MIMO signal detection based on tree-search involving nodes at multiple depths and a level-biased metric using soft decoding

The paper presents a category of detection algorithms for Multiple-Input Multiple-Output (MIMO) systems called Tree Search Techniques involving nodes at multiple levels. The techniques also combine the features of classical depth-first and breadth-first search methods. The descriptions of the algorithms are given. Initially the symbols are decoded using Minimum Mean Square Error (MMSE) criterion. An array is allocated for storing the nodes which belong to the category of legitimate paths leading to the correct solution. Amongst these nodes the ones which have the smallest euclidean distances are stored in an intermediate array which determines whether their children should be considered for extension of the search. This is done by juxtaposition of the partial solution of the node and the MMSE solution. The children of the selected nodes are placed in the initially allocated array and the iteration continues until the leaves of the tree are reached. It is seen that with the course of the iteration the array consists of nodes at multiple levels. A level-bias metric is also used which facilitates deriving a solution quickly at the expense of a trade-off between throughput and Bit Error Rate (BER) performance. Using Log-Likelihood function it is possible to obtain soft decoding. This leads to increase in the reliability of the decoding process. The proposed algorithms are simulated on a 4×4 16-QAM and 4×4 QPSK multiple-input-multiple-output system. A comparative study of the BER performance between the proposed algorithms with best-first tree search techniques and Sphere-Decoding techniques is done. Furthermore after decoding the symbols log-likelihood ratios are used for improving the reliability of the decoding process.