Classification of Phonological Parameters in Sign Languages
This work assists linguistic annotation and sign recognition models for sign language researchers and developers, representing an incremental improvement in domain-specific methods.
The paper tackles the problem of recognizing individual phonological parameters in sign languages, such as handshape and orientation, using a multi-label Fast R-CNN model on Danish Sign Language data, achieving improved performance by incorporating co-dependence between orientation and location parameters.
Signers compose sign language phonemes that enable communication by combining phonological parameters such as handshape, orientation, location, movement, and non-manual features. Linguistic research often breaks down signs into their constituent parts to study sign languages and often a lot of effort is invested into the annotation of the videos. In this work we show how a single model can be used to recognise the individual phonological parameters within sign languages with the aim of either to assist linguistic annotations or to describe the signs for the sign recognition models. We use Danish Sign Language data set `Ordbog over Dansk Tegnsprog' to generate multiple data sets using pose estimation model, which are then used for training the multi-label Fast R-CNN model to support multi-label modelling. Moreover, we show that there is a significant co-dependence between the orientation and location phonological parameters in the generated data and we incorporate this co-dependence in the model to achieve better performance.