Spectrum Variation of Clipped Speech Signals