As　it　is　difficult　to　get　an　accurate　mathematical　noise　model　under　various　dynamic　interference　conditions,　Strap-down　Inertial　Navigation　System　(SINS)　was　difficult　to　realize　self-alignment.　Although　the　fuzzy　adaptive　Kalman　filter　can　be　used　to　realize　the　SINS　self-alignment,　the　algorithm　complexity　is　high　and　the　error　term　is　unstable.　To　solve　this　problem,　a　novel　self-alignment　method　using　real-time　adaptive　filter　was　developed　in　this　paper.　This　method　is　based　on　the　theory　of　fuzzy　adaptive　adjustment　and　used　the　filter　stability　as　the　criterion.　The　exponential　function　was　used　to　replace　the　fuzzy　inference　calculation,　which　eliminated　the　complex　process　of　fuzzy,　fuzzy　inference　and　de-fuzzy.　The　stability　and　accuracy　of　the　adaptive　filter　was　also　improved.　Finally,　the　simulation　process　was　used　to　test　the　algorithm　of　this　method,　and　the　results　demonstrated　that　the　self-alignment　using　real-time　adaptive　filter　method　reduced　the　complexity　of　the　algorithm　and　had　better　stability　and　alignment　accuracy.　©　2016　IEEE.