In order to reduce the number of sensors arranged in current bridge detection, the wavelet packet decomposition and sample entropy are combined. A novel method detecting and identifying the structural damages of continuous girder bridge by dynamic response signal of single point has been discussed and studied. A novel damage identification index and method of continuous girder bridge has been proposed. In applications, the wavelet packet transform is used to decompose and reconstruct the acceleration response signal of bridge under moving load. The sample entropy values of reconstructed signal have been calculated. Moreover, the sample entropy of wavelet packet transform with different logarithmic acceleration energy has been established to act as a damage identification index. Its applicability and correctness are verified by a three-span continuous girder bridge with variable cross sections. Noise robustness of index is alsoanalyzed. The research results show that the index and method proposed in this paper can identify the damage location and extent by analyzing the acceleration response signals of any points on bridge, which is not sensitive to the noise.