Speech Analysis and Recognition Synchronised by One-Quasiperiodical Segmentation