Design And Implementation Of RF Modulated Signal Recognition Using Parallel Algorithm

Automatic modulation recognition is the central procedure in signal electronic reconnaissance and signal detection. It has been increasingly important and widely used for both civilian and military applications. And along with the development of satellite communication, the study on automatic modulation recognition commonly used in satellite channel becomes more and more indispensable. In this paper, several available automatic modulation recognition algorithms are analyzed, and a parallel algorithm based on a two-step 1-D FFT parallel algorithm and a recognition algorithm is designed. The complex algorithm is implemented on a multi-DSP signal processing system.Firstly, after analyzing the different algorithms which have been developed for automatic modulation recognition, this paper adopts the statistical modulation recognition method to accomplish the classification of the modulated signals in consideration of the communication environment of satellite channel. According to the characters of the interest signals, the orthogonal transform based on digital mixing frequency is proposed to extract the typical instantaneous features of the modulated signals. The paper analyzes the characters of statistical features of the modulated signals, and then, this paper proposes a kind of algorithm based on signal envelop spectrum and power spectrum. After that, the instantaneous feature and the spectrum feature are multiplied with decision tree classifier to recognize signals. The MATLAB simulation result shows that high classification rates are achieved with this method.In order to exert the hardware's performance, this paper studies two typical FFT parallel algorithms, and a standard six-step FFT parallel algorithm and a two-step FFT parallel algorithm are analyzed. Because of better expandability and flexible output control, the two-step FFT parallel algorithm is chosen as parallel computing fundamentals to complicated signal processing based on FFT.Finally, this paper implements the signal recognition parallel algorithm on a multi-DSP system. Two parallel schemes are presented based on task dividing and data dividing, respectively. The analysis indicates that, the algorithm based on task dividing has a lower time complexity, whereas the algorithm based on data dividing has a better expandability. The software development task is accomplishment by coding, debugging and classification threshold selection.