A trainable singing voice synthesis system capable of representing personal characteristics and singing styles