A novel LLR scaling factor selection using LDL Bayes classifier

In this paper, we suggest a novel learning algorithm that determines a scaling factor to adjust the log likelihood ratio (LLR) extracted from the symbol detector. The LLR scaling factor to overcome the constraints of HW can affect the performance depending on the determined value. It has the ambiguity that selects the value of LLR scaling factor since determining the value would be changed rapidly under the small difference of environment. On the other hand, among machine learning types, label distribution learning (LDL) is known to be beneficial for learning ambiguous data because it distinguishes the differences regarding each label as the distribution. After the proposed learning algorithm using LDL Bayes classifier learns the ambiguity of each scaling factor, the learned Bayes classifier determines the LLR scaling factor offering higher performance. As a result of the simulation, the proposed learning algorithm provides improved performance compared to the existing fixed LLR scaling factor.