The invention discloses a three-dimensional sound source localization method based on compressed sensing, so as to solve the technical problem of poor anti-noise performance of the existing three-dimensional sound source localization method. In the technical scheme, the measurement value of sound source signals is acquired through a microphone array, a selected three-dimensional sound source area is subjected to uniform mesh generation, and each mesh node is used as a hidden sound source position; according to a Helmholtz equation of a free field Green's function, a measurement matrix between the mesh node and the microphone array is built, and a three-dimensional narrowband sound source localization sparse representation model between the measurement value of the microphone array and unknown sound source signals is acquired; through carrying out singular value decomposition on the measurement value of the microphone array in the sparse representation model, a sound source localization sparse representation model after deformation is acquired; and finally, a compressed sensing OMP (orthogonal matching pursuit) algorithm is adopted to carry out iterative solution on the representation model after deformation, the sound source strength of each mesh node in the sound source area is acquired, and the sound source is localized. The anti-noise performance of sound source localization is improved.