Noise Statistical Characteristics Of Extremely Low Frequency Channel And Its Application

The Extremely Low Frequency(ELF) communication system is designed to operate in the frequency between 30 Hz and 300 Hz.Because the low attenuation rate in seawater and in the earth-ionospheric waveguide,the ELF communication can provide nearly global coverage and shore-to-submarine communication to submerged submarines below 100M seawater.So the ELF communication has extremely important value in military areas.Limited by the antennas and the transmitter of the ELF communication system,the available signal strength at the receiver is very small and the performance of the receiver system can have a significant rely on the ELF noise statistical characteristics.The ELF noise is known as atmospheric noise and its statistical characteristics is non-Gaussian. Conventional linear receivers which are effective and optimum for white Gaussian noise can drastically degrade the performance or even made ineffective when operating against the non-Gaussian noise.On the other hand,the performance of communications operating in such impulsive channels can be greatly enhanced if the true statistics of the channel are known and exploited.This dissertation focuses on the model of the ELF noise,verified noise model by real data,the parameters estimation of noise model and engineering application which implement the ELF receiver.The research fellow the method of combining theory with engineering application.The main contributions are as follow:1) For analyzing the Amplitude Probability Distribution(APD) of the noise from the device of GSM-06,the measured noise is processed by FFT filter which is used for suppressing the interfere of 50Hz.The Non-Gaussian property of ELF noise is testified by Lilliefors Hypothesis Testing.With considering the APD and the Q-Q plot(Quantile-Quantile Plot) of real ELF atmospheric noise based on the parameters estimation of Class A noise model,The conclusion is made by excellent agreement between the measured data and the theoretical curves which are provided by Class A. 2) This dissertation addresses the loss performance of Turbo decode in No-Gaussian impulsive noise by computer simulation and proposes a new modified decoding algorithm for Turbo code by using Middleton's Class A impulsive noise model based on statistical nature of the amplitude.The decoding algorithm improved performance by modified extrinsic and channel information of traditional MAP for suiting additive white Class A noise,In addition,The results of computer simulation have shown that the proposed algorithm has better performance than traditional MAP decoding algorithm in Class A noise.3) Two estimations of parameters Class A noise model is provided,the one is based on Characteristic Function Method(CFM) and the other is based on Bayesian estimator, the Bayesian estimator of the Non-Gaussian model parameters is derived and calculated by the Markov Chain Monte Carlo(MCMC) procedure.The considered estimator provides a novel method for advantage of low-complexity,fast convergence with small sample sizes.The estimator of CFM is small computations,fast convergence with large sample sizes,especially suit for the case which the property of channel noise is stabilization.The estimation parameters of the simplified Class B noise model is derived. The simplified class B distribution is represented by a weighted sum of confluenthypergeometric function.The applicability of Class B noise model is limited because it's very complexity form.In this chapter,the efficient estimation of the Simplified Class B model parameters based on least square gradient method is derived.The considered estimator is fast converges and low-complexity with performance approaching theoretical optima for large data samples.The kernel difficulty of application for Class B model is solved.4) The estimation parameters of two-dimensional Class A model is derived,for the optima reception of ELF signal is implemented in every direction.An efficient estimation of two-dimensional version Class A noise model parameters based on Markov Chain Monte Carlo(MCMC) is derived.The estimator can estimate five-parameter and hidden states for two-dimensional Class A noise model simultaneously.Simulation of this estimator indicates that this considered estimator is converges rapidly and low-complexity with performance approaching theoretical optima for small data samples,although it has large computations.The foundation of optima reception theory and engineering of application is provided for the future ELF communications in every direction.5) This dissertation further addresses the engineering implementation of the ELF receiver.The block diagram of the principal components of the receiver,such as filter, MSK-demodulation,Antijamming,and decoding is given when considering the base of design theory,optimum performance and details of the design receiver.For the signal-processing component of the receiver,the estimation parameter of Class A noise model is adapted.The performance of the ELF receiver is improved mostly in the ELF non-Gaussian noise when the statistics of the channel noise are exploited.