Kenji Kurosu

Neural network vowel-recognition jointly using voice features and mouth shape image

Abstract This paper describes a neural approach intended to improve the performance of an automatic speech recognition system for unrestricted speakers by using not only voice sound features but also image features of the mouth shape. In particular, we used the natural sample voice signals and mouth shape images that were acquired in the general environment, neither in the sound isolation room nor under specific lighting conditions. The FFT power spectrum of acoustic speech was used as the voice feature. In addition, the gray level image, binary image and geometrical shape features of the mouth were used as the compensatory information, and compared which kinds of image features are effective. For unrestricted speakers, a vowel recognition rate of about 80% was obtained using only voice features, but this increased to some 92% when voice features plus binary images were used. This method can be applied not only to the improvement of voice recognition, but also to aid the communication of hearing-impaired people.

Diagnosis of Engine Trouble by Fuzzy Logic

Abstract This paper consists of two parts. The first one is the principal component analysis of engine trouble. The authors investigate the reports of engine trouble of 120 large merchant ships over three years, and 89 items such as the location of failure, kind, damage, detection method, estimation, diagnosis, counterplan and economical loss are analyzed. As a result, it becomes clear what kinds of trouble are most effectively avoided, and what kinds of performance are expected for the safety system. The second part is the development of a new algorithm of diagnosis by fuzzy logic. The reason why the authors adopt fuzzy set theory is that the human perception and decision, which are very subjective and fuzzy, play the leading part in case of trouble. In this safety system, the signal from the mechanical sensors and that from the human sensory are processed together. For this purpose, a new inverse operation of fuzzy relation is developed. The results of diagnosis are indicated on C.R.T. as short sentences, which tell the operator clearly or vaguely the locations, causes and countermeasures of the trouble.

Fuzzy control of group with a leader and their behaviors

A method is proposed for analyzing group behaviors described by the dynamic equations of the individual members guided by a leader. Behaviors described by fuzzy rules were studied to see whether the leader can guide the individual members. The simulation results lead to the following conclusions: (1) Simple rules make it possible for members to form a group. (2) A leader with a fuzzy controller can guide the members or confine the members in a circle. (3) Some members placed far from the group are susceptible to straying. (4) The position of the front of the group is a greater determinant of guidance control than the center of the gravity. (5) Zigzag movements can make the group move slowly, enabling a delayed member to catch up to the group. (6) Parameter changes cause apparent effects on the behaviors of the group.

Design of Adaptive Control System for Flexible Robotic ARM

Abstract This paper presents a design method of a MRACS for a flexible robotic arm. The flexible robotic arm is simply modelled as a flexible cantilever beam revoluting at a joint. First, the mathematical formulation of the flexible beam is induced. Discretizing this model by sampling period, a discrete time equation is obtained. Then, a discrete MRACS is designed for the discrete model. A set of experiments has been carried out to confirm the theoretical results. In the experiment a CCD camera was used to detect the position of the end of the beam, which was driven by stepping motor through a microcomputer. Experimental results have shown that the beam can be controlled effectively by the proposed adaptive algorithms, thus the theoretical result is validated.

Adaptive Control for a Robot ARM and a Robot Drive System with Elasticity

Abstract First, this paper presents an application of a Model Reference Adaptive Control System (MRACS) to the position control, and the force feedback control of a flexible robot arm. Second, this paper investigates the application of an MRACS to the speed control, and the position control of a DC servo system, and a Direct-Drive(DD) servo system with a flexible transmission element. The flexible robot arm is simply modelled as a flexible cantilever beam. The servo systems with a flexible transmission element are treated as a servo mechanism with a one degree of freedom oscillatory load Using discrete-time mathematical models as a lumped parameter system, the discrete MRACS is designed for each discrete model Experimental results have shown that the control performance is good and the MRACS is effective.

Speaker‐independent vowel recognition combining voice features and mouth shape image with neural network

This paper describes a neural approach intended to improve the performance of a voice recognition system for unrestricted speakers using not only voice sound features but also image features of the mouth shape. The FFT power spectrum of acoustic speech was used as the voice feature. In addition, the gray-level image, binary image, and geometrical shape features of the mouth were used as the compensatory information and a comparison made of which kinds of image features are effective for voice recognition by a neural network. For unrestricted speakers, a vowel recognition rate of about 80 percent was obtained using only voice features. However, this increased to some 92 percent when voice features plus binary images were used. This method can be applied not only to the improvement of voice recognition, but also to aid the communication of hearing-impaired people.

Structural Model of Misoperation

Abstract In this report, general factors which cause human errors during plant-operation are chosen and these factors are classified into some groups by K.J. Method. Using the above results, a questionnaire is made in order to know the importance and the relation of these factors. The replies submitted to the questionnaire are analyzed by DEMATEL Method and the major factors of misoperation are summarized in the form of the structural model. Two case studies are mentioned, one of which is the misoperation in transportation and the other is in hot strip mill. From these results, the ways how to prevent misoperation are discussed.

CONDITION MONITORING OF MARINE ENGINES BY MEANS OF FACE PATTERN METHOD

The face pattern method is considered to be preferable for marine engine monitoring because it is suitable for understanding correlations of multi-variables at a glance and for noticing their sensitive changes in details. The state variables in abnormal states of a marine turbine plant system and a diesel one are displayed on a cathode ray tube by the face pattern method. Order from NSFI as No. 22482.

Failure Diagnosis of Plants in Operation

Abstract When plants are running normally, the ordinary process controllers, which have been developed till now, act excellently, but they are almost useless in an abnormal condition. Usually, simple automatic trip systems are equipped only for emergency use. These emergency control systems are so poor that the human operator is indispensable even in fully automated plants. On the other hand, if we equip the plant with many fault detectors, the plant may frequently make unnecessary halts due to noise or troubles in detectors. The purpose of this paper is to develop a new type emergency control system which keeps the plant safe while making the down time minimum. The new system has two major actions. One is the automatic diagnosis finding the place of troubles, either in the plant side (including normal state) or in the detectors’ side. The other action is the prediction of the plant state if it goes into the dangerous state or not. This is very important, because most of the mis-downs of plants are caused by the temporary fluctuation of some controlled variables. The basic idea of the diagnosis is the combination of “maximum likelihood” and “two out of three”. The prediction is done by “simulation”. After building the emergency control system, the authors tested it with a steam plant. This experiment gave fine results. The artificial troubles caused by us were completely diagnosed and predicted.