MIMO Precoding via Channel Vector Magnitude Selection

The patent transforms the channel matrix H into a transformed channel matrix H' by applying a unitary transformation based on the dominant eigenvectors of the channel covariance matrix. This parameter transformation concentrates the channel energy into fewer dimensions, allowing efficient scheduling and precoding that adapts to rapid channel changes while maintaining computational tractability. The transformation parameter (unitary matrix) is derived from statistical channel properties rather than instantaneous channel state, reducing complexity.

The patent applies different processing to different spatial dimensions of the channel by identifying and utilizing the dominant eigenvectors (local significant components) of the channel covariance matrix. Instead of treating all channel dimensions equally, the method focuses computational resources on the most significant spatial modes, achieving high throughput with reduced complexity by ignoring negligible dimensions.

The patent extracts only the dominant eigenvectors and eigenvalues from the channel covariance matrix, separating the significant spatial modes from the negligible ones. By taking out only the essential components (top N eigenvectors where N << number of terminals), the system can serve multiple terminals efficiently without processing all channel vectors in full dimensionality, thus reducing computational complexity while maintaining adaptability.

The patent reduces the dimensional space in which scheduling and precoding operations are performed by transforming from the original M-dimensional channel space to an N-dimensional transformed space where N is the number of dominant eigenvectors. This dimensionality reduction allows the base station to handle a large number of terminals with manageable computational complexity by operating in a lower-dimensional effective channel space.