| The automatic identification and parameter estimation of modulated signals have been important topics in the field of communications. In the non-cooperative field, it is required to monitor the wireless signals in space and intercept the information they carry with little priori knowledge. Software Defined Radio is a research hotspot in recent years, its core is the completion of a multi-institutional intercommunication. The above two aspects both need to know the modulation type and the corresponding modulation parameters before the subsequent correct demodulation. In this paper, based on the previous studies, a set of robust characteristics were extracted from the time domain, frequency domain, higher order spectral domain and cumulant domain. By support vector machine classifier, the recognition of ten signals including CW, ASK,2FSK,4FSK, MSK, BPSK, QPSK, OQPSK,8PSK,16QAM was completed. Then the carrier frequency and symbol rate were estimated.In terms of modulation identification, the signals were divided into two subsets including single-line signals and multi-line ones according to the spectrum at first. Then they were further divided into constant envelope and non-constant envelope ones by the parameter which characterized the envelope fluctuation. In the classification of frequency modulated signals, the paper especially took the small modulation factor and continuous phase MFSK signals into cosideration. Through the feature parameter constructed by the instantaneous frequency, the signals could be divided into the binary and quaternary frequency modulation. For the distinction of2FSK and MSK, the2FSK was considered to be three kinds of modulations under different conditions, and three feature parameters were extracted from the spectral. In the identification of MPSK signals, the features extracted from the improved higher order spectral could obtain better noise immunity and achievability. At last, the distinction between8PSK and16QAM was based on the higher order cumulants. Simulation results show that the correct recognition rate of each signal is greater than95%when the SNR value is greater than lOdB. In addition, all the algorithms have considered the pulse shaping technology in engineering applications. Compared with the methods based on cyclic spectrum and wavelet transform, the proposed algorithm has a smaller amount of computation, and is more conducive to the realization of the engineering and real-time signal processing. When it comes to the parameter estimation, this dissertation mainly researched the carrier frequency and the symbol rate estimation methods. The periodogram, square law, and spectrum symmetry carrier frequency estimation methods were studied, and the latter two means were improved. Computer simulation shows that the improved algorithms have better anti-noise performance than the original ones. In the aspects of symbol rate estimation, the thesis introduced the code rate detection methods of ASK, MFSK and MPSK based on the instantaneous characteristics. Besides, the paper focused on the introduction of the algorithm to estimate the OQPSK’s symbol rate. The formula derivation and computer simulation both verify the correctness of the presented algorithm. In addition, the estimated performance against roll-off factor, modulation coefficient, SNR value was also investigated.Finally, the proposed algorithms were implemented on the DSP-based hardware system. Analog signal was firstly sampled by AD, and then the discrete samples were cached by FPGA. DSP completed the identification and parameter estimation of unknown signal segments, and conveyed the results to the upper machine on PC to display. The algorithm’s performance on the hardware platform was tested. Results show that it declined with respect to computer simulation. Each signal’s correct recognition rate is higher than90%when SNR value is greater than lOdB. The hardware implementation further proves the correctness and realizability of the proposed algorithm in this thesis. |
PDF Full Text Request