Automatic Digital Modulation Recognition Based On Software Defined Radio

The basic idea of Software Defined Radio is to constitute an opening, standardization and modularization general hardware platform, and people can implement different wireless communication functions by loading different software. In order to implement the multimode communication system of Software Defined Radio, it's necessary to automatically recognize the modulation: signals corrupted by the channel noises. We study the important technology of Software Defined Radio deeply in this paper.Firstly we explain the concept of Software Defined Radio, the system structure and interrelated theories, then aiming at the constitution of modulation recognition system, we focus on the feature extraction and classifier design. Two types of features are studied in this paper: the instantaneous features and the multi-resolution features based on wavelet transform; three classifiers are studied: parallel decision classifier, artificial neural network classifier and support vector machine classifier. All of these are improved and applied in the modulation recognition system and a high recognition rate is acquired. After experimenting and emulating with MATLAB and LIBSVM software platform we can get the conclusion as follows:1. The instantaneous features are easy to extract, but are susceptible to the noises; the multi-resolution features are not sensitive to the change of the noises and even eliminate them.2. The parallel decision classifier has a simple structure and suit on-line processing; the artificial neural network classifier avoids configuring the threshold and has self-adjusting ability; the support vector machine classifier not only avoids configuring the threshold but also has good generalization ability, and has good performance even with less samples.3. All the modulation recognition models have good performance.