The invention discloses a low-complexity MIMO detection algorithm based on a GAIF (grouped alternative iteration filter), and the algorithm enables a result of a conventional detector to serve as an input initial value of an GAIF algorithm, and enables the input initial value to be converged to an output value which is smaller in Euclidean distance with a received signal vector. In each iteration process, the GAIF algorithm selects a plurality of most unreliable symbols according to a detection error deviation, and carries out the jointed detection correction of the symbols in a submask space of other symbol interference. When the Euclidean distance between the output result of the filter and the received signal vector cannot be reduced, the iteration of the GAIF algorithm terminates. The algorithm irons out the defects that a conventional algorithm is high in complexity or low in performance. Through converting a high-dimensional signal into a series of independent low-dimensional signals for detection, the MIMO detection algorithm takes the little increase of complexity as the cost, and remarkably improves the performance of a conventional detection algorithm.