In　the　processes　of　gene　expressing,　gene　expression　data　at　each　time　point　is　different.　Each　gene　expression　levels　in　a　time　point　affects　gene　expression　levels　in　next　time　point.　It　is　a　hot　point　to　construct　gene　regular　network　using　time　series　gene　expression　data.　There　are　many　methods　being　used　for　this　work,　such　as　Boolean　networks,　Differential　equations,　Bayesian　networks　and　so　on.　In　this　paper,　we　build　transfer　relationship　of　gene　as　gene　observation　matrix　by　correlation　coefficients　and　P-value　of　time-series　gene　expression　data　in　adjacent　time　points.　And　then　we　get　gene　states　transfer　probability　by　training　HMM　using　gene　observation　matrix,　and　build　gene　regular　network　corresponding　it.　By　comparing　with　real　network,　our　experiment　provides　good　result,　and　the　method　has　less　computation　complexity　than　other　regular　methods　like　dynamic　Bayesian　networks.　©　2012　IEEE.