A study of the influence of vocalization on lip motion for command input interfaces

Lip motion features such as changes in lip width and length can provide important information for authenticating personal identification or recognizing lip motions as commands. Unfortunately, lip motion can be significantly influenced by vocalization. In this paper, we investigate the differences in lip motion directly attributable to vocalization, with specific focus on the transition and variation of two lip motion features, namely, the aspect ratio (A_i) and area of the rectangle that bounds the lip (raS_i). The results of our experiment reveal that the transitions of non-vocalized data tend to be larger than those of vocalized data and that the inter-frame differential value tends to increase as well, by averages of 18.6% (raS_i) and 17.5 % (A_i).