An auditory analysis of a speech data base in French highlighted the use of rhythmic patterns by speakers. The rhythmic patterns are built by employing successive prosodic groups of the same length expressed in a number of syllables. These rhythmic patterns were not present to the same extent in the elocution of every speaker, but were speaker dependent. Although the auditory detection of the prosodic units in the speech signal is based on a whole group of acoustic parameters, an attempt is made here to retrieve the auditory detected prosodic units automatically. Therefore, one of the goals of this study is to evaluate and approximate the group of acoustic parameters used during auditory segmentation by three parameters such as the vowel duration, vowel energy and the slope of the F0 (measured on the last vowel of the prosodic unit), as they are easy to calculate and are considered as playing a prominent role in the prosodic demarcation of speech.