The parallel multistream model is proposed for integration sign language recognition and lip motion. The different time scales existing in sign language and lip motion can be tackled well using this approach. Primary experimental results have shown that this approach is efficient for integration of sign language recognition and lip motion. The promising results indicated that parallel multistream model can be a good solution in the framework of multimodal data fusion. An approach to recognize sign language with scalability with the size of vocabulary and a fast approach to locate lip corners are also proposed in this paper.
展开▼
