压缩感知重构极大熵方法_英文_.pdfVIP

应 　用 　数 　学 MA T H EMA TICA A PPL ICA TA 20 10 ,23 (2) :345352 Maximum Entropy Function Method f or Signal Reconstruction in Compressed Sensing WAN G Tianj ing (王天荆) 1 , YAN G Zhen (杨震) 2 ,WU Yej un (吴业军) 3 ( 1. Col leg e of S cience , N anj i ng Uni ve rs i ty of Tech nol og y , N anj i ng 2 10009 , Ch i na; 2 . I ns t i t ute of S i g n al an d I nf orm at ion P rocess i ng , N anj i ng Uni v ers i ty of P os ts Telecom m u ni cat ions , N anj i ng 2 10003 , Ch i n a; 3. D ep a rtm ent of B as ic Cou rse , N anj i ng I ns t i t ute of Tech ) nol og y , N anj i ng 2 11167 , Ch i n a Abstract :The emer ging t heory of Comp ressed Sen sing (CS) ha s led to t he remarkable result s t hat comp ressible signal can be recon st ruct ed u sing only a small number of measurement s. Significant at t ention in CS ha s been focu sed on Ba si s Pur suit (B P) ,exchanging t he sp ar seness con st raint wit h t he l1 norm . In or der to overcome t he non smoot h p roblem in l1 norm ,t hi s p ap er p ropo ses a new Maxi mum Ent rop y Function Met ho d (M EFM) to solve t he l1 op timization p roblem via smoot hing t he ob j ective function wit h maximum ent rop y function . Intimat ely relating to homotop y met ho d , M EFM p rovides a syst ematic app roach for deriving t he global op timal sp ar se solution . Finally ,t he numerical result s show t hat it i s an effective t echnique for signal recon st ruction . In a CS f ramework ,M EFM i s a u sefully alt ernating met hod to solve t he l1 op timization p roblem . Key words :Co mp re ssed sen sing ;Non smoot h op timization ; Ho motop y met ho d ;Maxi mum ent rop y f unction met ho d CLC Number :TN 9 12 　　AMS( 2