We recently wrote a paper comparing several acted speech databases portraying basic emotional categories.

Here are the the speaker partitions that were used in the paper.