This paper presents a new tactile sensors with multi artificial whiskers, which mimic whiskers of animals. The proposed sensor will distinguish the types and detect the attitudes or positions of contact objects from the condition of the deflection direction of each whisker that contacts objects. The sensor consists of plural sensing elements, each of which has a flexible whisker made of Ti-Ni alloy. The whisker is equipped with a plane synthetic rubber board. The recognition by the proposed tactile sensor is performed in the following two steps. First, the sensor will recognize the deflection direction of each whisker by the deformation of the synthetic rubber board. Next, it recognizes the atttitude or position of a contact object with its type from the combination of the deflection direction of each whisker. Both the recognitions were made by the neural network. The presents study examines the ways of the constructions of each neural network for the respective recognition steps, and confirms the effectiveness of constructed each neural network by the experiments and numerical simulations. Furthermore, the experiments verify the validity of the proposed tactile sensor.