Research On Modulation Recognition For Wireless Communication Signals

Automatic modulation recognition of communication signals, which is a hot topicin many fields, such as signal monitoring, interference identification, multi-userdetection, electronic countermeasures, radio spectrum monitoring, and so on, is atechnology to determine the modulation type of the received signal after appropriatelyprocessing without priori information when transmission noise exists, and to pave theway for further work. Especially, it is an extremely important and necessary function forsoftware radio receiver. Automatic modulation recognition technology plays a vital rolein modern wireless communication and has a broad engineering application prospects.With the rapid development of communication technology, wirelesscommunication environment has become increasingly complex and the signals receivedby the antenna are always mixed signals with multiple samples and aliasing in time,frequency and space simultaneously. The separation and recognition of the signals arevery difficult with less priori knowledge, so how to separate and recognize these mixedsignals quickly, accurately, safely and reliably have been a hot topic with much attentionin military and civilian domains.In this paper, some modulation types of digital communication signals areintroduced and simulated by MATLAB firstly. Then, a method based on independentcomponent analysis (ICA) and support vector machine (SVM) is introduced to separateand recognize the mixed signal which consists of seven common digital signals inwireless communication, namely2ASK,4ASK,2FSK,4FSK,2PSK,4PSK and16QAM. The FastICA algorithm is used to separate each independent component.Simulation result by MATLAB shows that each independent component is restoredideally from the source signal. By introducing SVM classification thought, a SVMmodulation recognizer is used to recognize the modulation types by extracting sixcharacteristic parameters from signals’ instantaneous information (instantaneousamplitude, instantaneous frequency, instantaneous phase), spectrum and complexity asrecognition feature sets. Since kernel function type, parameters and the penalty factorC directly affect the generalization performance of the SVM classifier, extensiveparticle swarm artificial bee colony (EPSABC) algorithm is adopted in this paper and itsperformance is evaluated by MATLAB. EPSABC which has more better performance is applied successfully to optimize SVM parameters. Finally， digital modulationrecognition is achieved based on EPSABC-SVM and each signal is recognizeddefinitely. In conclusion, the experimental results show good performance in theseparation and recognition of wireless signals with the algorithm.