In multimodal human-robot interaction (HRI), the process of communication can be established through verbal, non-verbal, and/or para-verbal cues. The linguistic literature shows that para-verbal and non-verbal communications are naturally synchronized. This research focuses on the relation between non-verbal and para-verbal communication by mapping prosody cues to the corresponding arm gestures. Our approach uses the coupled hidden Markov models (CHMMs), which could be seen as a collection of HMMs, one for the prosodic characteristics stream and six HMMs for the arm movements stream.