Frequency-Domain Two- to Three-Channel Upmix for Center Channel Derivation and Speech Enhancement

Twoto three-channel audio upmix can be useful in a number of contexts. Adding a front center loudspeaker provides a more stable center image and an increase in dialogue clarity. Even in the absence of a physical center speaker, the ability to derive a center channel can facilitate speech enhancement by making it possible to boost or filter the dialogue, which is usually panned to the center. Twoto three-channel upmix can also be a first step in upmixing from two to five channels. We propose a frequency-domain upmix process using a vector-based signal decomposition, including methods for improving the selectivity of the center channel extraction. A geometric interpretation of the algorithm is provided. Unlike most existing frequency-domain upmix methods, the current algorithm does not perform an explicit primary/ambient decomposition. This reduces the complexity and improves the quality of the center channel derivation.