关于几种变步长LMS算法的讨论

16720085ShipElectronicEngineeringVol.28No.5175LMS3(430033)LMS,LMS,,LMS,;LMSTP301.6DiscussiononVariableStepLMSAlgorithmsLiCunwuLinChunsheng(NavalUniversityofEngineering,Wuhan430033)AbstractLMSalgorithmisoneofveryimportantalgorithmappliedinself-adaptativefilter.BasedonthetraditionalLMSalgorithm,severalkindsofimprovedalgorithmshavebeenproposedinrecentyears.Inthispaper,theimprovedalgorithmsareintroducedanddescribed,andtheadvantagesanddisadvantagesofthosealgorithmsintheapplicationsareanalyzedanddis2cussed.KeywordsVariableStep,LMSAlgorithmsClassNumberTP301.61LMS,,,,LMSLMS,nX(n)=[x(n),x(n-1),,x(n-M+1)]H,M,W(n)=[w(n),w(n-1),,w(n-M+1)]H,d(n),:e(n)=d(n)-WH(n)X(n)LMSW(n+1)=w(n)+2e(n)X(n),,02max,max,,,,,LMS,,LMS,,LMS:,,,;,,,3:2007718,:2007830:,,,,,,:2LMS1)[2]LMS:e(n)=d(n)-XT(n)W(n)(n)=[1-exp(-|J(n)|)]W(n+1)=W(n)+(n)e(n)X(n),T;|J(n)|J(n);X(n)=[x(n),x(n-1),,x(n-N+1)]T;N;W(n)=[w0(n),w1(n),,wN-1(n)]T;e(n);d(n);J(n)e(n),e(n)X(n),e2(n),(n);,LMS0(n)23tr[R],R=E{X(n)XT(n)},tr[R]R,,023tr[R]LMS,,;,,J(n)(),(n)J(n),,,2)NVLMS,:w(n+1)=w(n)+1+e(n)2e(n)x(n),e(n)2=n-1k=0e2(n-k),e(n)LMS,(n)=1-e(n)2,,,(n)e(n),n,,e(n)2,(n),,e(n)0,(n),,e(n),(n),(n),3)LMS(VSS),(n),:(n+1)=max,(n+1)maxmin,(n+1)min(n+1),else(n+1)=a(n)+e2(n),01;(0)maxLMS,l,,LMS:,e(n),(n),;,e(n),(n),VSSLMS,,,,e(n)e(n-1),:(n+1)=(n)+p2(n),p(n)=p(n-1)+(1-)e(n)e(n-1),(01);(0)=max,p(0)=(0),,MVSS,VSS,,,e(n),e(n)e(n-1)3,,,LMSLMS,u,,,,671:LMS167[1].[M].:,2001[2],.LMS[J].,2001,29(8):10941097[3],.LMS[J].,2005,27(7):13071310[4],.LMS[J].(),2006,34(4):6164[5],.LMS[J].,2006,(11):4850(171)i()=rcos[-(i-0.5)]c,i=1,2,,M(2)c,-i=-i()--i(0),-i()H-iO,:-i()=rcos[+(i-0.5)]ci=1,2,,M(3)wiw-i424,f=1.05KHz[2]:19,4,-8.1,15%324,3,;,dB;,1719,24,,-5.9-8.1,27%15%,,,,,,(0,0),,01,524,;;,,[1],,.[J].,2007,26(1):6568[2].Matlab[J].,2004,21(4):25627[3],,.[J].,2006,25(5):270272[4],.[J].,2004,23(1):6771[5].[M].:,2003:195205[6].[M].:,2003:41341577120085