Modulation Recognition Method Based On CPU Software Radio

The modulation recognition is aimed to decide the modulation type and estimate the modulation parameters without any prior knowledge on signals. With the rapid development of communication technology, the modulation schemes of communication signals become more and more complicated and various, so it requires more and more to modulation recognition technology.Based on the principle of using simpler feature parameters to recognize more signals, recognition is realized in this thesis. The main contents are as follows:1. A modulation recognition method based on the combination of intermediate frequency feature parameters and higher-order cumulants is proposed. The scheme combines three intermediate frequency feature parameters, three feature parameters based on higher-order cumulants with SVM classifiers based on decision-making tree are used to recognize 11 kinds of digital modulation signals, and achieves more than 99.1% recognition accuracy when SNR is above 10dB. The proposed scheme has an advantage over the arithmetic only based on higher-order cumulants.2. Another modulation recognition method based on spectrum feature is proposed. The scheme puts forward a set of simple and effectual feature parameters which have a thick skin to noise and the parameters of signal such as information speed, sampling frequency, phase jitter and so on. It also uses SVM classifiers based on decision-making tree to recognize 10 kinds of digital modulation signals and achieves more than 98.8% recognition accuracy when SNR is above 5dB.3. It is presented that a scheme of software radio system based on CPU to realize the theory of the modulation recognition based on spectrum feature. Testing result of the software is presented, which shows that the recognition accuracy matches the simulation well, and the average time of recognition of every signal is about 10ms in the predefined condition.