Abstract: An ANFIS and wavelet denoising algorithm is proposed for additive colored noise. Modeling the nonlinear dynamic characteristic of noise by ANFIS and fuzzy approximating colored noise, the Auto Neural Fuzzy Inference System (ANFIS) and wavelet denoising system of speech signal is set up and trained. According to the different wavelet coefficients' transmission properties of edge signals and noises under the different scales of the wavelet transform, the median filter is designed. Colored noise can be successfully removed by using subtraction in the original speech signal. Experimental results show that the algorithm is effective.