Direction of Arrival estimation of asynchronous DS-CDMA systems using a neural network approach

Knowledge of the direction of arrival (DOA) of a signal is useful both in civil applications such as mobile communications and also in military related applications as well. Even though asynchronous wise both applications are same, they are not when the number of sources are considered. This paper considers the problem of estimating the DOA of users in an asynchronous DS-CDMA system where the technique is not constrained by the number of users in the system. Consequently in this work we do not assume multiple-access-interference (MAI) to be Gaussian distributed as the number of sources could be limited. In the proposed technique, explicit modeling of an asynchronous DS-CDMA system is done. Here we use a combined statistical and neural network (NN) approach to suppress the interference and noise. When selecting the input vector we adhere to our own technique which reduces the noise component largely and the size of the vector is smaller compared to some other techniques. NN training is used to mitigate the interference. To see how volatile the estimations in the variation of MAI component, we consider the worst case scenario and find a closed form expression for the maximum number of users that it can handle as interferes. Simulations under different environments are presented to verify the theoretical results. This technique shows highly accurate estimations as a result of being able to eliminate noise by the specific selection of input vectors and mitigate MAI via training. Even in the cases of limited number of users (as an example military related applications), where Gaussian distributed MAI approximation might not hold, this approach shows highly accurate estimations.