Rokuya Ishii

Accurate estimation of minimum filter length for optimum FIR digital filters

This paper presents an accurate estimation formula of minimum filter length for optimum (minimax criterion based) linear-phase finite impulse response (FIR) low-pass digital filters. Two estimation formulae have been already proposed; however, they have some problems which lead to wrong estimation. Discussing the problems based on our experimental results, a new estimation formula is newly proposed. The accuracy of the proposed formula is evaluated in comparison with that of the conventional formulae. Furthermore, the proposed formula is applied to the design of high-pass, band-pass, and band-stop filters. The estimation accuracy is discussed also in this case.

Neural net implementation on single-chip digital signal processor

Madaline rule II (MRII) and back-propagation (BP) algorithms have been implemented on a digital signal processor (DSP). Two kinds of modifications of MRII are proposed: a tree search for the best up-to-two-order combinations of neurons in a randomly chosen layer and an efficient way of setting the desired response value for the least-mean-square (LMS) adaptation of the neurons. A sigmoid table lookup function and some details of the implementation of the BP algorithm are presented. Perceptron span limitations, as the maximum number of neurons per layer, and processing times are given for both systems. This gives a good understanding of the general requirements for the implementation of perceptrons on DSP, such as memory space, data flow, and multiplier functional needs. The training behavior of the BP program on DSP is analyzed with reference to the example of handwritten character recognition. In spite of the low accuracy of DSP floating-point data, the perceptron simulation on DSP shows better results than a C-simulation program on a personal computer.<<ETX>>

An image processing method for use in a GUI for the visually impaired

This paper presents a method for processing the image of a human hand, in order to emulate the acts of moving a scroll bar by rotating the hand, and clicking on a mouse button by moving the thumb. This method can be used in a system that facilitates the operation of a graphical user interface (GUI) by a visually impaired person. The grayscale image acquired by a digitizer is thresholded and converted to a black-and-white image. The orientation of the hand, given by an angle /spl theta/ with the vertical axis, is calculated based on the central moments. The image of the hand is then rotated by /spl theta/ to a normalized position. The number of pixels in each column of the normalized image is counted, and the result is expressed in a histogram. The coefficient of asymmetry of this histogram is used to determine whether the thumb is positioned along the pointing finger or whether it is far from the other fingers, defining two states that correspond to a mouse button up or down. Experimental results showed a frame rate of 12.5 frames per second, demonstrating that this method is suitable for the real time processing demanded by the interface system.

Hand shape extraction from a sequence of digitized gray-scale images

Gesture recognition is currently a major trend in human-computer interface researches. In a gestural interface, the extraction of the hand shape from a digitized scene is the first step before the analysis and recognition of the gestures themselves. Aiming at the implementation of a gestural interface, in this paper the authors present a method to extract fingertip coordinates from a sequence of images. The authors propose a method comprising histogram analysis, using the results from movement detection in order to find threshold values, and applying gray level slicing. This overall processing solves some problems found in conventional methods, such as the influence of shadows. A potential application would be in the implementation of an input device that substitutes the mouse.<<ETX>>

A recognition method of restricted hand shapes in still image and moving image as a man-machine interface

This paper presents a recognition algorithm of restricted hand shapes and of a moving hand that is based on detailed analysis of restricted hand shapes as a man-machine interface. The hand shape recognition algorithm consists of two parts: hand region extraction process and hand shape recognition process. Furthermore, the algorithm can be applied to recognition of a moving hand. In this paper, recognition of a moving hand is reduced to recognition of still image. A moving hand is recognized by the proposed method in real time.

A GUI support system for a sight handicapped person by using hand shape recognition

The paper proposes an algorithm of hand-shape recognition and applies it to a graphical user interface (GUI) support system for a sight handicapped person. This algorithm is to recognize the angle of a hand and ten hand-shapes. As one application of this algorithm, we implement a system in which the angle of a hand controls a reading point in text on a screen in real time, and the text is read out with sound synthesis software. Further, in this system, more complicated directions to control a text reader can be done by recognized hand-shapes.

The property of bilinear transformation matrix and Schur stability for a linear combination of polynomials

Abstract In this paper, Schur stability for a linear combination of polynomials is studied. In order to investigate this stability, we first study some important properties of the transformation matrix derived by using the bilinear transformation. And then, under certain assumptions, necessary and sufficient conditions for the linear combination of k polynomials to be Schur stable are obtained.

An automatic design procedure of IIR digital filters from an analog low-pass filter

Abstract IIR digital filters are generally designed from an analog low-pass filter by any combination of the bilinear transformation, frequency transformation for analog filters and that for IIR digital filters. But the direct application of these transformations yields a complicated formula of the target transfer function, which has to be reduced by hand computation into the form of a rational polynomial. This hand computation will be replaced by an automatic procedure if the relation between the coefficients of the transfer functions is formulated. This paper derives the explicit formulae which connect those coefficients by matrices.

Image magnification by using spectrum extrapolation

Presents a new image magnification algorithm that produces high frequency components of the spectrum of an image. To estimate high frequency components, the authors investigate the iterative algorithm of Gerchberg-Papoulis with various extrapolation methods. The authors apply this algorithm to digital images. In the application of this algorithm, there is one problem that greatly affects its efficiency. The authors propose two techniques to overcome this problem. Using these techniques, the authors can obtain magnified images with higher quality.<<ETX>>

A periodically time varying digital filter containing an inverse discrete Fourier transformer and its application to the spectrum scrambling

An output sampling polyphase model containing an inverse discrete Fourier transformer is proposed and analyzed. The input-output relationship of the proposed model is obtained, and its periodically time varying property is discussed. An application of the proposed model to spectrum scrambling is considered. A structure suitable for realizing the spectrum scrambling within the scope of real computations is presented, and its design procedures based on the bifrequency map are described. It is demonstrated that the proposed structure can be used to implement a spectrum scrambler, whether FIR or IIR (finite or infinite impulse response) filters are used. A design example is given to verify the effectiveness of the proposed design procedures for the realization of a spectrum scrambling system when both FIR and IIR time-invariant filters can be used.<<ETX>>