In order to further solve the contradiction between the convergence speed and the steadystate error of the least mean ppower algorithm (LMP) and improve the performance of the adaptive algorithm, a convex combination scheme with different pvalues and variablestep size was proposed（combination of least mean p power, CLMP）to effectively solve this problem. In this scheme, independent large step filter LMP₁ and small step filter LMP₂(p₁>p₂) are connected in parallel in Gaussian environment, and the theoretical derivation of steadystate error expression was carried out by using separation assumption. Simulation results show that the steadystate performance of combined filter is better than that of single filter in Gaussian stationary environment, which provides a new idea for the research of adaptive filtering algorithm.