Cross-language speaker recognition using spectral moments