Multimodal integration technique in a map-based system using spatial and deictic relationship

The purpose of this research is to improve the multimodal integration technique, of multiple speech and sketch inputs in human computer interaction. In our technique, we integrated the sketch and speech representation model, taking into consideration the normalised relative distances between objects in the sketch, spatial preposition and the deictic words from the speech. We compared the proposed integration technique with the integration technique using only spatial information. The results showed an increase in the matching of correct pairs of spoken and sketched inputs, by 11%, that is, from 52% to 63%. The improved integration technique has shown that the probability of matching the wrong pair of spoken and sketched inputs can be reduced.