This paper proposes a compression method for storing the electroencephalogram (EEG) waveforms into digital storages under the condition of keeping the quality sufficient for diagnosis. The method is designed on the basis of features of the EEG waveforms. The data volume is reduced by storing the coefficients of the functions approximating the EEG waveforms. The waveforms are divided into segments, in each of which the frequency is similar, and are decomposed into basal wave components and other superimposed ones for efficient compression. Effectiveness of the method has been evaluated by applying it to several EEG waveforms. Quality of the reconstructed waveforms is band width of 0.5_??_120Hz, dynamic range of 60dB, and maximum error of 2μVp-p. The compression ratio is 0.648 on the average.
The redundant binary code (RBC) is used in a high-speed arithmetic algorithm because it allows carry-propagation-free addition owing to the existence of redundancy. The RBC is used conventionally in only an arithmetic algorithm. In this paper, an analog-to-RBC (A/RBC) converter is proposed. The performance of the A/RBC converter as analog-to-digital converter (ADC) system is improved by using the redundancy of RBC. The structure of A/RBC converter proposed here is based on successive approximation ADC principle. The digital-to-analog converter in the successive approximation ADC is replaced by the RBC-to-analog converter which have been reported by one of the authors. Moreover, a codeselecter and a memory are added to select and store optimum codes. There are several ways to represent an integer in RBC. In order to reduce linearity error of the ADC, the A/RBC converter selects the most optimum code out of several codes corresponding to an analog input signal. Experimental results are also presented. They show that ADC characteristics are improved by applying the A/RBC conversion algorithm proposed here.
This paper describes the in-service optical fiber monitoring for gigabits/s optical telecommunications where signal and monitor light waves are on a single-mode fiber using wavelength division multiplexing technique. In general, this monitoring is subject to crosstalk, which causes an increase in bit error rate in the signal channel and decreases in dynamic range in loss and reflection distribution measurement using optical time domain reflectometry in monitor channel. First, permissible conditions for the crosstalk power levels of signal and monitor channels were studied theoretically and experimentally. According to these conditions, the performance of optical monitor equipment and components have been clarified. It was found that optical time domain reflectometry using heterodyne detection was efficient for its large dynamic range and maximum crosstalk power level. As a result, a practical in-service monitoring system has been demonstrated.
Self-organizing multi-layer semantic maps are proposed and simulated. The semantic maps proposed by H. Ritter and T. Kohonen have a feature that the semantic relations in the input data are reflected by their relative distances in the maps. They consist of minimal two layers. Since the proposed maps are multi-layer type, they can do higher level information processing compared with the conventional minimal two-layer semantic maps. The computer simulation results indicate the effectiveness of the proposed maps: the maps can do hierarchical self-classification both in self-organizing symbol map and in role-based semantic map.
We apply a TDOF (Two Degrees of Freedom) robust controller to chaotic dynamical systems. We show that the TDOF controller is effective not only for rejection of chaotic disturbance but also for control of a chaotic plant.
Electric Power Utilities have several hundred thousands of power distribution facility management maps, each set consisting of topographic map and line map, for management of power transmition and distribution facilities. The power distribution facility management maps are now drawn and updated by hand, so it has been taken huge time to maintain these maps. Then, we developed the automatic input and recognition system for power distribution facility management maps to store the digital mapping data in computer databasse. In this paper, we describe this system and recognition technique. In this system, power distribution facility management maps, inputted by automatic drawing reader as image data and transfered to vector data by image processor, are recognized at figure elements with logical recognition method (using conditional value calculation technique to separate the figure elements like symbols and road, house shapes). But this logical recognition method cannot exactly recognize complicated parts of maps included overlapped symbols and partially omitted figures, so we adapt additional ratiocinative recognition method using knowledge-base for complicated parts of maps. We can input and recognize power distribution facility management maps automatically and get high recognition rate by this system.
This paper proposes a simple tuning method based on partial model matching for PID controller. In process control field, most of controlled objects are approximated by a 1st order lag plus dead time system. The partial model matching method can be applied to tuning PID controller for this typed system. Time scale factor σ has to be calculated as a minimum positive root of a 3rd order equation, and parameters of PID controller are calculated as functions of σ and process parameters, namely, process gain K, time constant T and dead time L. Comparing with conventional methods, in which PID control parameters are calculated as relatively simple functions of process parameters, the partial model matching method is complex and nontransparent in being applied to PID controller. This paper proposes a simple method to calculate σ when Kitamori model is used as a reference. In the method, σ is calculated by an approximate equation σ/L=1.37. And then this paper shows that σ/L is calculated as a function of L/T when Kitamori model and other models are used as a reference. Even though L/T changes, σ/L changes only a little when Kitamori model is used as a reference. Finally, it is showed that σ/L is calculated as a function of L/T in PI, I-PD and I-P controller, too. Since σ is response time constant of control system, rising time of above control systems which are tuned by the partial model matching method, can be easily compared.
The inverse problem is one of the most important problems in the medical diagnosis, because most of the medical diagnosis can be reduced to solving the inverse problems. This paper proposes one of the promising methodologies based on the Cauchy-Schwarz relation to estimate the current distributions in biological systems. A theoretical background of our method is given by means of a space power spectrum method which is one of the incomplete space Fourier series spectrums. A simple sample problem verifies the validity of our method by comparing it with the conventional method. Finally, we apply our method to obtaining the current distributions in a human heart. As a result, it is clarified that our method provides the unique current distribution patterns corresponding to correct ones.