For a multi-user (MU) massive multiple-input multiple-output (MIMO) system, its total throughput is directly related to the degree of mutual interference among user's channel vectors accessing the communication system. Selecting a group of users with the least mutual interference can maximize the total throughput, and thus, it is important to select users with low-mutual correlations. The paper designs two effective MU selection schemes with low-complexity. Specifically, based on channel characteristic, reference frames with low correlation are contructed. Then, the user whose channel is nearby an vector in the reference frame is selected, and different selected users are close to different vectors in the reference frame. Consequently, it can guarantee that the selected users have low mutual correlations. Comparing to the existing user selection schemes, the proposed algorithms exhibit better performance. Moreover, the schemes are more flexible, that is, the schemes can be applied into different environments, such as different number of antennas at base station (BS), different candidates, different number of selected users etc, based on different reference frames.