One　of　eminent　activity　recognition　approaches　is　based　on　the　extended　belief　rule-based　system　(EBRBS),　which　shows　an　excellent　robustness　comparing　with　many　traditional　approaches.　For　this　reason,　three　versions　of　EBRBSs,　namely　original　EBRBS　(O-EBRBS),　DRA-EBRBS,　and　Micro-EBRBS,　are　adapted　in　the　comparative　analysis　to　illustrate　the　performance　of　these　EBRBSs.　Results　demonstrate　that　the　Micro-EBRBS　can　produce　a　satisfied　accuracy　under　less　time　of　inference　scheme　and　fewer　numbers　of　activated　rules　comparing　to　the　O-EBRBS　and　DRA-EBRBS.