Currently, voice recognition is widely used in word processing software, car navigation systems, and for other purposes. Existing systems learn the features of an individual's voice and then compare them with input speech to recognize the speaker, or search through a recognition dictionary to find words having acoustic features closest to those of input speech in order to recognize what the speaker is saying, but these systems cannot recognize arbitrary speech spoken by an arbitrary speaker, and they have problems recognizing speech in a noisy environment. In this study the MT system was used in an attempt to establish a speaker-independent voice recognition technology that does not require learning processes or recognition dictiollaries. Recognition was performed on phonemes. Vowels and consonants were treated separately: vowels were recognized by extracting one waveform from a repetitive waveform; consonants were recognized by double checking. The results indicate that voice recognition from-individual phonemes is possible.

Qualitative scoring sheets are sometimes used to evaluate the credit-worthiness of an enterprise. Ensuring the objectivity of the qualitative evaluation and obtaining a scale that makes accurate credit evaluation possible have been issues of concern in this method. As scoring sheets are qualitative and attempt to score many different items, it would be desirable to have a unified evaluation with a quantitative value. In this study, we considered an array of scores as a type of pattern and used the MT system to obtain a unified value for the pattern from the evaluation scores. Furthermore, we assessed the degree of contribution of each scored item, using an orthogonal array to visualize regularities in the varying patterns of different enterprises. Thirty-six qualitative scores were thereby evaluated in terms of a single value, and we also succeeded in visualizing latent structures behind the score information by using a factorial effect diagram obtained from the orthogonal array.

One of the main activities of the quality assurance division to which the authors belong is approval of new products, in which a new product is evaluated as to whether it will stand the test of actual use in the field. Evaluation conditions were formerly based on specifications, but evaluation took a long time and left room for improvement in regard to accuracy. To reform the engineering process, eliminate backtracking, and reduce development lead times, we had to reexamine our evaluation methods. As a first step taken in a challenge to develop a "test-free" methodology, specification-based evaluation tests were eliminated. They were replaced by a functional benchmarking method in which the intrinsic function of a new product is measured, a mixture of noise factors derived from a series of studies is added, and the S/N ratio of the product is compared with that of an existing field-proven product. This method is expected to reduce evaluation time drastically and ensure better assessment accuracy.

As longer life is required for the molds used in die-cast molding, these molds are being made of tougher and more rigid materials which are hard to machine. In particular, drilling holes in mold materials has become less efficient, which is one of the factors that have increased the cost of the molds. Based on past case studies, the present study evaluated drilling performance in terms of electrical energy required, evaluated cutting efficiency in terms of electrical energy versus machining time, and also evaluated breakage and abrasion of drills. Optimal drilling conditions were derived from these evaluations, the drill feed rate was increased by a factor of five in view of the trends in factorial effect diagrams,and a new evaluation was carried out. The results were of highly practical value: die-cast mold drilling time was reduced to one fifth and surface roughness was reduced to one third of the previous values.

It is known that the number of sick people increases when sudden temperature changes occur at the turning of the seasons. Noise factors such as daily changes in environment, body, and behavior can be considered a necessary part of the data used for evaluating health, and for health tuning, it is also necessary to recognize the factorial effects of control factors which are prescribed for each person. In this study, tuning to a healthy condition was attempted by taking signals from living environment data, body data, and behavior data, which can be easily collected by untrained people, and by creating an MT system unit space scaled according to subjective feeling of health as the true value. This report is based on data collected in winter. Data collection and research will continue for the coming three years.

Press forming is used at many manufacturing sites as a means of producing large quantities of products having the same shape in a short time. Press drawing,which can create a seamless three-dimensional shape, has the advantage that procedures such as welding can be skipped and less material is wasted, but friction becomes a major problem in this process. As friction generates blemishes,cracks, and heat in the product, it is commonly believed that a lubricant must be used during the pressing process, but then the lubricating oil must be washed off before the press-formed products are shipped. To establish a lubricant-free pressing (dry pressing) technology, which would greatly contribute to improvements in two major problem areas, environment and cost, a fundamental study directed toward optimization of dry pressing conditions was carried out using quality engineering techniques. As a result, a method was developed in which distortion generated in a pressing machine is used to evaluate pressing stability, and a start was made on the optimization of the dry pressing conditions.