Aiming at the problem that noise residual and speech distortion can’t reach the minimum at the same time, this paper proposes an improved signal subspace combined with Wiener filter for two - level speech enhancement. The first stage uses the auditory masking parameter to improve the gain matrix of the signal subspace. After processing, it can effectively reduce the speech distortion and improve the signal to noise ratio. And then use the Wiener filter algorithm to enhance the enhanced speech signal. The experimental results show that compared with the traditional signal subspace algorithm and Wiener filter algorithm, this algorithm can effectively reduce the residual noise in the speech signal, reduce the distortion of the speech signal and improve the overall quality of the speech signal.