Abstract
This paper discusses the pivot pattern of tongue movement. In this pattern, there seems to be a point in the vocal tract where there is no motion, but there is motion at points of the vocal tract anterior and posterior to the pivot point. Based on tongue edge tracings of frames from ultrasound and x-ray dynamic imaging of the vocal tract, I will show that the pivot pattern is used in a variety of sequences, and I will discuss the possible causes of the pattern.

Abstract
In this paper we present the results of experiments on the synchronization of pointing gestures and speech. Evidence will be presented to show that there's a great deal of regularity in the way that pointing gestures are aligned on a small temporal scale with the syntactic boundaries of the phrases that they accompany. Furthermore, it will be shown that the alignment of pointing gestures to syntactic domains is sensitive to prosodic effects.

Abstract
In this paper we present a new method for extracting dynamic information from the speech signal. This method is based on extracting dynamic extractors from a reconstruction of the dynamical system's phase-space. We then summarize the performance results obtained from the new system as implemented in a Hidden Markov Model recognizer. This is the final component in a larger speech recognition system, which includes a high-level grammar and a gesture recognizer.

Abstract
To enable natural human-computer interaction in a virtual environment, information has to be captured from a number of human communication channels including speech, gesture, gaze, and facial expression. Furthermore, the information from different channels has to be aligned and correlated in order to obtain the overall meaning of the communication act. This paper focuses on relating and aligning the information from the speech and gesture signals. It will be shown that the speech gesture alignment problem is non-trivial, and that syntactic and prosodic information are a key to the alignment. We will then present the architecture of an adaptive Hidden Markov Model-based speech and gesture recognition system which incorporates the prosodic and syntactic alignment constraints.