Study On Pn Sequence Blind Estimation Of DS-CDMA Signal Based On Principal Component Neural Network

Spread-spectrum communication enables the transmission of useful signals over broadband.The bandwidth of the spread-spectrum signal is muchwider than the bandwidth required for the actual transmission of the desired signal,and this ensures the immunity,low interception and confidentiality of communications,thus secure communication is achieved.In order to meet the rapid growth of military and civilian demand for communication capacity and frequency band,DS-CDMA signals are used in modern communication systems widelyin many countries.However,the spread spectrum "concealment" strategy of DS-CDMA signals uses the spectrum spreading technique to reduce the power spectral density of the wanted signal until it is completely masked by noise,thus a serious challenge is brought to the non-cooperative communications party or the communication counter-party to detect useful information.Therefore,it is becoming increasingly important to study the pseudo noise(PN)sequence blind estimation of DS-CDMA signals.The main work of this thesis includes the following aspects:(1)Aiming at the problem that matrix decomposition method with the data vector that can be processed is limited and the complexity is high in DS-CDMA signal PN sequence blind estimation,the method which is based on Sanger multi-principal component NN is studied.Firstly,the period segmented DS-CDMA signals are chosen as NN input and the symbol function of each weight vector is used to represent the PN sequence of each user.Then the weight vectors of NN are trained repeatedly until convergence through entering the signal constantly.Finally,the PN sequence of each user can be rebuilt by the symbolic function of each weight vector.Furthermore,an optimal variable step convergence model is studied via the recursive least square(RLS),which improves the convergence speed of the network greatly.(2)For the problem that the convergence rate of the Sanger NN is slow in the blind estimation of DS-CDMA signal PN sequence,the method which is based on LEAP multi-principal component NN is studied.On the base of Sanger NN,this method adds the matrix operation to the weight updating formula of the NN,and the Schmidt orthogonalization of the connection weight is realized.(3)For the problem that the LEAP NN method has high complexity due to the matrix operation in the network and it can`t converge when the PN sequence is long,the method which is based on APEX multi-principal component NN is studied.On the base of Sanger NN,this method realizes the de-correlation of connection weights by adding lateral connections at the output of the NN.