Recently, acceleration sensors / gyro sensor and so on have been embedded in smartphones and be-come popular. We can gather various kinds of information simply and on a massive scale by using such familiar information device. Therefore, in this study, we focus on acceleration data gathered by smartphone embedded sensors. Analysis the acceleration / angular velocity data by statistical processing to restrain the influence of measuring condition and extract features. Then, we propose a simple assessment method that utilizes features and using a Support Vector Machine (SVM) to classify the road surface roughness. Moreover, confirms the effectiveness of the proposed method.