To　solve　the　data　collection　problem　in　wireless　sensor　network,　an　efficient　and　energy-saving　data　collection　scheme　based　on　compressive　sampling　theory　was　proposed.　The　matrix　projection　method　was　adopted　to　compress　the　sensors＇　sensed　data　in　the　data　collection　phase.　Meanwhile,　in　order　to　convert　the　constrained　l0　norm　minimization　problem　into　an　unconstrained　optimization　problem,　a　family　of　exponential　functions　was　utilized　to　approximate　the　l0　norm　of　the　original　signal　in　the　data　recovery　phase.　Furthermore,　a　series　of　weighting　functions　were　also　designed　to　accelerate　the　convergence　speed　of　the　recovery　algorithm.　The　experiment　results　have　shown　that　the　proposed　scheme　is　efficient　and　low　cost　in　terms　of　bandwidth　and　energy　in　the　data　collection　phase　and　also　provides　a　higher　recovery　rate　than　the　existed　recovery　schemes　within　an　appropriate　reconstruction　time　in　the　data　recovery　phase.　©,　2015,　Huazhong　University　of　Science　and　Technology.　All　right　reserved.