A new speech coding model based on a least-squares sinusoidal representation | IEEE Conference Publication | IEEE Xplore