Application of noisy-independent component analysis for CDMA signal separation

We propose a noisy-independent component analysis (ICA) based CDMA receiver for multiple access communication channels. ICA is a statistical method for transforming an observed multidimensional random vector into components that are statistically as independent from each other as possible. We apply noisy-ICA as a post processor attached to a subspace based CDMA receiver in the presence of Gaussian noise. The proposed algorithm reduces the bias caused by channel noise in ordinary ICA algorithms and further decreases the noise by dimension reduction. The downlink CDMA channel is investigated and we assume that only the code of the wanted mobile user is known (i.e., blind symbol separation). We compare the proposed receiver with noisy-ICA ability to the conventional matched filter, well-known linear MMSE multiuser detector and ordinary (noise free) ICA based receivers. Numerical simulations indicate that the performance of the noisy-ICA based receiver is superior to conventional detectors, and comparable to exact-MMSE (i.e., all user codes are known) detection performance in a synchronous multiple access CDMA channel. The performance of the ordinary ICA based CDMA receiver is improved with noise bias removal and principal component analysis (PCA) based dimension reduction.