Bridge between signal processing and information theory

Minimum mean square error (MMSE) is a common measure used in signal processing. On the other hand, mutual information is a common measure used in information theory. A connection between MMSE and mutual information has been revealed recently, which establishes an interesting MMSE-I relationship. In this talk, we will discuss the applications of the MMSE-I relationship in communication systems. In particular, we will examine linear precoder design for multiple-input multiple-output (MIMO) systems. The problem is conventionally studied using a signal processing approach based on the MMSE principle. Such an approach is most suitable to uncoded systems. In this talk, we will show that the problem can be reformulated for coded systems with iterative receivers, and optimized based on the MMSE-I relationship. This leads to a unified framework from both signal processing and information theory points of views. The new design approach has many attractive advantages, such as capacity approaching, robust against channel information uncertainty and flexibility.