A Novel Rate Combining Model for Coded Modulation with Applications to Power Allocation

In this paper, a new rate combining model for coded modulation with practical channel codes is proposed, resulting in a novel utility function. The mathematical properties of the new utility function, such as convexity, are studied. This utility function generalizes the harmonic mean function and demon-strates great potential in various applications in communication and learning problems. The proposed rate combining model is applied to the power allocation problem for parallel Gaussian channels with coded modulation inputs. Our results show that the optimal power allocation based on the new rate combining model outperforms other well known schemes such as mercury water-filling in parallel channels or multiple-input and multiple-output (MIMO) channels.