The new video compression method recognition- oriented is discussed using wavelet analysis on the basis of the characteristics of wireless automatic surveillance network. The characteristics of surveillance network are narrowband for video communication and need lower cost in the terminals because of the quantity. A new method is proposed in this paper for narrowband network, such as wireless network. The innovation of the method is that low- frequency parameters of wavelet transform that are used in classical compression method are not used, while the high-frequency parameters are used for the purpose of target recognition. This is done by emphasizing high frequency components and discarding low frequency components according to a new developed recognition-based approach for image compression. Computer simulation experiments show that high-frequency parameters can lower the bit rate, having good compression ratio, and has higher practical value for target recognition in surveillance system.