In　this　paper,　a　novel　supervised　learning　algorithm　for　automatic　classification　of　individual　musical　instrument　sounds　is　addressed　deriving　from　the　idea　of　supervised　non-negative　matrix　factorization　(NMF)　algorithm.　In　our　approach,　the　orthogonal　basis　matrix　could　be　obtained　without　updating　the　matrix　iteratively,　which　supervised　NMF　algorithm　is　unable　to　do.　Afterwards,　each　data　is　projected　onto　several　training　orthogonal　basis　matrices　and　three　classifiers　have　been　employed　to　compare　the　performance　with　different　methods.　In　addition,　feature　selection　is　also　applied　in　order　to　choose　the　most　discriminative　features　for　instrument　classification.　The　results　indicate　that　the　classification　accuracy　of　proposed　method　is　87.6%,　which　is　comparable　to　the　performance　of　supervised　NMF　algorithm　for　the　same　experiments.