Although　electro　larynx　speech　provides　an　important　means　for　the　laryngectomees　for　oral　communication,　the　resulting　speech　is　of　poor　intelligibility　due　to　the　radiated　noise　caused　by　the　instrument.　This　paper　concentrates　here　on　the　derivation　of　a　minimum　mean-square　error　spectral　amplitude　estimator,　and　on　its　application　in　electro　larynx　speech　enhancement,　also,　the　frequency　domain　masking　properties　of　the　human　auditory　system　was　also　embedded　into　the　proposed　speech　enhancement　algorithm,　so　that　to　further　improve　the　intelligibility　of　the　enhanced　electrolaryx　speech.　The　acoustic　analyzing　results　and　the　listening　evaluation　of　the　enhanced　electro　larynx　speech　suggests　that　the　radiated　noise　can　be　efficiently　reduced,　which　suggest　that　the　proposed　algorithm　has　a　good　performances　for　radiated　noise　reduction　for　electro　larynx　speech.　©　2012　IEEE.