Method and apparatus for soft output fixed complexity sphere decoding detection

摘要

Provided are an SFSD detection method and apparatus, and the method includes: QR decomposition is performed on a channel response matrix to acquire a Q matrix and an R matrix; the conjugate transpose of the Q matrix is multiplied by a received signal to acquire an equalized signal of the received signal; ML path detection is performed on the equalized signal, reserved nodes in respective layers are decreased layer by layer to acquire an ML path, and branches as many as iterations are reserved; ML complementary set path detection is performed on the branches, and all nodes of an acquired complementary set layer are reserved and reserved nodes in other layers are decreased layer by layer to acquire an ML complementary set path; and LLR information of each bit of each symbol of each layer is acquired according to the ML path and the ML complementary set path. For more-than-two-layer MIMO, the disclosure can acquire a detection performance approaching the ML performance and meet requirements on acceptable hardware implementation complexity.

主权项

1. A method for Soft-output Fixed-complexity Sphere Decoding (SFSD) detection, comprising:
performing QR decomposition on a channel response matrix to acquire a Q matrix and an R matrix; multiplying a conjugate transpose of the Q matrix by a received signal to acquire an equalized signal of the received signal; performing path extension starting from a top layer of the R matrix, which is a layer having only one non-zero element, to a bottom layer of the R matrix; reserving all nodes of the top layer of the R matrix; decreasing, layer by layer, reserved nodes in layers below the top layer of the R matrix to acquire Euclidean distances of respective branches; selecting a branch having a minimum Euclidean distance as a Maximum Likelihood (ML) path, and reserving branches as many as iterations; performing path extension starting from a top layer of acquired complementary set layers of respective reserved branches to corresponding bottom layer of the acquired complementary set layers; reserving all nodes of the top layer of the acquired complementary set layers; decreasing, layer by layer, reserved nodes in layers below the top layer of the acquired complementary set layers to acquire Euclidean distances of respective branches; selecting, layer by layer, starting from the top layer of the acquired complementary set layers, a branch having a minimum Euclidean distance as an ML complementary set path of each layer; and acquiring Likelihood Ratio (LLR) information of each bit of each symbol of each layer according to the ML path and the ML complementary set path.