For the separation problem of bridge deflection monitoring, it presents a single channel blind source separation algorithm based on EEMD-JADE. First, the single channel signal of bridge deflection is decomposed into a series of linear and stationary intrinsic mode function (intrinsic mode function, IMF) by traditional ensemble empirical mode decomposition (Ensemble Empirical Mode Decomposition, EEMD), and then using the discriminant method based on the energy entropy increment to identify and eliminate the false IMF component. The IMF component with larger energy entropy increment compose the input signal of the blind source separation model. Finally, the Joint Approximate Diagonalization of Eigen-matrices (Joint Approximate Diagonalization of Eigen-matrices, JADE) algorithm is used for blind source separation of the input signal. JADE can also separate the source signal well under the condition that the frequency difference of the source signal is small and the frequency is mixed, but the number of observed signals must be greater than or equal to the number of source signals. The Ensemble Empirical Mode Decomposition (EEMD) has good adaptability, which can decompose the mixed signals of single channel into multi-scale and form multi-channel signal, but the decomposition result has the endpoint effect and the modal aliasing. JADE algorithm can solve the end-point effect and modal aliasing problem in the decomposition result of EEMD, while EEMD also solves the prerequisite of JADE separation algorithm. The two algorithms have complementary advantages and can better separate the deflection components. With the model of background bridge established by the finite element software Midas/civil, the response of the bridge structure under the action of each single factor is obtained and it is superimposed together as a mixed deflection signal to be separated. The correlation coefficient between the result of the simulation signal separation and the source signal is above 0.98, and the separation effect is better. Finally, the measured deflection signals are collected for separation. The correlation coefficients of the deflection components separated at the symmetric position are above 0.9, which proves the applicability of the algorithm.