A method for identifying non-stationary sound source characteristics using partial coherence technology belongs to the field of noise in physics. In the case where multiple sound sources generate complex cyclostationary sound fields, the present invention adopts cyclostationary theory instead of Fourier transform, proposes a partial coherence analysis technology suitable for cyclostationary sound fields, and uses it in the partial sound field separation of cyclostationary sound fields, the method As follows: first determine the number and location of the reference source, arrange the reference source microphone array, extract the reference source signal, then design the microphone array to scan and measure the holographic surface, collect the holographic surface data, and then use the cyclostationary partial coherence analysis technology to separate Part of the sound field for each sound source. The present invention can analyze the complex cyclostationary sound field formed by multiple cyclostationary sound sources, obtain the partial sound field formed by each sound source, and also calculate the three-dimensional sound field formed by a single sound source from the sound pressure signal measured on the holographic plane. The sound field distribution realizes the visualization of the propagation path of each sound source.