Abstract: Due to the humanDs face, fingerprint, handwritten signature and other personnel identification method could not be well to meet the requirements of the mine underground personnel management system, based on a locality preservation projection (LPP) algorithm, a supervision algorithm on the locality preserved projecti on (SLPP) was provided and was applied to the gait recognition of the mine underground personnel.The gait data could be projected with the supervision on locality pre served projection.The expression method of the gait data in the low dimension was obtained.The nearest neighbor classifier was applied to the identification of the low dimension gait data.A test of the series gait identification was conducted in two gait databases and was compared with the classic dimension reduction algorithm LDA, t he supervision manifold learning algorithm DLPP, discriminant projection embed manifold learning algorithm and other gait identification method.The test results showe d that under the same test conditions, the supervision on locality preserved projection (SLPP) would have the highest recognition rate and thus the validity and feasibilit y of the algorithm could be verified.