Cone constrained constant modulus algorithm for blind adaptive multiuser interference suppression

An algorithm that structurally constrains adaptive filter coefficients within a cone around a given vector is derived. It is applied to the case when the given vector is the nominal spreading sequence of a user in a direct sequence code division multiple access system. The goal is to find the coefficients of an FIR filter that suppresses the multiuser interference. The output of the chip-matched filter produces the input of the FIR filter. The suppression is done by adaptively minimizing the constant modulus cost function. Knowledge of spreading sequences of other users is not needed. The size of the cone depends on the spreading sequence uncertainty and the worst case cross-correlation between the spreading sequences. The convergence and locking of the algorithm onto the desired user are investigated. The algorithm is compared with a back projection algorithm, a linearly constrained algorithm, and a constraintless algorithm. Under severe conditions the cone constrained constant modulus algorithm is the only one that locks onto the desired user.