Identification, Based On The Modulation Of The Neural Network Software Radio Signals

The basic idea of Software Defined Radio is to constitute an opening, standardization and modularization general hardware platform, and people canimplement different wireless communication functions by loading different software. In order to implement the multimode communication system of Software Defined Radio, it's necessary to automatically recognize the modulation , signals corrupted by the channel noises. We study the important technology of Software Defined Radio deeply in this paper. So that the receiver could work in multi-mode and multi-band based on a same hardware platform..Communication has developed with a rapid pace, and, the modulation mode of communication signals are changing outstanding especially. The study on modulate classification of communication signal has also been developing. Modulation -recognition has played an important role on signal confirm, interfere, discern, communication field. Neural network have an advantage on distributed store of information, extensive parallel processing and highly fault-tolerant characteristic .All these advantages can be used in the modulation-recognition. The learning ability and getting fault-tolerant have distinctive qualities to the uncertain modulation -recognition.This paper extract the characteristic parameters from signal instantaneous amplitude, instantaneous phase, instantaneous frequency to recognize the signals of 2ASK,4ASK,2FSK,4FSK,BPSK,QPSK base on the research on the characteristic set of digital modulation mode recognition. A hierarchical neural network classifier is designed for identifying this six modulation types. And three different BP algorithms are used to it such as momentum gradient algorithm, self-adaptive learning rate algorithm ,and Levenberg-Marquardt algorithm After experimenting and emulating with MATLAB software platform at the SNR is 4dB, 6dB, 8dB, 10dB and 15dB. we can get the conclusion that its successful recognition rate is over 98% at the signal-to-noise ratio (SNR) of not less than 10dB . and the system with gradient BP algorithm has higher successful recognition rate and automatic classification than the other tow algorithms.