A　new　method　for　extracting　spectrum　feature　of　spindle　unbalance　of　machine　tool　is　proposed.　The　flatness　error　of　workpiece　surface　includes　much　errors　information,　and　the　information　contains　high　frequency　signal　and　low　frequency　signal.　For　these　errors　information,　a　new　identification　method　of　turning　errors　of　workpiece　based　on　the　wavelet　transform　and　power　spectral　density　analysis　is　proposed.　According　to　the　focal　variation　character　of　wavelet　and　the　energy　value　of　power　spectral　density　analysis,　the　feature　of　spindle　unbalance　from　the　measured　flatness　error　of　workpiece　is　extracted　and　identified.