To analyze the parameter estimation performance of PARAFAC decomposition under Constant Modulus (CM) constraint, Cramer Rao Bound (CRB) expressions for PARAFAC decomposition under “First Row Known” and CM constraints were proposed. The fitting algorithm, named TALS CM, was also developed. Simulation results showed that, compared to the normal PARAFAC decomposition, CM constrained PARAFAC decomposition has a lower CRB. The performance of TALS CM algorithm was close to its CRB, which implied that TALS CM is asymptotically efficient.TALS CM algorithm has better performance than traditional TALS algorithm. Utilizing CM property of source signal improved the performance of the PARAFAC based signal processing algorithms.
