Towards the automated visualization and analysis of signed language motion - Method and linguistic issues

Abstract

This paper presents a semi-automatic method for the visualization and analysis of motion in signed language from a digital video. The method detects the different parts of the signer's bare skin on the video, tracks the motion of the different parts of the body and represents the frame-wise motion of the parts of the body with statistical descriptors. Results of an experiment are demonstrated and discussed. The main linguistic issue deals with the relationship between signs and transitional sequences. Evidence is provided for the claim that lexical-phonological and transitional sequences are qualitatively different in terms of their motion characteristics.