The invention provides a genetic property combining large-scale MIMO (Multiple-Input Multiple-Output) system LAS (Likelihood Ascent Search) signal detection algorithm, which is characterized in that a global optimization genetic algorithm is introduced on the basis of an original LAS algorithm, a standard cost function of a solution vector is adopted to act as an iteration stopping criterion, the number of iterations is set to be self-adaption, the required number of iterations is updated by the quality of the solution vector, the execution efficiency is improved, elements of the last high-quality solution vector are inherited at the same time, a variation factor is added to change values of elements at some positions in the vector, the same search region is jumped out, and search is enabled to be executed in different high-quality solution regions, so that the number of floating point calculation times required by the algorithm is decreased, and the system complexity is reduced. Meanwhile, the search capacity is improved, the bit error probability of the system is reduced, and thus the reliability of the system is improved.