The method of adding chaos noise to the Hopfield neural network (NN) in application to combinatorial optimization problems has been proposed, and many researchers have suggested that adding intermittency chaos noise near the three-periodic window of the logistic map will yield the best performance. However, the Hopfield NN with chaos noise sometimes remains in a group of several solutions. In this study, some Hopfield NNs with intermittency chaos noise are connected like hierarchical networks in order to find good solutions of quadratic assignment problems (QAPs). It is confirmed that connected Hopfield NNs with intermittency chaos noise provide numerous nearly optimal solutions.

View full abstract

A kind of templates of coupled Cellular Neural Network (CNN) are introduced, which are able to overstrike “directionally” pixels in images. The robustness analysis gives the template parameter inequalities which guarantee the corresponding CNNs to work well for performing prescribed tasks. As applications, the CNNs may be used to process unclear characters or art letters.

View full abstract

This paper describes the generation of adaptive gait patterns for a quadruped robot using central pattern generators (CPGs). A new CPG model, in which the dynamics of a joint servomotor is incorporated, is proposed and used to generate the periodical oscillation of the joint angle. The CPGs are mutually connected with each other through coupling parameters. The sets of coupling parameters are adjusted by a genetic algorithm (GA) so that the quadruped robot can realize stable and suitable gait patterns by interacting adaptively with the environment. Some experimental results obtained using a quadruped robot demonstrate that the proposed CPG network is effective in generating suitable gait patterns.

View full abstract

Independent component analysis (ICA) has been applied to magnetoencephalographic (MEG) data to determine the behavior and localization of brain sources. In this work, both nonaveraged single-trial data and averaged multiple-trial data are analyzed, in order to study the relationship between the performance of decomposition and the number of averages across data trials. To evaluate the performance of source decomposition, 1) the ratio of source decomposition, which indicates whether source components will be decomposed or not, 2) the power of decomposed components, which is calculated as the covariance of decomposed components, and 3) the accuracy of source estimation, are demonstrated. In addition, a number of existing ICA algorithms such as JADE, Fast-ICA, and the natural gradient-based algorithm with a robust pre-whitening technique are used to decompose MEG data. Our results show the relationship between the accuracy of source decomposition and the number of averages and, by applying the ICA approach, the number of averages can be reduced.

View full abstract

In this paper, we present an estimation algorithm for applying a Kalman filter to waveform data of the auditory brainstem response (ABR). The model parameters for extracting the features of the ABR waveform are obtained effectively by the proposed method. Usually, the ABR is analyzed by batch processing through signal averaging over a few minutes because the response has a slight potential. By using the Kalman filter, we expand the possibility of measuring it in real-time and minimizing the model error. First, we discuss, in particular, the type and the degree of the estimated transfer function. For this, we use the ABR waves obtained by adding less than 2000 click signals and estimate the transfer function in order to extract the features of the ABR waveform using the Kalman filter. Furthermore we analyze the characteristics of the coefficients of the transfer function in more detail to extract the ABR frequency characteristics. If we apply this method to clinical cases, we can expect that a significant reduction in the requirement for the addition of signals will contribute to reducing the burden for ABR monitoring in emergency medical care and surgical operations.

View full abstract

In medical radiography, noise has been categorized as quantum mottle, which is related to the number of incident X-rays, and artificial noise, which is caused by the grid etc. Noise due to these factors results in degradation of the diagnostic usefulness of the image. The results of a Wiener spectrum study showed that graininess increased as the dose was reduced, and noise affected all frequencies. However, in clinical images, these effects are seen only in the high-frequency domain above 0.3 cycle/mm. Moreover, the effects of a grid are restricted to a parallel component or a perpendicular component based on its structure. We attempted to reduce these types of noise simultaneously using a wavelet transfer method. In the standard wavelet transform, since the zone in which the Detail domain of level 1 occupies a large frequency range, detailed processing is difficult. Therefore, a method for reducing such types of noise using a wavelet packet was attempted. In addition, a tree structure of the wavelet packet to improve processing efficiency was devised. With regard to the noise reduction method, a new method in which the wavelet transform modules maxima (WTMM) method proposed by Mallet and et al. is applied to the wavelet packet is proposed. Based on the above considerations, evaluation was performed using clinical radiographs obtained at a standard dose and reduced dose with the noise reduction processing applied. The results showed that noise caused by quantum mottle and the grid can be reduced by this method without the need for threshold processing based on clinical experience.

View full abstract

Iodine contrast medium is used in Digital Subtraction Angiography (DSA) examination, in which the contrast medium is injected into a blood vessel, and images are collected in real time. However, when iodine contrast medium is used for a patient having iodine contraindication or an iodine allergy, critical side effects may occur. We compared the DSA image of Gadolinium (Gd) contrast medium and iodine contrast medium in order to investigate DSA examination using Gd contrast medium. As a result, the contrast provided by Gd contrast medium decreased due to the absorption of X-rays. When the gradation processing (linear gray scale transformation) built into DSA equipment is applied to the image in which the contrast has decreased, there is also the possibility that the image will become unusable for diagnosis due to increased noise. The present study devised a nonlinear gray scale transformation process using Gd contrast medium by which noise to the DSA image is not increased. As main techniques, the genetic algorithm which can perform optimization and search was applied to image processing. As a result of calculating the optimal gray scale transformation function using a genetic algorithm, and applying this function to a DSA image using Gd contrast medium, the obtained image contained information required for diagnosis.

View full abstract

In the present study we propose the combined use of wavelets and active contours to produce an alternative face segmentation method. Moreover, the implementation of this method in a hairstyle simulation system is discussed. The system was mainly designed as a web content service for mobile phones. Using our face segmentation method, an improvement over previous hairstyle simulation systems has been achieved. The improvement, measured as the subjective quality of the output, was produced by trading system speed for system quality. Nonetheless, we believe that future versions of the system will have lower computation times.

View full abstract

OFDM (orthogonal frequency division multiplexing) systems are very sensitive to carrier frequency offset. Although frequency offset is generally compensated for by an automatic frequency control (AFC), residual frequency offset remains after the AFC. Moreover, OFDM signals are also affected by phase noise. Thus, a kind of phase tracking is needed to compensate for these effects. In an OFDM system such as IEEE 802.11a, a phase tracking scheme that uses pilot subcarriers to detect the phase rotation caused by residual frequency offset and phase noise is employed. However, the spectrum of the AFC output signal is shifted by residual frequency offset. That is, the detected phase is the composite phase of the pilot subcarrier and the adjacent subcarrier because of inter-carrier interference (ICI). The compensation based on the composite phase sometimes overcorrects or undercorrects the phase rotation. In this paper, we propose a phase tracking scheme using iterative estimation of residual frequency offset in OFDM systems. The proposed phase tracking scheme estimates residual frequency offset in the frequency domain, and compensates for it in the time domain. The compensation in the time domain mitigates the spectrum shift of the AFC output. Therefore, the proposed scheme can reduce ICI, and track phases more accurately than the conventional phase tracking scheme.

View full abstract

We propose coherent ultrashort light pulse code-division multiple-access (CDMA) communication systems with multipulse pulse position modulation (MPPM). Coherent ultrashort light pulse CDMA is an asynchronous multiplexing technique for high-speed fiber-optic communications. MPPM is introduced to outperform both pulse position modulation (PPM) and on-off keying (OOK), when the optical bandwidths for these three modulation methods are the same. We theoretically derive the bit error rate (BER) of the proposed system. The performance characteristics are discussed with respect to a number of system parameters, and are compared with those of PPM and OOK. Our results show that MPPM is able to achieve a lower BER than that of both PPM and OOK when the bit rates for these methods are the same. We also find that there exists the optimum number of pulses in an MPPM signal block, which is inversely proportional to both the bit rate and the total number of multiple access interference (MAI) users.

View full abstract

The synthesis of analog electronic circuits involves the design of their input-output characteristic and the selection of suitable elements. The design of analog electronic circuits is difficult, and in general, there has been no general automated synthesis procedure. Therefore, a number of automated synthesis procedures of analog circuits have been proposed. Koza et al. proposed an automated synthesis procedure using genetic programming. The automated synthesis procedure can automatically create parameterized topologies of the desired analog circuit. However, this synthesis procedure uses an infinite number of devices, and therefore the scale of the generated circuits is very large in many cases. In a real design process, only the finite number device can be used. Also, the generated circuits are only evaluated by a numerical simulation, namely, SPICE. The generated circuit is required to have stability and reproducibility. In this paper, we propose a novel automated synthesis procedure for a two-port circuit which has a simple nonlinear characteristic, by means of a genetic algorithm. The proposed method generates a suitable circuit which uses minimum resources.

View full abstract