Abstract: The conflict measure of evidences is the key issue for the evidence theory.A new method for conflict measure based on singular value decomposition(SVD) of matrix was proposed in this paper.The new method firstly normalize the BPA matrix of evidences by a projection operation.Then the SVD algorithm was used on the normalized BPA matrix to capture the largest and the second largest singular values.The largest singular value meant the main combining direction and second largest one meant the strongest direction of conflict.So the ratio of the largest singular value and the second largest one was used to measure the conflict of evidences.Experiments of the Zadeh paradox,completely conflict evidences set,and nesting sequence of focal elements showed that the new method could measure the conflict of evidences with a good performance.