A　new　method　for　extracting　feature　of　the　dynamic　error　of　machine　tool　from　the　flatness　surface　is　proposed.　The　dynamic　error　is　identified　based　on　the　character　analysis　of　the　test　result.　By　Daubechies　wavelet,　the　information　that　contains　high　frequency　signal　and　low　frequency　signal　is　decomposed　from　the　flatness　error　of　workpiece　surface.　At　the　same　time,　the　wavelet　transform　and　power　spectral　density　analysis　are　combined,　and　the　dynamic　errors　of　machine　tool　from　the　measured　flatness　error　of　workpiece　are　extracted　and　identified.