Deterministic waveform-preserving blind separation of DS-CDMA signals using an antenna array

We consider deterministic blind separation of DS-CDMA signals using an uncalibrated antenna array, subject to: (i) unknown flat fading; (ii) unknown multipath; (iii) quasi-synchronous transmissions; (iv) oversaturation; (v) lack of (or, inexact) knowledge of the spreading codes; and (vi) loose power control. We establish a general identifiability result that allows blind recovery of the source signals without requiring independence or whiteness, nor finite alphabet or constant modulus. The results apply to both deterministic and stochastic sources; knowledge of the spreading codes is not required; and, actually, certain matrices involved may be allowed to loose rank without sacrificing identifiability. A provably monotonically convergent trilinear alternating least squares algorithm is proposed to recover the source signals in the noisy case. Aside from their theoretical generality and practical value, our results have important ramifications in eavesdropping of secure spread-spectrum communications, and the judicious choice of user codes.