Sparse code multiple access (SCMA) is a novel non-orthogonal multiple access technique, which fully exploits the shaping gain of multi-dimensional codewords. However, the lack of simplified multiuser detection algorithm prevents further implementation due to the inherently high computation complexity. In this paper, general SCMA detector algorithms based on Sum-product algorithm are elaborated. Then two improved algorithms are proposed, which simplify the detection structure and curtail exponent operations quantitatively in logarithm domain. Furthermore, to analyze these detection algorithms fairly, we derive theoretical expression of the average mutual information (AMI) of SCMA (SCMA-AMI), and employ a statistical method to calculate SCMA-AMI based specific detection algorithm. Simulation results show that the performance is almost as well as the based message passing algorithm in terms of both BER and AMI while the complexity is significantly decreased, compared to the traditional Max-Log approximation method.