Research On Blind Estimation Of The Parameter For Communication Signals

Communication reconnaissance is one of the most important components in modern electronic warfare. It is the premise for communication countermeasures. Blind detection of the communication signals can recognize the enemy signal and intercept the information in military applications. It is also applied for the government to survey the radios operation and determine its validity automatically in civilian purposes. Blind estimation of the parameters for communication signal is also a valuable support for the general receiving platform realized by software. The main research works and achievements are listed as follows:1. The training time was reduced by the proposed modified BP algorithm for modulation identification. The BP algorithm with momentum, the complex conjugate method and the adaptive learning for BP algorithm was studied. It was proved in theory and shown by the simulation results that the complex conjugate method was fastest for training and the best learning algorithm was the adaptive complex conjugate method.2. The method for the selection of initial weights was improved so that the training time was reduced. The excited function was sigmoid function and the adjustment of the error-performance function was used as an indication. The transfer function was approximated to a linear function round the zero. When the adjustment was more than the specified range, the transfer function recovered to sigmoid function. It was shown by the simulation that the performance of the network had been improved evidently.3. Removing the redundancy based on K-L transform was first proposed, which could determine the optimum topology by one step. The number of the extracted features decides the number of the nodes in input layer, and the number of recognized modulation types decides the number of nodes in output layer. So the nodes number in hidden layer is the key to determine the computation. The redundancy could be removed applying the K-L transform. It was shown the processing speed could be improved obviously by the optimized network with a higher stability. The performance of the network did not change after the optimization.4. A complete algorithm for modulation identification was presented. Researching on the learning algorithm, initial weights selection and the topology of the network, the proposed algorithm had the advantage of good performance with a higher speed in modulation identification.5. The algorithm based on multi features for modulation identification was proposed. The modulation identification was implemented by two steps. First it was the interclass identification, and then the intraclass identification was done. Different features were used for the interclass modulation identification of different signals in order to make the classifier simple. It was shown by the simulation results that the method could generate a higher percentage of correct identification.6. Fuzzy theory was applied for modulation identification. It had the advantage of easy to obtain the features with a higher processing speed. It was expected to improve the performance by imitating the way that people processes a complex signal.7. The VHF three-array nulling antenna was introduced. With a brief and effective algorithm, the adaptive antenna system is proved by experiments that the system can suppress interference effectively.