运动是人类赖以生存的基本能力。中国社会的快速老龄化加剧了老年人群运动功能衰退研究的相对欠缺。发展一套精确描述个体运动表现和运动能力的研究方法，可以为老年人下肢步行能力丧失、神经退行性疾病面部障碍等运动功能衰退的溯源和改善提供基础，具有重要的研究价值和社会意义。基于以上问题，本文主要开展了以下工作： 针对帕金森病患者面部运动障碍的诊断需求与隐私保护的矛盾，本文提出了面部运动特征不变原则，发展了提取面部运动特征并作身份投射的深度学习方法，建立了相应的综合评价体系，实现了保留医疗诊断信息的人脸隐私保护。构建了基于面部动态相似度模块、身份提取模块和面容替换模块的身份投射模型，解耦了帕金森病患者的面部运动特征与面容身份，实现了面部运动特征不变的身份投射重建，有助于深入理解面部运动障碍，为个体面部运动障碍的评估提供基础。 帕金森病患者是运动功能障碍的特例，而健康老年群体普遍面临下肢行走能力衰退。针对其运动学衰退规律，本文通过红外光学运动捕捉与逆运动学算法，建立了中国健康老年步态与健康年轻人步态的下肢运动数据集，填补了相关空白。相比年轻对照，老年人群的步幅、步速等指标下降，髋、膝关节更屈，踝除同侧触地外更背屈，膝、踝的活动范围缩小，对侧触地时刻髋关节变为伸向运动，同侧触地时刻髋、膝关节的运动由减速变为提速，反映了老年人步态的运动学特征变化。本文也将此方法推广应用于脊髓损伤截瘫患者神经调控步态研究，为其重建健康步态提供运动学依据。 为了研究老年人群步态动力学特征变化，本文通过肌肉骨骼模型模拟不同步态运动，实现从运动学特征推断更为本质的肌肉力量与控制特征。采用计算肌肉控制法计算不同步态下肢肌肉的发力、激活和控制模式差异，也发展了深度强化学习方法研究步态动力学。研究发现老年人步态需要的高伸髋力量和高趾屈力量可能超过肌力上限，偏离能量最低原则、增加能耗，推动了对老年人步行能力衰退动力学特征变化的理解。 本文通过面部运动特征不变性研究、光学动捕步态特征分析和肌肉骨骼模型步态动力学特征分析，实现了更精确的面部与下肢步行运动描述与分析，加深了对老年人运动障碍的理解，为延缓老年人运动能力衰退提供了现实依据。

Mobility is a basic ability for humans. The rapid aging of China has exacerbated therelative lack of research on the decline of elderly’s motor function. Developing researchmethods that accurately describe individual’s physical performance can provide a basisfor tracing and improving the decline of motor function in the elderly, such as lower limbwalking ability decline and neurological degenerative facial disorders, and has importantresearch value and social significance. Based on these problems, this thesis mainly carriedout the following work: In response to the contradiction between the diagnosis needs of facial movementdisorder in Parkinson’s disease patients and privacy protection, this work proposes theprinciple of invariant facial movement features, developed a deep learning method forextracting facial movement features and identity projection, established a correspondingcomprehensive evaluation system, and achieved facial privacy protection while retain-ing medical diagnostic information. This thesis constructed an identity projection modelbased on facial dynamic similarity module, identity extraction module, and facial iden-tity replacement module, decoupling the facial movement features and facial identity ofParkinson’s disease patients, and realized identity projection reconstruction with invari-ant facial movement features. It helps to deepen the understanding of facial movementdisorders and provides a basis for individual facial movement disorder evaluation. Parkinson’s disease patients are a special case of motor dysfunction, while the elderlyhealthy population generally faces a decline in lower limb walking ability. In response totheir kinematic decline pattern, this thesis established a lower limb movement databasefor the gait of Chinese healthy elderly and young people through infrared optical mo-tion capture and inverse kinematics algorithm, filling in relevant gaps. This work foundthat, Compared with young controls, the elderly population had decreased stride length,walking speed, increased hip and knee flexion, increased ankle dorsiflexion except foripsilateral heel strike, decreased range of motion of the knee and ankle, shifted hip to ex-tending motion at contralateral heel strike, and shifted the hip and knee to accelerationat ipsilateral heel strike, reflecting the changes in the kinematic features of elderly gait .This method was also applied to the study of gait in a spinal cord injury paraplegic patientunder neuromodulation, providing a kinematic basis for his reconstruction of healthy gait. To study the changes in dynamic characteristics of gait in the elderly population,this thesis simulated different gait movements using a musculoskeletal model to infer themore essential muscle strength and control characteristics from the kinematic features.This thesis used the computated muscle control method to calculate the differences inforce, activation, and control of the lower limb muscles in different gait movements, anddeveloped a deep reinforcement learning method to study the dynamic characteristics ofgait. This study found that the high hip extension force and high plantar flexion forcerequired for elderly gait may exceed the muscle strength limit, deviate from the principleof minimum energy, increase energy consumption, and could promote the understandingof the dynamic features of gait decline in the elderly. Through the study of invariant facial movement features, optical motion capturedgait feature analysis, and musculoskeletal model gait dynamic feature analysis, this thesisachieved an accurate description and analysis of facial and lower limb walking move-ments, deepening the understanding of elderly motor dysfunction and providing a basisfor slowing down the decline of elderly mobility.