Glottal activity region based processing for speech synthesis