Recently, mobile information devices and environment of mobile communication have progressed quickly. Under the situation, the technology which can monitor human activity and position has been needed in many fields and many developments and researches about it have been performed. Actually, a lot of systems using GPS and PHS have been proposed and put to practical use, though, there are still serious problems in the time-resolution and the limit of its usable area. For that reason, an autonomous positioning system not to depend on infrastructures has been developed, which can estimate walking human's position continuously and high time-resolution by combinations of same sensors with walking human and Map-matching to modify errors occurred by the sensors. But these positioning systems can be used only when users walk on smooth or flat surface, not applicable to the other various human's motion modes such as walking up and down the stairs, taking bicycle, car, train, bus etc. This paper, after investigating the characteristics of each motion mode pattern, try to estimate the various human's motion modes using the time-series motion data measured from a human body.