Measures such as speaking rate and articulation rate are commonly used in the assessment of various speech disorders. Such measures are also used to adapt automatic speech recognition systems to avoid performance degradation encountered with high speech rates. This paper illustrates a hidden Markov model-based method to estimate the number of syllables, which can further be used to estimate speaking rate or articulation rate. The method uses the Viterbistate sequence for syllable estimation. The algorithm wastes tedon two datasets; the TIMIT dataset achieved an error of 14.6% and correlation of 0.81, while the Switchboard data yielded an error of 21.2% and correlation of 0.88.