Speaker recognition by humans and machines