The latest progress in the multiuser detection and independent component analysis (ICA) is reviewed systematically. Then two novel classes of multiuser detection methods based on ICA algorithms and feedforward neural networks are proposed. Theoretical analysis and computer simulation show that ICA algorithms are effective to detect multiuser signals in code-division multiple-access (CDMA) system. The performances of these methods are not identical entirely in various channels, but all of them are robust, efficient, fast and suitable for real-time implementations.