Speaker identification by knowledge inheritance and eigenvoices