Blind code estimation of multi-antenna direct-spread CDMA with long-code signal using decomposition technique

Aiming at the problem of blind estimation of the spreading code of multi-antenna Direct-Spread CDMA Long-Code (DS-CDMA LC) signals with poor interpolation effect, the idea of segmentation is used to construct the third-order tensor of the received signal into blocks according to the spreading gain, which is equivalent to Tucker decomposition. The sub-tensor of the model then uses the variable step gradient descent algorithm for the sub-tensor to perform Tucker decomposition to obtain the spreading code segment and the receive gain matrix. According to the same characteristics of the sub-tensor's receive gain matrix, the cross-correlation matrix of each sub-tensor's receive gain matrix is used to estimate the permutation matrix to remove the ambiguity of the sequence of spreading code fragments. Finally, the autocorrelation of the spreading code is used to remove the amplitude ambiguity of the spreading code segment to obtain the spreading code of each user. The simulation results verify the effectiveness of the algorithm.