This paper presents a framework for hand gesture recognition based on 3-channel electromyography (EMG) sensors. In the framework, the start and the end points of meaningful gesture segments are detected automatically by checking the cross points of EMG signals and their moving average curves. Then, a classifier combining k-Nearest Neighbor (kNN) and Decision Tree algorithms is used for achieving gesture recognition. For gesture-based control application, a real-time interactive system has been built up for household appliance using 10 kinds of hand gestures as control commands. Our proposed framework facilitates intelligent and natural control in gesture-based interaction.