Abstract: We consider the nonparametric regression model =g（X_i）+ε_i, where the residual ε_i is a stationary φ-mixed process with unknown common density function f（x）, g（x） = E（Y|X =x） is the unknown regression function. In this paper, first we define the residuals estimates based on the nonparametric estimator \hatg_n(x) forg（x）, then we construct the estimate \widehatf_n(x) for f（x）on the basis of the residuals, the pointwise weak and strong consistency as well as the uniform strong consistency of \widehatf_n(x) are estabilished under some suitable conditions.