©　2004-2012　IEEE.　This　brief　proposes　a　bias-compensated　proportionate　normalized　least　mean　square　(BCPNLMS)　method　for　identifying　sparse　system　when　subjected　to　the　noisy　input.　The　proposed　BCPNLMS　algorithm,　which　combines　the　proportionate　scheme　and　the　unbiasedness　criterion,　is　able　to　identify　the　system　parameters　with　better　steady-state　accuracy　and　faster　convergence　speed　than　conventional　NLMS,　bias-compensated　NLMS,　and　PNLMS　algorithms.　Robustness　and　high　identification　accuracy　with　noisy　input　can　be　achieved　by　introducing　the　bias-compensation　term　derived　from　the　unbiasedness　criterion.　Simulation　results　on　sparse　system　identification　confirm　the　excellent　performance　of　the　proposed　BCPNLMS　in　the　presence　of　both　input　and　output　noises.