Multirate discrete Fourier transform (MR-DFT) has been proposed as one of the DFT algorithms by us. The feature of the MR-DFT is to use only one multiplication for each Fourier coefficient. Therefore the number of multiplications can be decreased comparing with the conventional DFT. In this paper, for applications of the MR-DFT, we consider how the noise affects the estimation performance of the MR-DFT. Then we propose two improved algorithms of the estimation performance, that is, a shifted multirate discrete Fourier transform (SMR-DFT) and a parallel shifted multirate discrete Fourier transform (PSMR-DFT). The SMR-DFT prevents an aliasing of the additive noise by a frequency modulation. The PSMR-DFT performs the parallel processing of the SMR-DFT with different sample points using the periodicity of an input signal. The estimation performances of the proposed algorithms are compared with MR-DFT by the computer simulations. The SMR-DFT could prevent the degradation of the estimation performance of the MR-DFT in the lower components caused by the wide bandwidth noise. Hence its estimation variance could become 0.61 times of the MR-DFT. Furthermore, the PSMR-DFT could improve the estimation variance of the SMR-DFT to 0.46 times. Comparing the number of operations of the proposed algorithms with the MR-DFT, the number of additions has increased, however, the number of multiplications has the same complexity.

抄録全体を表示

In recent years, it is said that communication speed of next-generation cellular phones will reach 100Mbps. Accordingly, demand for robust error correction codes with high-speed processing is increasing. High-dimensional discrete torus knot code (torus code) performs well in fields with many errors. We propose a new and unique synchronization system that uses the torus code. It performs error correction and synchronization processing effectively and simultaneously. According to results of comparison with a conventional system (HDLC:High Level Data Link Control), performance (Loss of Lock Probability) of the proposed synchronization system is as much as 5 figures better than the conventional HDLC in the 10^-4 field (Input BER). By applying a LSI technology, we have developed a FPGA(3Dm5, 125bit, r=0.512) with the proposed synchronization system and have built a high-speed MPEG communication device which uses the newly developed FPGA. The high-speed MPEG communication device can transmit a video signal of 20Mbps.

抄録全体を表示

Many laser radars have been studied as in-car radar. Generally, a distance is measured from delay time using either a periodic signal or a single pulse. But the received signal becomes to be buried in noise with increasing distance. Recently a chaos has been paid an attention and a chaos laser radar has been suggested. When a received signal is correlated with a transmitted one, it takes long processing time to perform repeatedly multiplications and integrations. A new type of chaos laser radar which processes by only an addition is proposed. It can quickly process and is comparatively strong for noise. A proto type of chaos laser radar is produced, in which a semiconductor laser with low output power below 1mW is used. This laser radar is able to measure the distance up to 80m. It is confirmed that the chaos laser radar is effective.

抄録全体を表示

We have proposed the Horizon View Camera (HVC), which is a unique system of the object detection by a single camera and has some unique features. The HVC is to put a single camera on the ground, and the optical axis of the camera is directed toward the horizon by using a mirror. We have shown its effectiveness by measuring distances to objects using the obtained images from the HVC with straight motions. In this paper, we proposed the HVC-90 that is more effective system than the previous HVC for object detection. The HVC-90 is constructed using the orthogonal optical axes of the camera and the mirror, and it can obtain two kinds of images. One is the image including only the object, and another is the image including only the ground. This system gives new features in addition to original HVC. By using them, we made it possible to the distance measurement in the case of any movements, and we show its effectiveness by some experiments.

抄録全体を表示

Relating audio-visual events is important for constructing an artificial intelligent system, which can acquire the audio-visual knowledge of moving objects through active observation without a supervisor. This paper proposes a method for relating multiple audio-visual events observed by a camera and a microphone according to general laws without object-specific knowledge, which copes with including entire object movement and sound location change. As corresponding cues, we use Gestalt’s grouping law; simultaneity of sound onsets and changes in movement, similarity of repetition between sound and movement. Based on the correlation coefficient between auditory and visual sequences, the component of frequency at sound onset is related to the spatiotemporal invariant sequence (STI sequence) of movement. We experimented in the real environment and obtained satisfactory results showing the effectiveness of the proposed method.

抄録全体を表示

To elucidate perceptual filling-in mechanisms in peripheral vision, we investigated dependency of filling-in occurrence on spatio-temporal frequency of dynamic textures surrounding the filling-in target. We first measured spatial frequency sensitivity of the filling-in target in static texture. Then, the time to filling-in, when dynamic textures which have variously limited spatio-temporal frequency are surrounding the filling-in target, were measured. According to the hypothesis of filling-in process which has already proposed by the authors, the tendency of inducing filling-in, i.e., the attenuation factor of perceptual power for filling-in target in dynamic textures, is estimated as a function of spatio-temporal frequency. It was suggested that surrounding texture with stronger perception promotes filling-in more intensively.

抄録全体を表示

When we come across flowers in hills and fields, we might want to know their names. We can usually obtain information on flowers from flower picture books, but it is not available without flower names. It is difficult to search a flower from the appearance, in this study, we propose a flower information retrieval system using picture. We extract a flower area from a picture with a natural background using domain knowledge. Then we compute some features from the obtained flower area; flower color, size, the number of petals, circularity, and form similarity. We use these features as reference keys. A form similarity, which enables to retrieve visually similar flowers, is computed by matching the outline of two figures using dynamic programming. In the reference experiment, a reference ranking average is fourth and this shows our system validity.

抄録全体を表示

In this paper we propose a least mean p-th adaptive notch filter that has a cost function of E[e^p(n)], where e(n) is the estimation error. The structure of the adaptive filter is a tandem connection of the second-order adaptive notch filter with an allpass filter. In general, the bandwidth of the notch filter is preferable to be extremely small for the theoretical and practical viewpoint. However, the convergence speed of the weight becomes slow if the bandwidth is reduced. The transfer function of the notch filter has a following special characteristic, that is zero in the center frequency and unity in other frequency. The equivalent broad bandwidth can be obtained when the cost function is chosen as E[e^p(n)]. The higher convergence speed and the excellent stationary performance are obtained by the use of combination of E[e^p(n)]. Finally, the convergence performance of the estimation accuracy is verified by the computer simulation.

抄録全体を表示

In this paper we propose the structure and the adaptive algorithm of a hierarchy type adaptive filter. This adaptive filter is composed of the filter that has the steady arbitrary transfer function instead of the delay of the conventional FIR type adaptive filter. In general, the unknown system has the multiple reflection of the signal at the signal source impedance and the load impedance. The multiple reflection enlarges the length of the impulse response of the unknown system, and the estimation becomes difficult by the conventional FIR type adaptive filter. The proposed adaptive filter estimates the relatively short impulse response that does not include any reflection and the reflection coefficients independently. Finally, the convergence performances of this adaptive filter are verified by the computer simulation.

抄録全体を表示

It is well-known that stability margins (gain and phase margins) are important quantitative indicators for evaluating stability in feedback control system synthesis. However, when we use conventional techniques based on such stability margins, it may be difficult to suppress the vibration from high-order modes of mechanical system. This paper proposes the robust controller synthesis which achieves both the conventional stability margins and the second phase margin which is a quantitative indicator for suppressing the vibration. The basic idea is to synthesize controller such that the Nyquist locus of open-loop transfer function encircles the immediate outer side of the circle which is specified by the conventional stability margins and the second phase margin. This is formulated as modified H_∞ mixed sensitivity problem with the weighting constants which are decided by the center and radius of the circle.

抄録全体を表示

Recently, many kinds of robots are developed, and there are a lot of robots which work in human living space. One of the most important interactions between a robot and human is when a human informs a robot of an object’s location. The purpose of this work is to make an interface for informing a robot of object location in a human living space with several objects. We assume that the robot has found a user by sound source localization. At the beginning, the robot recognizes pointing gesture and verbal cues of the user, and detects candidates of object location. The system recognizes pointing direction by a stereo camera, and recognizes verbal cues. The direction of the pointing gesture and the directive word are used to restrict the searching space. When multiple object candidates are detected, the system asks the user for additional features such as color name or relative location among those, and then finds one of them. We have conducted experiments on a dialog task. There were three objects in the searching space. The system is able specify the object by dialog, after which, the robot moves toward it.

抄録全体を表示

In this paper, position and angle measurements of a miniature robot by using position sensitive detectors (PSDs) are proposed. The miniature robot carries three PSDs and laser beams are irradiated on the PSDs’ surfaces. The position and angle of the robot are obtained by the positions of the laser spots. Since the PSD, which is sensitive according to the laser spot, provides a continuous signal, the measurement sensitivity is theoretically determined by the sensitivity of the PSD. The results obtained by preliminary experiments in the linear and rotational displacements are described.

抄録全体を表示

In recent years, complication of power system is progressing. Under such a situation, measures for high-speed stabilization are needed in order to prevent beforehand that the influence of the fault spreads for the whole system. If the fault occurs in power system, improvement of transient stability is expected by carrying out of generator shedding, i.e. some generators are separated temporarily. In this paper, the stability estimation system after generator shedding is constructed. Moreover, the neural network for selecting the suitable shedding generator taking account of stability after the fault was newly constructed.

抄録全体を表示

The tectonic activities that precede significant earthquakes release electromagnetic (EM) waves that can be used as earthquake precursors. We have been observing EM radiation in the ELF (extremely low frequency) band at about 40 observation stations in Japan for predicting significant earthquakes. The recorded signals contain, however, several noise components generated from the ionosphere, human activity, and so on. Most background noise in observed signal is attributed to lightning in the tropics. In this paper, we propose a method based on PCA (principal component analysis) to detect signals from large data sets. The good performance of the proposed method is confirmed.

抄録全体を表示