In recent years，data compression for Time Difference of Arrival(TDOA) estimation has become the hotspot in the signal processing field. Application field like emitter location using TDOA based on sensors network requires practical data compression technology，which can reduce the difficulty of platform design and implementation，and get a trade-off between the data amount for transmission and the accuracy of TDOA estimation. The compression rule based on maximizing Fisher Information is simplified to which based on maximizing the sum of the squared discrete Fourier transform magnitude，and the efficiency of data compression is improved，according to the feature that sub-band with lowest TDOA estimation Cramer-Rao Bound has the highest signal to noise ratio generally after the narrow-band communication signal being compressed by sub-band extraction. The sub-band extraction method is developed by combining with frequency extraction in order to reduce the effect of limited observing time. Simulation results show that，the proposed measures can obtain much higher accuracy of TDOA estimation than sun-band extraction method under the same compression ratio；tens nanosecond order of TDOA estimation accuracy can be achieved by transmitting and utilizing only 128 frequency points under signal to noise ratio above 0 dB.