In this work, we propose a novel variant of the conventional p- Constant Modulus Algorithm (p-CMA)algorithm. The rationale behind our work is that the exponent p in the standard CMA can play significant role inthe convergence speed and steady-state performance of the algorithm. It was observed that the larger value of pcan provide faster convergence at the cost of higher steady-state excess mean-square-error (EMSE). On theother hand, smaller value of p can give lower EMSE at the cost of the slower speed of convergence. Thus, inthis work, we develop a time variant p parameter by utilizing the energy of the estimation error such that thevalue of p takes larger values in the beginning of adaptation and reduces to smaller values near steady-state.Thus, it can achieve both faster convergence and smaller steady-state EMSE. Simulation results are provided tosupport our theoretical claims.
展开▼
