Time variation of fish school form is investigated for constructing a mathematical model of schooling behavior of fish. In our earlier papers, the form of fish school was approximated by an axis, and an autoregressive model was proposed to estimate the time variation of the axis length. However, the model of school axis was not sufficient for representing school form. In this paper, the form of fish scool is expressed by an ellipse. The time series data of the position of each individual swimming in a water tank is obtained from the video data by means of image processing. As an example, we observe a fish school consisting of ten individuals. Three approximation methods, i. e. probability ellipse method, four individual method and equivalent periphery ellipse method, are proposed for determining the parameters of the ellipse. Of these methods the validity of probability ellipse method is shown from the viewpoint of accuracy.
In this paper, several methods to achieve fast on-line algorithms for blind equalization are developed. Firstly, the dead zone function on equalizer output is used to accelerate the convergence speed of the conventional distribution matching algorithm. Secondly, Godard's method is applied over the dead zone schemes. Furthermore, based on the asymptotic approach for the design of cost funtion, higher speed adaptive algorithm is derived and it is shown that the dead zone is reasonable for the case that the transmitted signal depends on the uniform distribution. The present cost function is designed to give the most precise estimate ωL obtained from L samples of received signal in the sense that asymptotic covariance limL→∞E [L (ωL-ω*) (ωL-ω*) T] is minimized, where ω* is the ideal equalizer. Several computer simulations show that this cost function is generally effective to accelerate the convergence speed of on-line algorithms.
I propose a self-organizing model of hierarchical information. The hierarchical representation and learning algorithm are designed for the high-level information processing using neural networks. In this paper, I deal with both learning acqusition and representation issues in a single framework. The whole complex knowledge structure can be acquired through the piecewise relational information among hierarchical information. The self-organizing hierarchical information is implemented from an object-oriented programming paradigm, Inferences are proceeded by sending and receiving messages among information-objects. Since they are proceeded inside information objects, the inference processes are autonomous and parallel.
A practical method of power spectral estimation is proposed, that can automatically provide the precise estimates of stationary random signals. First, by reviewing the conventional automatic estimation by AR model fitting, the necessity of some means for the order determination under the assured whiteness of residue error series is stressed, even when the whiteness, assumed by most of the criterions such as AIC, is violated by the conventional. Secondly, based on the fact that the residue series include much information about the original series in the above cases, yet, the automatic precise estimation by multistep fittings of AR model is presented, that iterates the model fitting until the whiteness of residue series is ascertained. Finally, the fundamental characteristic and effectiveness of the proposed method are made clear through automatic power spectral estimation of time series, generated numerically for both cases with and without fitting to the AR model.
As a path control method of mechanical system, we proposed a new method using preview control with Superficial Error Evaluation Term (SEET) based on a different viewpoint of the usual position error in order to evaluate strictly the path control error. However, there were some problems in previous report, such as not to guarantee robustness of control system, and also the transformation of SEET to discrete-time expression were not specified. In this report, we improve this method about the following two points, aiming to construct a more practical control method. (1) We design both SEET and a control system from a consistent viewpoint of discrete-time system. With this, the transformation of SEET to discrete-time expression becomes unnecessary, and we make the meaning of SEET clear. (2) A method to guarantee steady-state robustness of the control system is given, using a theory of type-I servo control system, in order to cope with parameter uncertainties and continued disturbance.