In this paper, we deal with user scheduling algorithms for a codebook based precoding of a multiuser multiple-input multiple-output (MIMO) downlink system considered in IEEE 802.16m and 3rd Generation Partnership Project Long-Term Evolution (3GPP LTE) standardization. We especially focus on per-user unitary rate control (PU2RC) and semi-orthogonal user selection schemes. Also, we examine the disadvantage of the codebook based precoding with these schemes - a loss of sum-capacity when the codebook size is larger than the number of total users who request the communication service to a base station. Finally, we propose two user scheduling algorithms to improve the sum-capacity when the number of total users is smaller than the codebook size. First, we introduce feedback information retransmission algorithm in which each user sends a new codeword again depending on a request of the base station. Second, we address the adaptive feedback algorithm where users transmit one or two codeword indices called preferred matrix index (PMI) depending on their channel condition. The improvement of the sum-capacity is verified through the simulation.