As few research works have been done on robotic finger-type soft touch-sensor system for classifying the multiple features of object such as size and softness, we have developed a soft touch-sensor array system of 5-links robotic finger-iype together with a statistical data processing of spatial-temporal on-off switch data that are obtained in the course of grasping an object. The finger-type touch-sensor body is covered with a semi-circular soft silicone rubber. With compliance of 0.004m/N within which small mechanical touch-sensor switches (compliance = 0.001mlN) are implanted at 5 x 4 7 grid points. The spatial and temporal on-off switch data are decomposed into the frequency distribution of 8-neighborhood-cell patterns. By applying Maharanobis distance categorization method, the dynamic found more effective to achieve the 100% multi-sensing is f ple-feature classification of the 12 tested circular cylinders with respect to size and softness than the static one.