The present invention provides a capacity maximization-based sparse code multiple access codebook design method. The method includes a first step of determining a codebook parameter S (J, K, F, M, N) of the sparse code multiple access (SCMA), wherein J represents a maximum quantity of users that the SCMA holds, F represents an SCMA mapping matrix generated based on resource blocks and users, K represents a quantity of the resource blocks, M represents a modulation order, and N represents a quantity of non-zero elements in each code word; a second step of constructing a one-dimensional constellation set S0, and calculating based on the one-dimensional constellation set S0 so as to obtain a one-dimensional complex constellation point set Sd; a third step of constructing an N-dimensional complex constellation set according to the one-dimensional complex constellation point set, and selecting an optimal combination sequence as a benchmark code word; and a fourth step of reranking code words of all users according to the benchmark code word, and allocating reranked code words to an indication matrix to generate an SCMA complex codebook. In design of a multi-user codebook, a superimposed coded modulation concept is introduced to improve system performance, and a power difference characteristic of an N-dimensional code word is maintained.