Consistent Speaker Identification via Wigner Smoothin