Statistical eigenvoice: speaker features within S+N framework and a way towards language-independent voice conversion | IEEE Conference Publication | IEEE Xplore