Resumen
Sign language aids in overcoming the communication barrier between hearing-impaired individuals and those with normal hearing. However, not all individuals with normal hearing are skilled at using sign language. Consequently, deaf and hearing-impaired individuals generally encounter the problem of limited communication while interacting with individuals with normal hearing. In this study, a sign language recognition method based on a movement-in-a-video detection scheme is proposed. The proposed scheme is applied to extract unique spatial and temporal features from each gesture. The extracted features are subsequently used to train a neural network to classify the gestures. The proposed movement-in-a-video detection scheme is applied to sign language videos featuring short, medium, and long gestures. The proposed method achieved an accuracy of 90.33% and 40% in classifying short and medium gestures, respectively, compared with 69% and 43.7% achieved using other methods. In addition, the improved accuracies were achieved with less computational complexity and cost. It is anticipated that improvements in the proposed method, for it to achieve high accuracy for long gestures, can enable hearing-impaired individuals to communicate with normal-hearing people who do not have knowledge of sign language.