In　recent　years,　human　motion　recognition　acceleration　sensor　has　become　one　of　the　active　research　direction　and　the　emerging　field　of　computer　pattern　recognition,　more　and　more　researchers　use　acceleration　sensor　in　motion　monitoring　of　human　experimental　environment　and　natural　living　environment.　Human　motion　recognition　mainly　includes　data　acquisition,　data　processing,　feature　extraction　and　selection,　and　classifier　design.With　the　further　development　of　microelectronics　and　sensors,　sensors　can　be　embedded　in　mobile　devices,　portable　devices　and　so　on.　The　use　of　acceleration　sensor　for　human　motion　monitoring　has　broad　application　prospects.　On　the　one　hand,　the　daily　exercise　behavior　is　closely　related　to　the　physical　and　mental　health.　It　is　of　great　significance　for　the　development　of　scientific　exercise　and　fitness　program　to　improve　the　health　status.　On　the　other　hand,　it　is　of　great　value　to　monitor　the　abnormal　motion　of　human　body.　This　kind　of　abnormal　exercise　can　bring　some　harm　to　the　human　body.　It　can　provide　timely　help　to　the　fall　and　reduce　the　risk　of　falling.　Through　the　comparison　between　threshold　and　peak　vertical　acceleration　sensor　data　waveform　threshold　trough　peak　and　valley　time　intervals　of　three　characteristic　values,　and　combined　with　the　principle　of　decision　tree　classification　to　achieve　accurate　recognition　of　squatting,　jumping　and　running　action.　A　human　fall　detection　algorithm,　the　algorithm　falls　by　extracting　human　motion　features　in　the　process,　focusing　on　the　use　of　the　three　axis　acceleration　sensor　correlation　to　fall　over,　fell　back,　left　and　right　fall　four　to　fall　to　distinguish　fall　behavior.　The　experimental　results　show　that　the　average　recognition　rate　of　human　motion　behavior　is　more　than　90%,　which　has　a　certain　practical　value.　©　2017　IEEE.