As mobile robots are becoming increasingly more popular in workplaces, public places and homes, the understanding of surface dynamics is one of the new challenges to improve their mobility and reduce overall disturbance in the human environment. This work proposes a comparison between different classifiers of identifying surfaces by using a 3-axis accelerometer sensor on a holonomic robot. The robot is equipped with four omni-directional wheels and the sensor is mounted on its frame to measure the whole body vibration. Four typical indoor floors were tested over four robot motions. Final results show that a mixture of statistical and spectrum density based features are sufficient to identify surfaces with more than 85% overall accuracy.