Abstract: An improved method based on orthogonal wavelet transform for adaptive speech denoising is proposed,which can increase the convergence rate of the adaptive speech denoising process.Orthogonal wavelet transform plays a part in whiting,and enables the adaptive filter input signals to be orthogonal.With orthogonal wavelet analysis,the energy in the adaptive filter of every wavelet subspace is lowered,the dynamic domain of the input auto-correlated matrix of the adaptive filter is reduced,and the distribution of eigenvalues is more concentrated,so the convergence rate is increased.