On the performance of source separation with constant modulus signals

Cramér-Rao bounds (CRBs) are developed for narrow band source separation, when the sources are constrained to have constant modulus (CM). The bounds are appropriate for multi path CM sources, in blind, semi-blind, or fully known cases. Source separation bounds are contrasted for calibrated and uncalibrated arrays. It is shown that, from the CRB perspective, calibration adds no additional information. Closed-form CRBs are given for a single source, and two-source examples are also presented. Optimality of the analytical constant modulus algorithm (ACMA) for blind CM-source separation is demonstrated, achieving the appropriate constrained CRBs over a wide SNR range in challenging multipath scenarios.

[1]  B. C. Ng,et al.  On the Cramer-Rao bound under parametric constraints , 1998, IEEE Signal Processing Letters.

[2]  Brian M. Sadler,et al.  Bounds on SIMO and MIMO Channel Estimation and Equalization with Side Information , 2002, J. VLSI Signal Process..

[3]  Arogyaswami Paulraj,et al.  An analytical constant modulus algorithm , 1996, IEEE Trans. Signal Process..

[4]  Brian M. Sadler,et al.  Bounds on uncalibrated array signal processing , 2000, Proceedings of the Tenth IEEE Workshop on Statistical Signal and Array Processing (Cat. No.00TH8496).

[5]  Brian M. Sadler,et al.  Bounds on bearing and symbol estimation with side information , 2001, IEEE Trans. Signal Process..