Testing the existence of a low-dimensional perturbations or signals is of fundamentalimportance in statistical theory and application like factor analysis and signal processing.The paper aims to develop a new test for high dimensional spiked covariancematrix basing on projection. The asymptotic distributions of the proposed test areobtained under some regular conditions. We further explore the power enhancementtechnique if we have auxiliary information that the covariance matrix is somewhatsparse. We assess the finite sample performance of the proposed test by examiningits size and power via Monte Carlo simulations, which show that the proposed testachieves much more power. A dataset of 97 stocks is studied using the proposed test.