In this paper, we design and develop a ground surface condition recognition algorithm using shoe-mounted inertial sensors. We use a pair of small sensor boxes mounted on feet with accelerometers and gyro sensors inside and detect walking steps using them. Firstly, we detect stationary stance phase using accelerometers and gyro sensors. Then, based on this information, we estimate the “Angle of Inclination” (AoI) and stability of the ground. Moreover, we estimate whether the road surface is flat or not(unstable or unpaved (covered by gravel rubble or dirt) otherwise) based on variance of AoI.
In addition, as for small undulation of surface due to dips, humps and bumps, which is hard to recognize only by a few samplings, we rely on continuous sensing data aggregated spatially from multiple users based on dead-reckoning techniques. We have developed a prototype of the proposed method. In the experiments, we show that our method cannot estimate not only walking steps correctly and but also AoIs of roads in rough trends.