Incorporating global variance in the training phase of GMM-based voice conversion | IEEE Conference Publication | IEEE Xplore