Gesture recognition for elderly care is an approach to classify gestures performed by the elderly to convey specific messages. Nursing homes or caretakers are often hired to take care of senior citizens and are responsible to keep them safe. Hence, this study will be useful to assist caretakers in providing needs requested by the elderly when they are absent. A collection of dynamic hand gesture recognition (HGR) datasets consisting of ten gesture classes with 630 videos is used to build the models. The ten gestures will represent requests for help to perform daily activities namely eating, toileting and dressing. Specifically for this research, we have infused Islamic hand gestures such as gestures to perform prayers and read the Quran. In this paper, we adopted the action recognition pre-trained model into the HGR by using CNN RNN and Transformer with CNN models. The result of this study shows that Transformer with CNN model has higher accuracy in recognizing hand gestures compared to CNN-RNN model.