Modulation Recognition Method Of PSK Signals Based On Compressed Sensing

In recent decades, digital signal processing technology has gotten rapidly development. However, as the frequency band of the modulation signals becoming more and more border, the task of dealing with the wideband signals is much more difficult. Using traditional Nyquist sampling law for processing needs ever higher sampling rates. And it is very stringent requirements for physical device. Not only would it increase the complexity and the cost of realization, but also it would be beyond the performance limit of existing analog-to-digital converters(ADC). At this time, the birth of a new signal processing modes—compressed sensing theory brings hope to breaking the limitation of bandwidth for the sampling rate. This thesis infers a received signal’s modulation type by exploiting the differences of MPSK(Multiple Phase Shift Keying) signal features. We complete the recognition task of the MPSK signal’s modulation type in the case of low speed sampling rate by combing the compressed sensing theory with pattern recognition methods. The main research content of this article is as follows:1. This thesis introduces the basis of compressed sensing theory, describes the basic framework of compressed sensing, presents the traditional identification methods of the PSK signal, and find the point of those methods and compressed sensing. Then all of these lay the groundwork for introducing compressed sensing theory into the PSK signal modulation pattern recognition task.2. A new PSK signal modulation recognition method is proposed based on the high power of the Fourier transform. Due to greater than Nyquist rate sampling is essential for extraction in the Nth Power Nonlinear Transformation(NPT), we reduce the number of measurements, decrease the requirement for the front-end device and complete the recognition task by introducing the compressive sensing(CS) and combining CS with NPT. In the process of reconstruction in compressive sensing, we compare the greedy algorithm represented by orthogonal matching pursuit(OMP) with convex relaxation algorithm represented by basis pursuit(BP), then do experiments and simulations.3. This thesis proposes a PSK signal modulation recognition method based on compressive signal processing(CSP). Due to requiring amounts of calculation in the process of reconstruction in compressive sensing, we use CSP to solve directly the recognition task in the compressive measurement domain. The compressive measurements in fact already contains all the useful information of the original signal, using the compressive measurements to reconstruct a full-scale signal is a waste of resources, it only reduces the amount of data transmission process and the actual amount of data needed for processing has not been reduced. The feature of compressive measurements provides an opportunity to directly deal with it. In the back end decision, we respectively use the BP( Back Propagation) neural network algorithm(classification algorithm) and improved k-means algorithm based on simulated annealing thought(clustering algorithm), and improve the using range of the algorithm.