devices. This paper presents a useful method of analyzing the behavior of RCC based on the material mechanics approach. This makes it easy to design a new RCC for a specific purpose. Simulation and experimental results are also presented.