Research On Signal Feature Parameters Estimation And Modulation Recognition Of Digital Signal

With the rapidly development of the modern communication technology, the users are increasing in the limited bandwidth, which causes a complex electromagnetic environment and the lack of the resources of the frequency bandwidth. Therefore, the characteristic parameter extraction of unknown communication signals is becoming more and more important. The modulation mode, symbol rate, carrier frequency and bandwidth of the digital modulation signal are key to the correct demodulation and the base of signal interference and decryption. They provide the structural characteristics of unknown signals, and have important applications both in military and civilian fields.Firstly, the four categories of used digital modulation signals are analyzed from several aspect, such as modulation mode, symbol rate, center frequency and bandwidth. And the popular algorithms of parameter estimation and modulation recognition are introduced. The algorithms in some typical applications are analyzed in detail, and their tendency are also of briefly introduced.Secondly, Corresponding to three important parameters, the three estimation algorithm, the symbol rate estimation algorithm based on wavelet transform, the bandwidth estimation algorithm based on average periodic diagrams and the carrier frequency estimation algorithm based on higher-order cyclic cumulant, were presented iis the thesis. The wavelet transform, so called "mathematical microscope", obtains the information in time domain and frequency domain by shifting the mother wavelet and modifying the scaling factor, respectively. The algorithms based on cyclic cumulants have better performance in Gaussian noise. This feature results in a high accuracy in carrier frequency estimation when the step-size is small enough. The algorithms based on the average period gram performs well in bandwidth estimation due to averaging the period gram of the segmented received signal. In this thesis, the algorithms and its corresponding theories are studied and the simulations are given respectively.Finally, we proposed two different algorithm in this paper for the modulation identification to determine the modulation mode under blind, considering the scalability and the rapidity. The former takes high-order cumulant as the feature parameter and extracted it in the same method, and chooses the support vector machine as the classifier. It has a complete classification architecture and good expansibility. It uses a similar approach to obtain the characteristic of the various of the signals. However, it costs more computational complexity compared with other algorithms. The latter algorithm analyzed the spectrum, square profile and other characteristics of the high power of the spectrum of the signal to extract their identifying characteristics of waveform characteristics, then based on the tree classification and a large number of tests determined the optimal threshold for recognition. The algorithm has a relatively low complexity, and a better recognition in real time. But it achieved a considerable cost in expansibility.