Channel estimation usually refers to estimation of the frequency of the path between the transmitter and receiver. Several channel estimation algorithms have been developed, such that LS, MMSE estimation with decision feedback and channel estimation by neural network. However in a practical environment where Gaussian impulse noise can be present, the classical methods may not be effective for this impulse noise. The LS-SVM estimation algorithm makes use of reference signals to estimate the total frequency response of the highly selective multipath channel . The simulations will show the effectiveness the proposed method which has good performance and high precision.