An improved signal subspace estimation method and its application to multiuser detection

An improved signal subspace estimation method is proposed and applied to blind multiuser detection. Firstly, a rough signal subspace is estimated from the received signal by direct eigendecomposition or subspace tracking. Then, the estimated signal subspace is modified with the desired user's signature waveform that is generally known. The new multiuser detector based on the improved signal subspace can offer substantial performance improvement over recently proposed multiuser detectors based on signal subspace estimation with a little attendant increase in computational complexity. Moreover, the estimated dimension of the signal subspace can be significantly reduced with little performance reduction when there are some interference signals weaker than the desired signal. Numerical simulation results are provided to support our claims.