Abstract: A novel data-level fusion method for emitter identification is proposed for the large scale communication in cooperative reconnaissance, which uses the superiority of compressed sensing in representing an original signal by using few measured data. In the proposed method, the Gabor time-frequency data of intercepted signal in a receiver is compressively measured with a Gaussian random measurement matrix. By transmitting few compressively measured data rather than the original signal, the large scale communication problem is alleviated. In the fusion center, a correlation fusion rule is proposed to calculate the combined weight of measured data according to the correlation among compressively measured data. To identify the signal type, a dictionary library is trained with every possible signal, and the reconstruction error in the sub-dictionary is calculated. The signal type with minimum reconstruction error is just the identification result. The simulated result proves that the proposed method achieves a pretty good balance in identification rate and communication scale, especially under low signal-to-noise ratio. Compared to the existed algorithms, the correlation fusion rule keeps more details of original signal. Key

Key words: ordnancescienceandtechnology, emitteridentification, data-levelfusion, compressedsensing, time-frequencytransform, correlationfusionrule