Objective Alzheimer?? s disease AD lacks effective treatments causing significant harm to patients and society. To address the limitations of existing methods this paper proposed an innovative approach to detect Alzheimer?? s disease using long-term visual features. Methods First 17 human keypoint coordinates were extracted from videos of individuals walking in their daily environments covering key regions such as palms arms shoulders ankles knees hips torsos and heads. These keypoints were combined into a dataset with long-term visual features and the dataset was fed into a sequential neural network for training. Results The experimental results showed that the proposed method performed well in Alzheimer?? s disease detection with a final detection accuracy of 0. 96 and an F1 score of 0. 93. This demonstrated the superior performance of the method and further emphasized the importance of long-term visual features in Alzheimer?? s disease detection. Conclusion The proposed method provides an effective approach to Alzheimer?? s disease detection filling the gap in current diagnostic options. Meanwhile the experimental results highlight the significant role of temporal processing networks in this task. Future studies should explore and optimize the structure and function of temporal processing networks in greater depth to further improve the accuracy and reliability of Alzheimer?? s disease detection. This innovative approach offers a new direction for the early diagnosis and treatment of Alzheimer?? s disease and holds promise for research and applications in related fields.