Blind adaptive NPCA algorithm based on neural network tracking subspace

Recently developed adaptive multiuser detection techniques are briefly reviewed, and the novel principal component analysis (NPCA) blind adaptive multi-user detection algorithm based on bilinear neural network tracking signal subspace is proposed. It can be obtained blindly, i.e., only requiring prior knowledge of the signature waveform and timing of the desired user. Compared with previous blind adaptive linear minimum-mean-square error (MMSE) and projection approximation subspace tracking with deflation (PASTd) algorithms, the proposed NPCA algorithm offers lower computational complexity and better performance in terms of the steady-state signal interference noise ratio (SINR) and bit error rate (BER). Moreover, applying this NPCA algorithm, detector's traceability and robustness against signature waveform mismatch are improved. Numerical simulations and experiments have shown the superiority of the proposed algorithm.