A Step towards Textual Instructions to Virtual Actions

Extracting large body movement dataset from textual instruction could be useful for serious games for health to learning by demonstration in robotics. The interpretation of instructions for the automatic generation of the corresponding motions (e.g. exercises) and the validation of these movements are difficult tasks. In this article we analyzed a step towards achieving automated extraction of motions from textual instruction of physical exercise. To achieve this, we have recorded 36 different exercises using a XSENS motion suit. Using the XSENS data we have developed one automated physical exercise engine with three different modes. In this engine all exercises are generated in skeleton mode. Out of these three mode, in first mode it is limited to the mentioned exercises in the application. In second mode user can generate any combination of exercise (upto four) of the mentioned exercises. And in the last mode user can generate any exercise using text input or uploading instruction sheet to the system. We are still working on the third mode, which is not fully efficient now.