In order to realize the compound tactile sensing of mechanical finger, a pressure sensor and a temperature sensor are packaged in the same polymer sensing unit, and a fiber Bragg grating (FBG) is used as a sensing element. The characteristics of the pressure sensor disturbed by temperature of target object are analyzed. A back propagation neural network is used to process the tactile sensing signal of FBG, and thus the recognition of the positive pressure applied on the surface of sensing unit is achieved accurately. The simulation and experimental results show that this method eliminates the effect of the target object′s temperature on the strain sensor, and the uncertainty error of strain sensor is reduced. The compensation improves the stability of the pressure measurement and the measurement accuracy. The temperature drift rate of pressure sensor is 1.2×10-4 nm/℃ after compensation. The research can be applied to the FBG tactile sensing array installed on the mechanical finger. The temperature interference to the strain sensing can be suppressed, so that the tactile and sliding measurement system of flexible mechanical fingers have a broad application prospect.