A computer program is described which converts the speech signal into a string of intonation units, connected to the syllable chain. The main steps in this system are [1] the segmentation into syllabic units, [2] the perceptual integration of prosodic parameters within each syllabic nucleus, [3] the assignment of prosodic features and [4] the parsing of the resulting chain of feature bundles according to a language-specific intonation grammar.