Constrained constant modulus algorithm in multiuser detection

A novel blind adaptive algorithm, constrained constant modulus algorithm (C-CMA) for multiuser detection is proposed in the paper. We analyze the convergence behavior of C-CMA in the noiseless case, from which an approximate relation between the constant in C-CMA and the power of the desired user, which guarantees the algorithm converges to the desired stationary point, is given. Simulations illustrate that both the constant and step size make a critical effect on the convergence, and it is better to select a suitable value to guarantee C-CMA converges to a desired stationary point while the convergence speed is bearable. It is also shown that if it is well initialized, C-CMA attains a better performance than the minimum output energy (MOE) algorithm in Honig et al. (1995). Therefore, it is a feasible approach to use the MOE algorithm firstly, after it reaches steady state, and transfer to C-CMA to improve the performance. Lastly, we compare the performance between C-CMA and CMA. It is shown that C-CMA has better performance in the low noise case. But the convergence behaviors of the algorithms are both sensitive to the constant and step size.