改进AlexNet的细粒度汉字手势识别研究
Fine-Grained Chinese Character Gesture Recognition Using Improved AlexNet

射频识别在人机交互中的应用不断拓展，但现有方法在细粒度手势区分、跨场景适应性和模型轻量化方面仍存在不足。为此，提出一种改进AlexNet融合注意力机制的细粒度汉字手势识别方法。该方法通过标签矩阵与双天线融合增强信号采集完整性，结合非视距信号基线修正与马尔可夫转移场映射强化时序特征表达，并在轻量化卷积神经网络中嵌入注意力模块，在降低计算复杂度的同时保持较高识别准确率。实验结果显示，系统总体识别准确率为98.88%，跨用户与跨场景平均准确率分别为97.49%和97.74%。研究表明，该方法在细粒度汉字手势识别中具有较强鲁棒性与泛化能力，对智慧课堂人机交互及教育信息化发展具有应用价值。
Radio Frequency Identification (RFID) has been increasingly applied in human-computer interaction (HCI), yet existing approaches still face challenges in fine-grained gesture differentiation, cross-scenario adaptability, and model lightweighting. To address these issues, this study proposes a fine-grained Chinese character gesture recognition method based on an improved AlexNet integrated with a Squeeze-and-Excitation (SE) attention mechanism. The method enhances signal acquisition integrity by combining label matrices with dual-antenna fusion, applies non-line-of-sight baseline correction and Markov Transition Field mapping to strengthen temporal feature representation, and embeds an attention module into a lightweight convolutional neural network to reduce computational complexity while maintaining high recognition accuracy. Experimental results demonstrate an overall recognition accuracy of 98.88%, with average accuracies of 97.49% and 97.74% under cross-user and cross-scenario conditions, respectively. These findings confirm that the proposed approach exhibits strong robustness and generalization in fine-grained Chinese character gesture recognition, and provides a reliable solution for smart classroom interaction and the advancement of educational informatization.