It is necessary to ascertain tactile texture recognition of humans in developing tactile devices that create or detect lifelike texture. Tactile texture has been researched in the field of psychophysics and neuroscience, but its function-structure relationship cannot be still uncertain by their method. In this paper, modeling of tactile texture recognition mechanism by numerical simulation system is discussed. The model is made of two parts ; reception part and perception part. Reception model is built by finite element model of human finger and filters emulating mechanoreceptors response properties. Perception model is built by feed forward type of neural network with 3 layers. This network is considered for configuration and module structure of primary somatosensory area. Finally, validity of the total model is confirmed by comparing output data of this model to the result of sensory evaluation.