Blind Identification of Signals with Stochastic Frequency Offset in Non-Cooperative Environment

With no knowledge of the transmitted data and multiple unknown parameters of the receiver, blind identification is vulnerable to frequency offset. In this paper, we present a frequency offset estimation algorithm based on the kernel method and an unsupervised method termed as density and metric based spectral clustering analysis (DMSC), so as to identify the modulation type of signals. The stochastic frequency offset is estimated by maximum likelihood based kernel method. Then a new spectral clustering method is therefore proposing to identify the modulation type of signals. We send 8QAM signals of which the frequency offset obeys Rayleigh, normal and gamma distribution. The classification error rate of our method is lower than k-means and traditional spectral clustering method.