Name
Abstract | ||
|---|---|---|
Gesture recognition is very useful in everyday life for tasks related to computer-human interaction. Gesture recognition systems are usually tested with a very large, complete, standardised and intuitive database of gesture: sign language. Unfortunately, such data is typically very large and contains very similar data which makes difficult to create a low cost system that can differentiate a large enough number of signs. This makes difficult to create a useful tool for allowing deaf people to communicate with the hearing people. The present work presents a sign recognition system for the Spanish sign language. The experiments conducted include separated gesture recognition and sequences of gestures. This work extends previous work by augmenting the size of data set (91 signs), higher than most of the state of the art systems (around 20 gestures). Apart from that, the proposed recognition system performs recognition of dynamic gestures, in contrast to most studies that use static gestures, which are easier to recognise. Finally, the work studies the recognition of sequences of gestures corresponding to grammatically correct phrases in Spanish sign language. For both tasks, Hidden Markov Models are used as recognition models. Results presented for classification of separated gestures are compared with other usual classification techniques, showing a better recognition performance. The current data set, which has been captured using the Leap Motion sensor, will be publicly available to the research community in gesture recognition. |
Year | DOI | Venue |
|---|---|---|
2017 | 10.1007/978-3-319-58838-4_46 | PATTERN RECOGNITION AND IMAGE ANALYSIS (IBPRIA 2017) |
Keywords | Field | DocType |
Gesture classification, Sign language translator, HTK, Leap Motion, HMM, K-NN, DTW | Everyday life,Recognition system,Pattern recognition,Gesture,Leap motion,Computer science,Gesture recognition,Speech recognition,Sign language,Artificial intelligence,Hidden Markov model,Gesture classification | Conference |
Volume | ISSN | Citations |
10255 | 0302-9743 | 1 |
PageRank | References | Authors |
0.40 | 1 | 2 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Zuzanna Parcheta | 1 | 1 | 2.09 |
| Carlos D. Martínez-Hinarejos | 2 | 38 | 10.86 |