随着全球人口老龄化加剧，脑老化引发的认知与运动功能衰退已成为重大公共卫生挑战。传统筛查工具存在评估维度单一、成本高昂及基层适用性不足等缺陷，难以满足社区大规模筛查需求。为此，本研究提出基于运动-认知双任务范式的脑老化风险评估新方法，通过多模态数据融合与算法创新，构建高效、低耗的筛查系统。论文主要工作及成果如下： 设计了一套运动-认知评估系统，系统由运动跑台、多视角摄像头与认知交互平台集合而成，实现运动-认知双任务的同步量化。硬件端通过三摄像头阵列捕获三维步态视频，软件端设计语音交互的色词测试认知任务。创新性地提出阈值法结合隐马尔可夫模型（HMM）的步态时相划分算法，在正常步态中的平均识别准确率达到91.03%，异常步态（如中风患者）的步态相位分割优于传统方法，解决了复杂步态模式下的特征提取难题。 系统地采集了90例康养中心的中老年人的认知功能、身体机能以及双任务测试数据，测试项目包括蒙特利尔认知评估量表（MoCA）、主观认知下降量表（SCD-9）、老年抑郁量表（GDS）、肺功能测试以及运动-认知双任务测试，对身体机能、认知能力与量化的双任务指标之间的潜在关联进行了理论分析和验证。结果发现双任务得分与肺功能指标、MoCA评分显著相关（p<0.05）。通过标准化回归模型及Bootstrap中介效应检验方法揭示了年龄对上述关联的介导效应，为双任务评估的生物学合理性提供证据，并指明年龄是后续研究的核心变量。 构建了脑老化风险等级预测模型。对社区队列152例简化版本数据库进行分析，利用Spearman分析证实年龄与双任务得分、主观认知下降得分（SCD-9）、步速、动态认知得分、患慢性疾病个数、是否需要扶扶手等指标的强相关性。基于LASSO回归模型筛选关键指标：双任务得分、动态认知得分、步态对称性（髋关节左右差异率、步态周期左右差异率），构建生理年龄预测模型（MAE=7.58岁，R²=0.65），其预测年龄偏差（ΔAge≥5岁）可用于划分脑老化风险等级。最后，该系统仅需5分钟测试即可输出结果，成本为传统方法的1/20。 本研究通过“系统开发-机制解析-模型构建”的全链条创新，为脑老化早期筛查提供了高效解决方案。通过运动与认知功能的同步量化评估，构建的预测模型可精准识别高风险人群，为社区分级管理提供技术支撑。未来通过多中心验证与设备优化，有望成为老龄化社会健康管理的关键基础设施。

With the intensification of global population aging, cognitive and motor functional decline caused by brain aging has emerged as a major public health challenge. Traditional screening tools suffer from limitations such as unidimensional assessment, high costs, and poor accessibility in primary care settings, failing to meet the demands of large-scale community screening. This study proposes a novel brain aging risk assessment method based on a motor-cognitive dual-task paradigm, constructing an efficient and low-cost screening system through multimodal data fusion and algorithmic innovation. The main contributions and findings are outlined as follows: A motor-cognitive assessment system was designed, integrating a treadmill, multi-view cameras, and a cognitive interaction platform to synchronously quantify dual-task performance. Hardware components included a triple-camera array capturing three-dimensional gait videos, while software components featured voice-interactive cognitive tasks (e.g., Stroop color-word tests). Innovatively, a threshold method combined with Hidden Markov Model (HMM) was proposed for gait phase segmentation, achieving an average recognition accuracy of 91.03% in normal gait and outperforming traditional methods in abnormal gait analysis (e.g., stroke patients), effectively resolving feature extraction challenges in complex gait patterns. Systematic data collection was conducted from 90 older adults in a senior care center, encompassing cognitive assessments (Montreal Cognitive Assessment (MoCA), Subjective Cognitive Decline-9 (SCD-9), Geriatric Depression Scale (GDS)), pulmonary function tests, and motor-cognitive dual-task evaluations. Theoretical and empirical analyses revealed significant correlations between dual-task performance scores and pulmonary function indicators (p<0.05) as well as MoCA scores (p<0.05). Standardized regression models and bootstrap-mediated effect analyses demonstrated the mediating role of age in these associations, providing biological plausibility for dual-task assessment and identifying age as a critical variable for subsequent research. A brain aging risk stratification model was developed. Analysis of a simplified community cohort dataset (n=152) using Spearman correlation confirmed strong associations between age and dual-task scores, SCD-9 scores, gait speed, dynamic cognitive performance, chronic disease prevalence, and balance dependency. Key predictors—including dual-task scores, dynamic cognitive performance, and gait symmetry metrics (hip joint laterality difference rate, gait cycle asymmetry)—were selected via LASSO regression to construct a physiological age prediction model (MAE=7.58 years, R²=0.65). The predicted age deviation (ΔAge≥5 years) effectively stratified brain aging risk levels. The system delivers results within 5 minutes at 1/20 the cost of traditional methods. Through integrated innovations in "system development-mechanistic exploration-model construction," this study provides an efficient solution for early brain aging screening. By synchronously quantifying motor and cognitive functions, the predictive model enables precise identification of high-risk populations, offering technical support for community-based hierarchical healthcare management. Future multicenter validation and device optimization could position this system as critical infrastructure for aging society health management.