A high-resolution method for sound sources localization and identification based on LP norm eigenvalue decomposition is proposed in this paper. According to the principle of signal subspace method, the response function of subspace is obtained by making eigen decomposition of cross spectral matrix. The signal reconstruction model in each feature subspace is established via pre-defined sound source types for the reference solutions, and then by utilizing the sparse constraint condition of LP norm to calculate the optimal solution, the high-resolution localization and identification of sound sources can be achieved. By the theory derivation and numerical simulation, the method proposed not only can obtain the localization results but also can reflect the absolute contribution of each coherent source comparing with several existing beamforming algorithms. The parameters affecting the performance of sources identification are also reasonably chosen by numerical simulations. The method is capable of locating and identifying sources with high-precision and high-resolution which has good prospect in engineering applications.