A novel feature extraction method for handwritten numeral recognition is proposed based on character's geometric structures. A group of stable and reliable global features are defined and extracted. Further, a floating feature detector is proposed to detect and extract tiny segments as fine features. A neural network is employed as the recognizer to conduct experiments on evaluating the feasibility of the new approach. This proposed method demonstrates that the combination of fine features with global features can greatly improve handwritten character recognition rate compared to those of merely global features being used.