We propose a new code word selected scheme that directly selects code words to maximize the rate lower bound and combine bit allocation algorithm to reduce the sum rate loss

which is based on analyzing the disadvantages of the traditional limited feedback interference alignment scheme in two-cell interfering MIMO-MAC. Meanwhile

we adopt MAX-SINR criterion to decode the target signal. Unlike the traditional limited feedback schemes which are based on minimum chordal distance or alignment degree criterion

we select optimized code words from the perspective of maximizing the rate lower bound. Further

we generate the sets of code words which are close to perfect precoder

and select the optimized code words in these sets through low complexity and suboptimal global searching. The simulation results verify that the proposed algorithm effectively improves the performance of system and improves the rate lower bound compared with the existing typical algorithms.