This paper is a study on the valuation and mathematical formulation of human touch sensation. First, the feelings of touch are collected on five natural and chemical fabrics through the SD (Semantic Differential) method questionnaires sent to twenty three examinees. The feelings are valued on each fabric by using the weigthed mean on graded factor feelings. The multiple regression analysis is then introduced to represent the overall evaluation of touch on every fabric by a linear function of the valuations on feelings of touch. Next, a soft tactile sensor made of a PVDF (Polyvinylidene Fluoride) film patch and rubber layers is assembled and slid over the same sample fabrics to collect the surface tactile information on the fabrics. The features on the collected data are then extracted by calculating the temporal average of absolute out signal and using the intensity of power spectral density on the medium frequency range. Comparison of the results show that the PVDF sensor well describes the tactile perception of forefingers.