In　human-robot　interaction,　it　is　essential　for　robots　to　accurately　model　and　predict　the　movements　of　their　human　partners,　allowing　them　to　adjust　their　behavior　in　advance　to　avoid　conflicts　with　humans　and　to　improve　the　efficiency　and　safety　of　the　interaction　process.　This　paper　aims　to　develop　a　highly　accurate　motion　model　to　model　and　predict　the　motion　trajectory　of　the　human　hand.　Firstly,　the　causes　and　characteristics　of　the　errors　generated　by　the　existing　minimum　jerk　model　(MJM)　in　modeling　the　trajectories　of　real　human　arm　reaching　movements　are　investigated,　and　a　new　motion　model　called　MMJM　is　proposed　based　on　a　2nd　order　Fourier　modification　term.　Secondly,　how　to　use　the　proposed　MMJM　to　predict　the　human　arm　reaching　motion　is　studied.　Finally,　the　effectiveness　of　the　proposed　MMJM　is　validated　by　the　human　arm　reaching　movement　experiments,　which　shows　that　the　MMJM　has　better　modeling　accuracy　and　prediction　accuracy　than　the　existing　MJM.