Aiming at the problems that the influence of bridge deck roughness on bridge vibration is hard to remove and the low accuracy of high-order modal recognition during the current indirect measurement method based on vehicle response， this paper proposes an approach to separate the effect of bridge deck roughness and bridge vibration response by using blind source separation identification based on the acceleration response signal of the axle contact point， to identify the bridge modality. Firstly， the principle and method of obtaining the bridge vibration estimation signals by using the second order blind identification （SOBI） algorithm and taking the acceleration response signals of two groups of vehicle-bridge contact points as input signals are described in detail. Then， the bridge modal identification technology flow and framework based on vehicle response are established using signal band-pass filtering and Hilbert transform combined with the mode modification strategy of fulcrum data extension. Finally， the applicability and effectiveness of the proposed method were verified by numerical examples， and the influences of vehicle spacing， bridge deck roughness， and vehicle frequency on the applicability of the method were analyzed. The results show that the proposed method can effectively filter the influence of bridge deck roughness， achieve accurate identification of high-order modes of bridges， and show robustness under the influence of multiple key factors. Moreover， the proposed method has the characteristics of high precision， simple operation， and good applicability， which can provide a new idea for bridge modal identification based on vehicle response.