We demonstrate noise-induced phase synchronization among simple digital counters, aiming at its practical application to phase synchronization among isolated digital systems (driven by precise clock sources) accepting common noisy pulse sequences. We employ a simple digital up-counter where i) relaxation oscillations in nonlinear analog oscillators are emulated by linear counting and carry overflow in a low-bit digital counter and ii) the counter value, which is equivalent to the phase of the digitally emulated oscillator, is increased by the precise base clocks, as well as by external noisy pulses, and the autonomous increase of the value (by the base clock) is inhibited by the noisy pulses, depending on the phase of the counter at which the pulse is given. Through extensive Verilog simulations, we demonstrate that the proposed digital counters can be synchronized by applying a common noise sequence, even if initial values of both the counters and phases of the counter's base clocks have different values.

抄録全体を表示

We numerically investigated the dynamics of new reaction-diffusion-type excitable media where the diffusion coefficient is represented by memristive dynamics. The medium consists of an array of excitable Oregonators and each Oregonator is locally coupled with each other via memristors, which were claimed to be the fourth circuit element exhibiting a relationship between flux φ and charge q. Through extensive numerical simulations, we found that i) memristor conductances were modulated by excitable waves passing on the memristor depending on the memristor's polarity, ii) the velocity of the excitable wave propagation is thus modulated by the change in the memristor conductance, and the degree of modulation is inversely proportional to the time constant of the memristor model, and iii) nonuniform spatial patterns are generated under the periodic boundary condition (ring structure of excitable media).

抄録全体を表示

We present steady-state analysis for the class-E power amplifier outside the class-E zero-voltage switching and zero-derivative switching (ZVS/ZDS) conditions. The analytical expressions in this paper include the MOSFET-body-diode effect. By carrying out circuit experiments, it is shown that the analytical predictions agreed with the experimental results quantitatively, which indicates the validity of the analytical expressions.

抄録全体を表示

A chaotic neuron and a chaotic neural network have been applied in a variety of research, e.g., combinatorial optimization problems, dynamical associative memory, and chaos information processing. In addition to such research, a dynamical logic element based on the chaotic neuron model was proposed for novel information processing. This dynamical logic element can switch its logic function by changing the threshold and coupling coefficient to the inputs. Moreover, it has chaotic transition regions among digital (logic) states. The dynamical logic element has been implemented by a switched-capacitor (SC) chaotic neuron circuit technique, and then the basic operation of the circuit was verified through experiments. However, the dynamical logic element has several disadvantages. Therefore, we recently modified the element, and showed its improvement experimentally. In the present paper, we further improve the logic element, and propose a SC circuit implementation of the improved dynamical logic element. In addition, we verify the digital and analog characteristics of the proposed circuit through circuit experiments.The dynamical logic element has been implemented by a switched-capacitor(SC) chaoticneuroncircuittechnique, and then the basic operation of the circuit was verified through experiments. However, the dynamical logic element has several disadvantages. Therefore, we recently modified the element,and showed its improvement experimentally. In the present paper, we further improve the logic element, and propose a SC circuit implementation of the improved dynamical logic element. In addition, we verify the digital and analog characteristics of the proposed circuit through circuit experiments.

抄録全体を表示

We describe a prediction system for solar radiation using simplified fuzzy reasoning based on weather forecasts and the mesoscale spectral model grid point value (MSM-GPV). The proposed prediction system has a multistage-type structure that consists of two parts: a function approximation part using a normal distribution function and a correction part using the simplified fuzzy reasoning. The output of the prediction system is hourly averages of predicted solar radiation from 6:00 to 18:00. Using solar radiation data acquired in the past, a self-tuning algorithm tunes fuzzy correction rules so that the prediction error is decreased. Several simulation results confirm the usefulness of the proposed prediction system.

抄録全体を表示

A particle swarm optimization (PSO) is one of the powerful systems for solving global optimization problems. The searching ability of such PSO depends on the inertia weight coefficient, and the acceleration coefficients. Since the acceleration coefficients are multiplied by a random vector, the system can be regarded as a stochastic system. In order to analyze the dynamics rigorously, we pay attention to a deterministic PSO, which does not contain any stochastic factors. On the other hand, the standard PSO may diverge depending on the random parameter. Because of this divergence property, the standard PSO has high performance compared with the deterministic PSO. Since the deterministic PSO does not have stochastic factors, the diversity of the particles of deterministic PSO is lost. Therefore its searching ability is worse. In order to give diversity to the deterministic PSO, the golden angle is applied to the rotation angle parameter of the deterministic PSO. We confirm the performance of the searching ability of the proposed PSO.

抄録全体を表示

We propose an ant colony optimization (ACO) including dull ants caused by genetic algorithm (GDACO). The GDACO algorithm combines ACO including dull ants with the genetic algorithm (GA). GDACO searches solutions using the pheromone of ACO and the genetic information of GA. In addition, GDACO consists of two kinds of ants: intelligent and dull. The dull ant is caused by mutation and cannot trail the pheromone. We apply GDACO to quadratic assignment problems (QAPs) and confirm that GDACO obtains more effective results than ACO and GA.

抄録全体を表示

We present a radial-basis adaptive resonance theory map and its application to classification problems. In order to classify input data, the system generates radial-basis categories. Two key parameters can control the number and size of the categories. Performing numerical experiments for two types of traditional expectant mothers, we have investigated the effects of key parameters on the classification functions.

抄録全体を表示

The brain-computer interface (BCI) links a human brain and a computer directly. Using the BCI, people with motor disorder, such as amyotrophic lateral sclerosis (ALS), or spinal injury can control their environment or communicate with people around them. A single-trial BCI is needed to realize fast communication via a BCI-based speller, which types words into the computer, driven by brain signals. A single-trial BCI based on independent component analysis (ICA) is proposed. In the conventional method, the accuracy of 76.7% is achieved by using correlation analysis to obtain the existence of P300, which is an evoked potential in electroencephalogram (EEG). However, there is still room for improvement of the accuracy by selecting optimal sensors. By considering this possibility, better results can be realized. Therefore, in this study, we use the backward elimination method for sensor selection (to reduce the number of electrodes) and 81.0% accuracy is achieved. The aim of this study is the identification of the important electrode points for the achievement of the single-trial BCI. The proposed method exhibits a high recognition accuracy.

抄録全体を表示

Voice activity detection (VAD) is used to detect speech/nonspeech periods in observed signals and it is a very important technique for various speech signal processes. However, there is a serious problem in that the accuracy of detection of speech periods drastically reduces if the current VAD technique is used for noisy speech and/or for mixtures of speech/non-speech such as those in music and animal sounds. Thus, VAD needs to be robust to enable speech periods to be accurately detected in these situations. This paper proposes a robust method of VAD using empirical mode decomposition (EMD) and modulation spectrum analysis (MSA) to resolve these problems. The proposed method reduces noise by using EMD, and then determines speech/non-speech periods by using MSA. Five experiments on VAD in real environments were conducted to evaluate the proposed method by comparing it with traditional methods (OTSU's, the G.729, and power envelope thresholding methods). The results demonstrated that the proposed method could accurately detect speech periods more accurately than the traditional methods.

抄録全体を表示

One of the greatest difficulties in automatic speech recognition (ASR) is how to deal with variations in speech signals caused by nonlinguistic information, such as age and gender. Various methods have been proposed to compensate for the variations and one of them is speech structure. Speech structure, which extracts only contrastive features and discards absolute features, is proven to be transform-invariant mathematically and to be very robust with the nonlinguistic variations experimentally. Although the conventional speech structure extracts local and distant contrastive features, it doesnot extract dynamic features explicitly, which are supposed to exist in the contrastive features. In this paper, we reformulate speech structure based on trajectory Hidden Markov Model (HMM) and derive trajectory structure (TSR), in which dynamic and contrastive features can be defined and used in ASR. We carry out an experiment of n-best rescoring of isolated word recognition using trajectory structure and obtain a 28.5% relative decrease in the word error rate.

抄録全体を表示

It is well known that the performance of automatic speech recognition (ASR) systems is easily affected by acoustic mismatch between training and testing conditions. This mismatch is often caused by various kinds of environmental noise or distortion. To reduce the effect of mismatch, feature normalization, feature enhancement and model adaptation have been studied intensively. Cepstral mean normalization (CMN), mean and variance normalization (MVN), and histogram equalization (HEQ) are well-known methods of feature normalization. Stereo-based piecewise linear compensation for environments (SPLICE) is one of the feature enhancement methods. In this paper, we describe how to combine these methods to effectively improve the robustness of ASR systems. In the experiments performed on the Aurora-2 database, a good combination showed a 41% improvement in word error rate over SPLICE only, and a 25% improvement over the conventional combination of SPLICE and CMN.

抄録全体を表示

We propose a novel speech recognition method using multiple recognizers of various recognition units. Our method can improve the word recognition accuracy and reject incorrect recognition results when the correct recognition results are obtained. We use a Support Vector Machine (SVM) to select a single recognition result or reject recognition results from multiple recognition results. Experimental results show that the proposed method achieves higher word recognition accuracy and classification accuracy in comparison with a single recognizer and ROVER.

抄録全体を表示

In asynchronous direct sequence-code division multiple access (DS-CDMA), the chaotic spreading code is theoretically effective in reducing the bit error rate (BER), and its performance has been shown through computer simulations. In this study, we implement and evaluate the chaotic CDMA system using the software-defined radio technology. By the results of our experiments, clarified that the chaotic spreading code leads to a lower BER than that of the conventional spreading code, as well as in our implemented wireless systems.

抄録全体を表示

In this paper, a multipoint relay with energy awareness (EAMPR) is proposed for a system data-sharing protocol, abbreviated as EAMPR-SDS. It adopts a system datasharing concept which combines broadcasting routing with data fusion so that data exchange between wireless sensor nodes can be easily realized in the real world. Owing to the energy constraint of WSN, the proposed protocol employs EA-MPR to improve the transmission efficiency and extend network lifetime. Unlike some broadcasting optimizing mechanism such as self-pruning, it assigns a sending node to decide which of its neighbors should forward the shared data on the basis of the neighbors' rest energy as well as the neighbor information. Therefore, instead of all the nodes in the network, only a subset of nodes needs to forward the shared data while still guaranteeing complete network connectivity. The simulation results prove that in comparison with the existing system data-sharing protocol and self-pruning, EAMPR-SDS can significantly reduce the redundant broadcasting and its corresponding power consumption.

抄録全体を表示

In the real-time segmentation of the moving objects in an image sequence, methods based on background subtraction are used widely. Bipolar radial reach correlation (BP-RRC) has realized robust background subtraction through the evaluation of a local texture per pixel to suppress the influence of brightness variations. However, this method has the weak point that it is affected by nonstatic backgrounds. We propose a robust and stabilized background subtraction algorithm that can cope with texture changes caused by various illumination changes and nonstatic backgrounds by improving this texture background model. We verified the proposed method using image sequences that include a loose lighting variation, a rapid lighting variation, and a moving object in the background. We proved that the adaptability of the proposed method is higher than that of the conventional method on a nonstatic background as a result of verification.

抄録全体を表示

We clarified the effect of ensemble learning on the performance of systems with neural networks, using onedimensional numeric sequences as input patterns for the detection of abnormal shadows in X-ray images of lungs. In order to implement the ensemble learning, the input patterns, which were one-dimensional numeric sequences obtained from two-dimensional images, were preprocessed using several averaging and differential filters. Then, we combined several systems with neural networks constructed using the input patterns with different preprocessing conditions. From the results, we found that the ensemble learning improved the performance of the systems with neural networks using one-dimensional numeric sequences. The best value of areas under ROC curves in the systems with ensemble learning was superior to those in previous systems with twodimensional information as input patterns. Thus, the systems proposed in this study are thought to be useful for medical diagnosis.

抄録全体を表示