For　solving　the　problem　on　feature　extraction　of　test　signal　in　high-dimensional　space,　the　compression　sampling　and　reconstruction　methond　based　on　sparse　signal　is　presented　here.　Test　signal　is　denoteed　with　sparse　representation　in　specific　transform　domain,　the　observation　matrix　which　is　unrelated　to　transform-based　is　constructed　for　compressed　sampling,　Data　dimension　reduction　and　signal　feature　extraction　are　completed　simultaneously,　and　these　discrete　data　can　reconstruct　the　test　signal　in　compression　domain　by　norm　optimization　technology　perfectly.　An　illustration　verify　this　method　finally.　©　2010　IEEE.