Research On Digital Modulation Recognition

Digital signal modulation recognition is an important research subject in the field of communication. In the aspect of military, and also civilian, it has comprehensive applied prospects. At present, the way in which we identify the type of modulation signal is mainly depends on the manual work at domestic. However, it is was inefficient, subjective and harmful to the operator, and further more hardly fit to the communication requirement gradually. Therefore, the automatic type recognition of modulation signal is on the agenda. The relative departments need it eagerly.The paper research the recognition problem of digital modulation signal from the prospect of pattern recognition.At first, the paper briefly introduces the basic information of modulation signal referred. It helpful to future analyzes and process.And then, the paper focuses on the feature extraction which is the most important part of pattern recognition. At this part, we research the extraction of the energy feature by Wavelet Packet Transform, and analyze the feature. Subsequently, we introduce the Hilbert-Huang Transform (HHT), especially, its main components-Empirical Mode Decomposition (EMD). According to the method and characteristics of empirical mode decomposition, we propose a new feature by improve the wavelet packet energy feature. The way is calculate the Intrinsic Mode Function (IMF), by EMD. And then get the instance phase of IMF. The wavelet packet energy of the instance phase is the final feature.The first part of classifier of digital modulation signal recognition uses the instance phase wavelet packet energy and the wavelet packet energy as the feature set, and selects the features from it by Genetic Algorithm (GA).That is an alternative part. And the second part applies the Support Vector Machine (SVM) to classify. Then, the work of design is completed.The paper tests these two types' features by the simulated signal. The result primarily proves the efficiency of the new feature. And the performance of the new feature is better than the old one. We also test recognition rate of the entire method which reach 95% as the SNR is 4db.That is better than expectation. Finally, we test the method by the real signal, which was received form shortwave environment. The performance shows the effective of the method.