Adaptive OPAST algorithm can ensure the orthogonality of weighted matrix in every iterations, and accelerate the convergence rate, improve the performance of error stably. However, it also enhance the complexity of algorithm. In this paper, an improved OPAST algorithm is proposed. The algorithm reduce cumulative error in the process of every iterations by inducting the dual Gram-Schmidt orthogonalization, ensure the algorithm has a better stability. Simulation results show this algorithm has stability convergence.