**1. Introduction**

162 Advanced Topics in Measurements

Sebe, N., Cohen, I., Garg, A., Huang, T. S. (2005). *Machine Learning in Computer Vision,* ISBN-

by Springer, Printed in the Netherlands

13 978-1-4020-3275-2 Springer Dordrecht, Berlin, Heidelberg, New York, Published

In this study, to measure the human finger's tactile sensation capability of recognizing a fine surface texture using psychophysical experiments, a computer-controlled measurement system that presents fine step-heights of 0 to 1000 µm to human subjects' fingers at various presentation angles were developed. The measurement system can control four parameters of fine step presentation, i.e., the step-height, presentation velocity, presentation angle, and presentation temperature. In psychophysical experiments of this study, the measurement system calculates the amounts of step-heights based on the Parameter Estimation by Sequential Testing (PEST) method (Taylor & Creelman, 1982) and presents the step-heights to subjects' fingers in order to measure difference thresholds and subjective equalities for fine step-heights. Those values are considered to be the fine step-height discrimination capability of finger's tactile sense.

Human finger's tactile sense is a measurement system that can detect subtle surface roughness and smoothness by touching the surface. This finger's tactile sense is much more robust than the tactile sensors developed so far for robot tactile recognition. These sensors for robot still cannot reach the performance of recognizing such fine roughness or smoothness as humans can. Therefore, it is important for engineering, as well as for psychology, to investigate the finger's tactile recognition mechanism.

So far several researchers have examined the finger's tactile sense mechanism in detail using microneurography and psychophysical experimentation. In the microneurography, a tungsten microelectrode was inserted into tactile-related nerve fibers in an arm of humans or monkeys and the reactions of the tactile sense to the stimuli presented to the hand were examined via the signals sensed by the microelectrode. In the psychophysical experiments, on the other hand, several magnitudes of stimuli were presented to human hands and the responses of the tactile sense to the stimuli are analyzed through the replies to questions regarding the stimulus magnitudes.

The microneurography found out that the human tactile organs consist of four types of mechanoreceptive units: Fast adapting type I unit (FA I), Fast adapting type II unit (FA II), Slowly adapting type I unit (SA I), and Slowly adapting type II unit (SA II) (Vallbo & Johansson, 1984; Salentijn, 1992), and it is considered that FA II responds to a subtle mechanical vibration, FA I or FA II to surface unevenness and SA I to a pattern like Braille

Measurement System of Fine Step-Height

same condition, *S*0.5 should be equal to *δs*.

Fig. 1. An example of a discrimination characteristics curve.

The values of *Δ<sup>U</sup>* = *S*0.75 − *S*0.5 and *Δ<sup>L</sup>* = *S*0.5 − *S*0.25 are the upper and lower thresholds for *δs*, respectively. Moreover, the average of the thresholds, *Δ =* (*ΔU* + *ΔL*)/2, is called the difference threshold. In addition, these thresholds usually have very close values because the upper and lower thresholds become almost equal. Also the value of the ratio of *Δ* to *δs* is called the Weber fraction. The value is known to be constant over the range of stimulus

magnitude in tactile sensing mechanisms, as well as in visual and auditory.

Discrimination Capability of Human Finger's Tactile Sense 165

determined using the experiments are important values for investigating the human tactile sensation. The meanings of those values are explained in the following (Gescheider, 1985). In an experiment, human subjects touch several pairs of stimuli with their fingers and try to distinguish them. One of the stimulus pair is the standard stimulus and the other is the comparison stimulus. The magnitudes of the standard and comparison stimuli are denoted by *δs* and *δc*, respectively. The standard stimulus is designed to be constant and the comparison stimulus is variable. Several pairs of *δs* and *δc* are presented to the subjects and for each pair they are asked to tell which stimulus of *δs* and *δc* they feel stronger. When *δc* is smaller than *δs*, the proportion of the responses that *δc* is chosen as stronger than *δs* is supposed to be low. Conversely, when *δ<sup>c</sup>* is greater than *δs*, the proportion of the responses that *δc* is chosen as stronger than *δs* is supposed to be high. Figure 1 shows a characteristic curve of the proportion that *δc* is chosen as stronger than *δs*. The horizontal axis shows the comparison stimulus while the vertical axis shows the proportion of the subjects selecting the comparison stimulus. The comparison stimulus magnitudes for the proportions equal to 0.25, 0.5, and 0.75 are denoted by *S*0.25, *S*0.5, and *S*0.75, respectively. The value of *S*0.5 is called the subjective equality for *δs*. If the standard and comparison stimuli are presented under the

dots, respectively (Heller, 1989). On the other hand, the psychophysical experiments (Miyaoka et al., 1993, 1996, 1997) determined that the human tactile mechanism is able to detect a mechanical vibration of 0.2 µm in amplitude and a surface unevenness of 3 µm in amplitude. Also, the psychophysical experiment (Kawamura et al., 1996) revealed that FA I plays an important role in discriminating the magnitudes of step-height of around 10 µm. From these experimental results, it is considered that, like the human visual sense, the human tactile sense has several kinds of module mechanisms, and it is supposed that the human tactile modules are classified based on the stimulus magnitudes they can detect and discriminate and their information processing characteristics: the subtle stimulation detection module, fine texture recognition module, two-dimensional pattern recognition module, and three-dimensional shape recognition module. So far the authors have been investigated the tactile sensation capability of recognizing fine step-heights with respect to the fine texture recognition module.

Using a measurement system that presents fine step-heights of about 10 µm to subjects' fingers (Miyaoka et al., 1996; Kawamura et al., 1996), the difference thresholds for a 10 µm step-height were determined when the subjects actively touched the step-height with their fingers moving over the step-height and when they were passively touched the step-height presented to their fingers by the movement of the step presentation device. As a result, the difference thresholds for a 10 µm step-height in the active- and passive-touch experiments agreed approximately. Therefore, it was concluded that the finger's discrimination capability of fine step-heights of about 10 µm does not depend on the touching manners. Also, the paper (Kawamura et al., 1998) suggested that when the subjects discriminated a pair of the 10 µm step-heights presented at the different presentation velocities of 20 and 40 mm/s to their fingers, they perceived the height of the fast moving step-height to be a larger stimulus than that of the slowly moving step-height due to the influence of the stimulus velocity. Furthermore, the authors developed a measurement system that can create fine step-heights of 0 to 1000 µm and present the step-heights to subjects' fingers at various presentation angles (Kawamura et al., 2009).

In this paper, to measure the finger's tactile sensation capability of discriminating fine stepheights, the developed measurement system is used. In the psychophysical tests, the presentation angle of a step is defined as the angle to finger's length and several pairs of fine step-heights of 0 to 100 µm are presented to subjects' fingertips at various presentation angles. This paper first describes the measurement system that controls the amounts of step-heights according to the experiment procedure based on the PEST method in order to determine subjective equalities and difference thresholds for fine step-heights, then examines the effects of the touching manner of human finger, finger's motion direction, and fingertip region on the tactile recognition of fine step-heights. In the psychophysical tests, first, the subjects discriminate step-heights of 10 to 100 µm in active- and passive-touch manners using the center of their fingertips. Next, the subjects discriminate step-heights of around 10 µm using the top and center of their fingertips in various motion directions of their fingers.
