Speech modeling and processing by low-dimensional dynamic glottal models