Fast maximum-likelihood decoding for BLAST systems: Decomposed Matrix Structure technique

In this paper, we propose a technique which is efficient in reducing the required computational complexity of the maximum-likelihood decoding for BLAST system. Our technique is based on the determination of reducible computational complexity for any given BLAST system. In this fashion a specialized decoding can be defined. Our idea is based on the determination of zero entries of the upper-triangular matrix R which occurs in QR-decomposition step of the sphere decoding algorithm. Through the determined zero at k^th row of the matrix R, at least 8 arithmetic operations can be canceled for every time of node visiting in the search tree at level k. Furthermore, properties of the matrix structure can be exploited in sphere decoding in order to reduce complexity of the metric computations for each branch. For this case, considerable saving in the computational complexity can be obtained.