【Objective】In order to improve the reconstruction accuracy of the particle velocity and ensure the ac curacy of source identification, 【Method】this paper proposes a compressed equivalent source based-NAH mod el with the sparse acoustic vector sensor array CESM-v by substituting the particle velocity measurement for the traditional pressure measurement and combining the acoustic vector sensor with the sparse sampling and sparse regularization. Several numerical simulations and experiments were carried out to exam the efficiency of the pro posed model. Through multiple analysis of the particle velocity reconstruction, the equivalent source strength re construction and the error calculation, the proposed CESM-v model was compared with the traditional equivalent source based-NAH with particle velocity measurement (ESM-v) and the existed CESM-p model. 【Results】The proposed CESM-v model always performs better than the other two models. Even though using sparse sam pling, the CESM-v model can generate comparable effect of the particle velocity reconstruction to that of tradi tional ESM with adequate sampling, and can give higher and more stable accuracy in particle velocity recon struction than that of the CESM-p model. In short, both simulation and experimental results show that the CESM-v model has better stability and reliability, can obtain good particle velocity reconstruction results with a small number of sampling points, and then the accuracy of sound source identification can be ensured. 【Conclusion】With the robustness of sound field reconstruction, the CESM-v model can be extended to practical projects such as rail transportation for noise source identification and fault diagnosis.