To bring products giving a high degree of user satisfaction to market, their ease of use must be evaluated quantitatively. An attempt was therefore made in the present study to use the Mahalanobis-Taguchi system to detect stress, which is a human emotional variable, from biomedical information. This method was tried out on five experimental volunteers. Four tasks intended to produce a resting state were prepared for the unit space. Fifty-five features were extracted from plethysmogram and electroencephalogram spectra and the Mahalanobis distance was found. The results of this experiment confirmed that in a stressed state induced by mental arithmetic, the Mahalanobis distance from all four restful tasks was large. From an analysis of feature effects, a tendency to agree with the plethysmogram and electroencephalogram findings was observed.

View full abstract

In the two step optimization procedure advocated by quality engineering, first a robust set of design conditions with high noise tolerance is found; then, if necessary, the conditions are tuned toward the target characteristic. When the target characteristic is expressed as a curve, orthogonal polynomials are useful in the tuning process. Two types of orthogonal polynomials are available: Chebyshev general orthogonal polynomials, which have constant terms; and proportional orthogonal polynomials, which lack constant terms. Criteria for choosing between these two types have been unclear. In the present study, this issue was addressed with respect to the operating force curve of a switch by using theoretically derived orthogonal polynomials of both types and comparing the tuning processes. Both types were applied to the same materials, making possible a concrete comparison between them. Both types of orthogonal polynomials led to nearly the same final tuning accuracy and yielded nearly identical design specifications, but the proportional orthogonal polynomials were found to be advantageous because they were easier to work with and were more generally applicable.

View full abstract