Digital Modulation Signals Recognition In Shortwave Communications

The short-wave communication which is a kind of modern communication technology maintains vitality with the superiority of its mobile flexibility, the device simplicity and remote communication, although with the development over several decades. However, the short-wave communication had serious impact on the work of the receiver, because of its character of massive bad noises and many kinds of signals. The algorithm of digital modulation signal recognition in short-wave communication is studied in the thesis according to these above characters of short-wave communication, which will improve working conditions and enhance their working efficiency.Firstly, mathematical models of common digital communication signals are introduced, at the same time the method which is used to estimate the modulation signals parameters is given.The combination of wavelet packet and cumulants is used in characteristics extraction part. Wavelet packet can choose different points in time,frequency points on the time-frequency resolution freely, different modulation signal wavelet packet decomposition has different energy distribution. The cumulants play an important role in noise suppression because Gaussian white noise's cumulants is zero when it has more than second-order, and it can be used to classify the modulation signals which have similar energy distribution. Therefore, the method of characteristics extraction which uses the combination of wavelet packet and culumants works effectively.The classifier based on support vector machines is mainly studied, and the hierarchical classifier based on support vector machines is constructed. The support vector machines based on structural risk minimization has tremendous advantages in small samples and better generalization ability. This hierarchical method has lower computational complexity and faster training speed than the single-stage classifier.On the basis of theory study, the experiment based on modulation signals collected at the scene in the short-wave communication is done. With wavelet packet and culumants being characteristics, hierarchical classifier based on support vector machines being classifier, the experiment successfully realizes the automatic recognition of various signals in a large SNR range. At the same time, the performance with different kernel functions and different methods to classify is given. The results presented in this paper show that the digital modulation signals recognition in shortwave communications algorithm is very effective.