Optimum joint detection using soft decision combined with maximum gradient search

The number of computations required by the optimum maximum likelihood receiver to make a decision on the received signal vector grows exponentially with the number of users in a multiuser CDMA environment which makes it too complex to implement practically. In this paper we suggest a modification in the approach of this receiver to arrive at the maximum likelihood decision. Instead of comparing the received vector for all possible combinations of the transmitted data vector, we use the gradient method to scale the surface of the decision metric in the direction of the optimum maximizing point. Combining this with soft decision by restricting the movement by one dimension at a time after making a decision along the respective axis, we can arrive at the maximum likelihood decision with a number of computations that increases quadratically rather than exponentially with the number of users. Results show that the performance of the optimum receiver using this approach is virtually indistinguishable from the standard computationally intensive approach.