Communication Signals Modulation Recognition Algorithm Research

With the development of technology, the amount of modulation type becomes larger and the modulation becomes more complicated. How to identify the signal modulation efficiently has become a challenging research direction and attracts considerable scholarly attention. Automatic identification of signal modulation refers to identify the modulation of signal which has been processed by channel existing noise and inference without prior information of signal or prior information is negligible.This paper adopts three methods to solve the problem of signal modulation in different SNR, these three methods are the combination of Short Time Fourier Transform(STFT) and Affinity Propagation(AP), Non-negative Matrix Factorization(NMF) and AP, as well as NMF and Learning Vector Quantization Neural Network(LVQNN) respectively. STFT is an algorithm which could show different time domain and frequency domain properties, thus we could analyze the signal based on the spectral configuration obtained by STFT. The AP has been applied in many areas such as image processing, biomedical sciences, text categorization, etc. In this investigation, we extract the features of signal by utilizing the AP which could reduce dimensions. The advantages of AP are fast calculation speed and good cluster effect, thus, we could identify and classify the features extracted based on Semi-supervised affinity propagation clustering(SAP). Owing to its remarkable capability of self learning, generalization and nonlinear data process, the neural network is also been studied for classification. Experimental simulations are provided to demonstrate the superior performance of the methods aforementioned.