Compound gestures are more complex and challenging in activity recognition due to more detailed behavioral context which is needed to be identified. This study proposed a context-aware healthcare monitoring system that recognizes fine-grained compound gestures of the elderly by using wireless body-worn inertial sensors and smart phone based RFID detectors. The Dynamic Time Warping (DTW) pattern matching technique was implemented on the classification of motion acceleration patterns. The proposed system aimed to provide solutions on the three major issues of activity recognition, and they are ambiguity, generalization, and segmentation. Through collecting the location information, object usage condition, and processing the motion pattern by DTW, the inference engine estimates the recognized activity. The overall recognition accuracy on compound gestures in this study is 88%. The results indicate that in general, the proposed healthcare monitoring system provides good ability on the recognition of fine-grained activities in a cost-effective approach.