In this paper, we construct a system that recognizes user's walking behavior using smart insole with FSR (Force Sensitive Resistor). We analyze the data measured by smart insole using CNN (Convolution Neural Network to classify user's behavior as walking, going down stairs, going up stairs, standing and lying down. Especially when the user's walking behavior is walking, we divide user's walking by normal walking and abnormal walking (out toe, in toe, bent waist). At the same time, we count the steps using average value of pressure applied the smart insole. So we can calculate ratio of abnormal steps.