Automatic Modulation Recognition And Parameters Estimation For Communication Signals

Automatic modulation recognition and modulation parameters estimation are key problems in non-cooperative communication systems and Software Defined Radio. Non-cooperate communication is commonly used in civilian and military applications, where it is required to monitor the activities of communication signals in space and intercept the signal information content with little prior knowledge. Software Defined Radio is a key technique in future communications, which require the receiver to work in multi-mode and multi-band based on a same hardware platform. So it is necessary to recognize the modulation type and estimate the modulation parameters automatically. Signal-noise-ratio (SNR) is a key parameter of signals, which has a significant effect on system performance. So it is necessary to estimate SNR of intercepted signal. Direct spread spectrum sequence (DSSS) signal has an ultra-low power spectrum density. How to detect this signal and estimate its parameters without the knowledge of spread spectrum sequence in non-cooperative communication systems is a great challenge with a strong background of application. This dissertation studied these problems based on the statistical characteristic of communication signals.First, we proposed a combined modulation recognizer, which adopted the statistical pattern recognition approach and combined the advantages of decision tree and artificial neural network. Without the prior knowledge of modulation parameters and SNR, 11 types of commonly used modulation signal, including CW, BPSK, QPSK, OQPSK, 16QAM, 32QAM, 64QAM, 2FSK, 4FSK, GMSK and OFDM, can be classified correctly by the proposed recognizer.Secondly, we studied the modulation parameters estimation of intercepted signals in non-cooperative communications. The research on the roll-off factor of M-PSK and M-QAM signals shows the fact that the inaccuracy of the factor has a little effect on the system performance. Thus it is not necessary to estimate this factor in the receiver design, but to configure a fixed value. The research on the modulation index estimation for M-FSK signals shows that the proposed estimation algorithm can improve the timing synchronization performance significantly and has a trivial effect on the final BER performance. The research on the baud rate estimation for M-PSK, M-QAM and OQPSK signal shows that the proposed estimation algorithm can obtain a baud rate value with a short observation interval, which is precise enough for the demodulation requirement. Thirdly, we proposed two non-data aided SNR estimation algorithms. One is the modified algorithm that is based on eigenvalue decomposition and the principle of minimum description length, the other is the algorithm based on Gibbs sampler. These two algorithms can estimate the SNR precisely in both AWGN channel and multipath channel. Additionally, the latter one can give an estimation of channel information.Finally, we studied the detection of DSSS signal and its parameter estimation in non-cooperative communications using the strong self-correlation and weak cross-correlation of spread spectrum sequence. We estimated the baud rate, chip rate and carrier frequency of DSSS signal based on the cumulant of self-correlation function of intercepted signal. Then we proposed two measurement functions. One is used to synchronize the spread spectrum sequence to determine the correct starting position of the sequence in the intercepted sequence; the other is used to estimate the chip value. The computer simulation results verified our theoretical analysis.In conclusion, the proposed automatic modulation recognition algorithm and modulation parameters estimation algorithm can effectively improve the performance of receiver in non-cooperative communication systems and Software Defined Radio.