LMA approach in person-independent sign language recognition

In the research of singer independent sign language recognition, the contradiction brought about by the difference among data has made this difference an imperative problem to be analyzed. To understand sign language from the perspectives of human kinematics and linguistics is helpful to remove the individuality of various signers, at the same time ensuring the common characters of sign language words; this is an effective way to tackle the contradiction brought about by these differences. Based on the principles of movement observation science, especially the effort theory in LMA, taking into consideration the structurality of sign language, the paper describes and defines each dimension of effort in sign; then normalize sign language data of unspecific person to some extent, the normalized data is to be used in training and recognition. This method has been assessed under multiple experiment environments and been proved to greatly improve the recognition results.