Takehiko OGAWA Keisuke KAMEYAMA Roman KUC Yukio KOSUGI
A new neural network for locating a source by integrating data from a number of sensors is considered. The network gives a solution for inverse problems using a back-propagation algorithm with the architecture to get the solution in the inter-layer weights in a coded form Three different physical quantities are applied to the network, since the scheme has three independent ports; an input port, a tutorial port and an answer port. Our architecture is useful to estimate z" in the problem whose structure is y=f(x,z) where y is the observed data, x is the sensor position and z is the source location. The network integrates the information obtained from a number of sensors and estimates the location of the source. We apply the network to two problems of location estimation: the localization of the active nerves from their evoked potential waveforms and the localization of objects from their echoes using an active sonar system.
Hidetoshi YOKOO Masaharu TAKAHASHI
This paper proposes a new lossless data compression method, which utilizes a context sorting algorithm. Every symbol in the data can be predicted by taking its immediately preceding characters, or context, into account. The context sorting algorithm sorts a set of all the previous contexts to find the most similar context to the current one. It then predicts the next symbol by sorting previous symbol-context pairs in an order of context similarity. The codeword for the next symbol represents the rank of the symbol in this sorted sequence. The compression performance is evaluated both analytically and empirically. Although the proposed method operates character by character, with no probability distribution used to make a prediction, it has comparable compression performance to the best known data compression utilities.
Md. Kamrui HASAN Takashi YAHAGI
We present a new method for the identification of time-invariant multichannel autoregressive (AR) processes corrupted by additive white observation noise. The method is based on the Yule-Walker equations and identifies the autoregressive parameters from a finite set of measured data. The input signals to the underlying process are assumed to be unknown. An inverse filtering technique is used to estimate the AR parameters and the observation noise variance, simultaneously. The procedure is iterative. Computer simulation results that demonstrate the performance of the identification method are presented.
Fernando Gil V. RESENDE Jr. Keiichi TOKUDA Mineo KANEKO
A new adaptive AR spectral estimation method is proposed. While conventional least-squares methods use a single windowing function to analyze the linear prediction error, the proposed method uses a different window for each frequency band of the linear prediction error to define a cost function to be meinemized. With this approach, since time and frequency resolutions can be traded off throughout the frequency spectrum, an improvement on the precision of the estimates is achieved. In this paper, a wavelet-like time-frequency resolution grid is used so that low-frequency components of the linear prediction error are analyzed through long windows and high-frequency components are analyzed through short ones. To solve the optimization problem for the new cost function, special properties of the correlation matrix are used to derive an RLS algorithm on the order of M2, where M is the number of parameters of the AR model. Computer simulations comparing the performance of conventional RLS and the proposed methods are shown. In particular, it can be observed that the wavelet-based spectral estimation method gives fine frequency resolution at low frequencies and sharp time resolution at high frequencies, while with conventional methods it is possible to obtain only one of these characteristics.
Kazuhiro NAKAI Gaishi YAMAMOTO Toshio NAKAMURA
A filter configuration that allows configuration of any transfer function used the state variable is discribed as an application of the second generation current conveyors (CCIIs) to RC networks. The filter types discussed are low-pass filter (LPF), high-pass filter (HPF), band-pass filter (BPF), all-pass filter (APF), and band-elimination filter (BEF). The filter circuit consists of four CCIIs and allows tandem connections. The device sensitivity and CCII's sensitivity to transfer coefficient are relatively low. The filter circuit that allow simultaneous configuration wewe fabricated. An experimental result at around 10kHz was obtained for the filters. In the case, the LPF, HPF, BPF, APF, and BEF characteristics are obtained at Q value of 5.0.
Continuous nonlinearity" is stressed as a fundamental principle in pattern recognition including handprinted Kanji character recognition. Continuity" in template matching and spatial nonlinearity" in structural analysis should be unified toward deriving a higher level of recognition algorithm. At the same time, continuous nonlinearity in the temporal axis is important, as is the case of simultaneous processing of segmentation and recognition for touching characters. The above viewpoint is discussed in the following examples: nonlinear normalization, directional pattern matching, locally maximized similarity, relaxation matching, dynamic programming matching, segmentation of character string using dynamic programming, and exhaustive matching for character extraction on complex background.
Tsutomu MIYASATO Haruo NOMA Fumio KISHINO
This paper describes the results of tests that measured the allowable delay between images and tactile information via a force feedback device. In order to investigate the allowable delay, two experiments were performed: 1) subjective evaluation in real space and 2) subjective evaluation in virtual space using a force feedback device.
For similarity methods to work well, the image must be blurred before being input. However, the relationship between the blurring operation and the similarity is not fully understood. To solve the problem of this relationship, in this paper, the effect of blurring is investigated by expressing figure f(x) in the form of the sum of higher derivatives of f (x,σ), and then a simple similarity between figures was mathematically formulated in terms of the relation between visual patterns. By modifying this formulation, we propose pluralized simple similarity to increase the allowance in different view of multiple similarity. The similarity maintains higher allowance without any discernible loss of distinguishing power. We verify the effectiveness of the pluralized simple similarity throughout some experiments.
It is well known that the Hopfield Model (HM) for neural networks to solve the TSP suffers from three major drawbacks: (D1) it can converge to non-optimal local minimum solutions; (D2) it can also converge to non-feasible solutions; (D3) results are very sensitive to the careful tuning of its parameters. A number of methods have been proposed to overcome (D1) well. In contrast, work on (D2) and (D3) has not been sufficient; techniques have not been generalized to larger classes of optimization problems with constraint including the TSP. We first construct Extended HMs (E-HMs) that overcome both (D2) and (D3). The extension of the E-HM lies in the addition of a synapse dynamical system cooperated with the corrent HM unit dynamical system. It is this synapse dynamical system that makes the TSP constraint hold at any final states for whatever choices of the HM parameters and an initial state. We then generalize the E-HM further into a network that can solve a larger class of continuous optimization problems with a constraint equation where both of the objective function and the constraint function are non-negative and continuously differentiable.
Jian KANG Hiroshi ECHIGO Koichi OHNUMA Shigeo NISHIDA Risaburo SATO
The reception level of a round-trip signal from a VSAT (Very Small Aperture Terminal) was monitored continuously for three years starting October 1991.For these experimental measurements, a commercial satellite channel (up-link 14GHz/downlink 12GHz, bandwidth 100kHz) was used and rainfall was measured simultaneously. Data gathering time interval of 2 seconds was adopted to elucidate very rapid variation and lower percentage statistics. In this paper, attenuation due to rainfall is shown using the data obtained in this three-year period. It is shown that so far, the measured rain attenuation agrees very well with the values estimated using the CCIR model, and limits the range where the cumulative time exceeds 0.01%, even for our VSAT system in Tagajo, Miyagi Prefecture, Japan.
Fang SUN Shin'ichiro OMACHI Hirotomo ASO
In this paper, a new algorithm for selection of candidates for handwritten character recognition is presented. Since we adopt the concept of the marginal radius to examine the confidence of candidates, the evaluation function is required to describe the pattern distribution correctly. For this reason, we propose Simplified Mahalanobis distance and observe its behavior by simulation. In the proposed algorithm, first, for each character, two types of feature regions (multi-dimensional one and one-dimensional one) are estimated from training samples statistically. Then, by referring to the feature regions, candidates are selected and verified. Using two types of feature regions is a principal characteristic of our method. If parameters are estimated accurately, the multi-dimensional feature region is extremely effective for character recognition. But generally, estimation errors in parameters occur, especially with a small number of sample patterns. Although the recognition ability of one-dimensional feature region is not so high, it can express the distribution comparatively precisely in one-dimensional space. By combining these feature regions, they will work concurrently to overcome the defects of each other. The effectiveness of the method is shown with the results of experiments.
Caiming ZHANG Takeshi AGUI Hiroshi NAGAHASHI
A new global method for constructing a C2 piecewise quartic polynomial curve is presented. The coefficient matrix of equations which must be solved to construct the curve is tridiagonal. The joining points of adjacent curve segments are the given data points. The constructed curve reproduces exactly a polynomial of degree four or less. The results of experiments to test the efficiency of the new method are also shown.
Hidetoshi MIYAO Yasuaki NAKANO
In the traditional note symbol extraction processes, extracted candidates of note elements were identified using complex if-then rules based on the note formation rules and they needed subtle adjustment of parameters through many experiments. The purpose of our system is to avoid the tedious tasks and to present an accurate and high-speed extraction of note heads, stems and flags according to the following procedure. (1) We extract head and flag candidates based on the stem positions. (2) To identify heads and flags from the candidates, we use a couple of three-layer neural networks. To make the networks learn, we give the position informations and reliability factors of candidates to the input units. (3) With the weights learned by the net, the head and flag candidates are recognized. As an experimental result, we obtained a high extraction rate of more than 99% for thirteen printed piano scores on A4 sheet which have various difficulties. Using a workstation (SPARC Station 10), it took about 90 seconds to do on the average. It means that our system can analyze piano scores 5 times or more as fast as the manual work. Therefore, our system can execute the task without the traditional tedious works, and can recognize them quickly and accurately.
This paper describes advances in the study of handwritten Kanji character recognition mainly performed in Japan. The research focus has shifted from the investigation of the possibility of recognition by the stroke structure analysis method to the study of the feasibility of recognition by the feature matching methods. A great number of features and their extraction methods have been proposed according to this approach. On the other hand, studies on pattern matching methods of recognizing Kanji characters using the character pattern itself have been made. The research efforts based on these two approaches have led to the empirical fact that handwritten Kanji character recognition would become more effective by paying greater attention to the feature of directionality. Furthermore, in an effort to achieve recognition with higher precision, active research work has been carried out on pre-processing techniques, such as the forced reshaping of input pattern, the development of more effective features, and nonlinear flexible matching algorithms. In spite of these efforts, the current character recognition techniques represent only a skill of guessing characters" and are still on an insufficient technical level. Subsequent studies on character recognition must address the question of how to understand characters".
Kanad KEENI Hiroshi SHIMODAIRA Tetsuro NISHINO Yasuo TAN
Devanagari is the most widely used script in India. Here, a method is introduced for recognizing Devanagari characters using Neural network. The proposed method reduces the number of output unit necessary for a conventional neural network where the classification is based on a winner take all basis. An automatic coding procedure for representing the output layer of the network and a different method for the final classification is also proposed. Along with the automatic coding procedure, a heuristic method for representing the output units by exploiting the structural information of Devanagari character is also demonstrated. Besides, it has been shown by random representation of the output layer that the representation effects the generalization/performance of the network. The proposed automatic representation gave the recognition rate of 98.09% for 44 categories.
Akira WATANABE Yuuji KOUI Shoichiro SENO Tetsuo IDEGUCHI
We propose an architecture of a high-speed internetworking device using central control method. Co-operations of hardware and software is required to realize high relay performance. For the hardware, we have designed an original bus arbitration control method to achieve a high throughput of a data bus. For the software, we have devided a normal relay processing from other processing and built it as a basic function of the monitor. By this method, relay perfomance improves dramatically, because of a multiple effect of the reduction of software overheads and the improvement of cache hit ratio. We have developed the prototype device and confirmed the effects of the proposed method.
Handprinted Kanji character recognition was the most important research topic for 1980s. Today there are several commercial products of handprinted Kanji OCR, but there are many unsolved problems. In this paper three types of research approach are focused for handprinted Kanji character recognition. The first approach is directional pattern matching, which is generally applied to handprinted Kanji character recognition. The second approach is potential field feature extraction, which activated Kanji character recognition research in the early stage of research. The third approach is shape matching. This paper surveys these research approaches, and both contributions and problems of them are discussed.
Among various characters invented by the human being ,Kanji (Chinese character) is outstanding in its diversity and complexity, in contrast to Roman alphanumerics. So machine recognition of handwritten characters requires particular technics. This paper concerns commercially available optical character readers (OCRs) and the recognition techniques using therein. Methods of character recognition is classified the pattern types and the method of extracting features, to discuss the present states of the character recognition. In regard to the structural analysis, the proposed techniques are classified and discussed in accordance with the extraction method of line segment and axial correlation. Finally, various techniques are compared with one another by using a common database, so as to understand the present states of character recognition and to discuss the technical trends.
Masato SUZUKI Nei KATO Hirotomo ASO Yoshiaki NEMOTO
In recognition of handprinted characters, it is important to dissolve distortions of character caused by writer's habits. In order to dissolve distortions and to obtain better features, many image conversion methods have been proposed. But there are distortions that cannot be dissolved by these methods. One example is the case of parallel strokes which are spread out in fan shape. In this paper, in order to dissolve distortions, we propose a new image conversion method, Transformation based on Partial Inclination Detection (TPID)", which is employed just before normalization, and is intended to dissolve several kinds of distortions in images of each character. TPID constructs transformation functions from inclination angles which are detected in some subspaces of the character's image, and converts images using the transformation functions. TPID is especially suitable for correcting the inclinations of horizontal and vertical strokes of a character. This has a powerful impact on the quality of the characteristic features. In recognition experiments using ETL9B, the largest database of handprinted characters in Japan, we have obtained a recognition rate of 99.08%, which is the best to our knowledge.
Masataka OHTSUKA Yoshihiko KONISHI Makoto MATSUNAGA Takashi KATAGI
In this paper, authors propose a linear array antenna using two bifilar helical antenna elements placed along the helix axis to reduce beam direction movement according to frequency change. The beam direction movement of this proposed array antenna is smaller than that of a conventional bifilar helical antenna. Also, the gain of this proposed array antenna is higher than that of the conventional helical antenna for a cross point angle of radiation patterns at the different transmit and receive(Tx and Rx) frequencies. The conventional helical antenna is suitable for vehicle antennas in mobile satellite communication systems such as the MSAT system because it owns circularly polarized omni-directional radiation pattern and its thin pole form. However, this antenna has a disadvantage that the beam direction in an elevation plane moves according to frequency change. In the proposed array antenna, the beam direction movement is about 9 smaller than that of the conventional bifilar helical antenna on condition that antenna total length is 4.83 λ0, antenna diameter is 0.12 λ0, and frequency change is from 0.957f0 to 1.043f0(f0 is center frequency and λ0 is free space wavelength at f0). Also, the Tx and Rx gains of this proposed array antenna at the cross point angle between Tx and Rx beams are about 2 dB higher than that of the conventional bifilar helical antenna on the same condition.