Modulation Classification For BPSK And QPSK Signals Based On SVM

Modulation classfication technology of communication signals is an important branch in the signal analysis and processing filed. It is applied extensively in many techonolgies such as reconnaissance, monitoring and the best interference pattern synthesis technology, and becomes a significant research theme in the communication electronic warfare realms.With the rapid development of communication technology, especially digital communications technology, it has created a variety of communication systems coexisted. The modulation and access technologies in these communication systems are different, which becomes a great obstacle in the multi-system internet communication. Wireless communication environment becomes increasingly complex, and modulation type becomes more diverse, the received signal is often the weak target signal which is mixed with noise, such as white Gaussian noise, impulsive noise, etc. Therefore, it is necessary to improve the modulated recognition performance at low SNR environment.In this paper, the development of modulation classification is firstly introduced. The recognition method based on decision theory and the pattern recognition method based on feature extraction are analyzed systematically; The modulation classification based on SVM is developed. The problem of modulation classification of BPSK and QPSK based on HOC and SVM at the low SNR is studied in detail. To improve the recognition accuracy of signal modulation at low SNR, combined antenna array with SVM, a modulation recognition algorithm of BPSK and QPSK based on SVM is proposed. In this algorithm, to increase the SNR of the received signals, the equal gain combination is exploited to the received signals of the antenna array. Signal feature extractions based on the normalized fourth-order and sixth-order cumulants are then applied to the combined signals. Taking advantage of these feature information, a SVM classifier can be constructed and the modulation recognition scheme of the received signals can be given. Furthermore, the influence of the length of the observed data and the number of the sources to the probability the accurate identification of BPSK and QPSK is analyzed.Through analyses and simulations, the modified digital modulation recognition algorithm achieves the request of task, which has an important significance of theory and practice to further increase the performance of digital modulation recognition in low signal to nosie circumstance.