Split feasibility problem where the sets C and Q are nonconvex is considered in this paper. This class of problems are widely applied in compressed sensing, matrix factorization, outlier detection and many other fields. In this setting, split feasibility problem can be formulated as a non-convex and non-smooth problem with variable separation, so that alternating minimization algorithm can be used. We focus on the alternating minimization algorithm with Bregman distance studied in Chao et al. [7]. To further enhance efficiency, Armijo-type search method and linearization technique are employed. Under suitable assumptions on the sets involved, we show that, the sequence generated by our algorithm is bounded and globally convergent when applied to the nonconvex split feasibility problem. Some mild conditions that ensure the sequence converges to the solution of problem are proposed in this paper.