Research　on　various　eco-environmental　sounds　is　very　important　for　people　to　understand　a　particular　area.　However,　eco-environmental　sounds　have　many　specific　properties　such　as　the　diversity,　high　background　noise　and　non-stationary　structure　which　make　many　traditional　audio　features　hard　to　characterize　them　accurately.　In　this　paper,　a　novel　feature　extraction　technique　based　on　Matching　Pursuit　algorithm　is　adopted　to　obtain　proper　features　which　can　describe　the　specific　properties　of　eco-environment　sounds　accurately.　In　order　to　further　improve　the　performance　of　our　classification　system,　we　make　use　of　the　support　vector　machine　to　automatically　classify　the　eco-environmental　sounds　into　seven　classes.　The　experimental　results　show　that　our　system　is　effective　to　classify　the　environmental　sounds　quickly　and　accurately.　©2010　IEEE.