采用支持向量机评估老年人步态对称性的研究

Novel Approach for Evaluation of Elderly Gait Symmetry

采用支持向量机(SVM)评估老年人步态的对称性.将鉴别老年人下肢左、右两侧的步态模式相似性问题转化为二分类问题,通过识别老年人下肢左、右两侧的步态模式,确定其两侧步态模式相似性的差异,判断其步态的对称性.采集24名健康老年人下肢左、右两侧的步态数据,采用交叉验证方法评估SVM泛化能力,测试了多项式核、径向基核、线性核.结果表明,多项式核、径向基核的泛化能力优于线性核,基于多项式核的SVM识别左、右两侧老年人步态模式的分类正确率较高(88%),可有效地提取步态模式的非线性信息.SVM有望成为评估老年人步态对称性的一个有效的工具,有助于及早预防老年人跌倒和老年性疾病发生,提高老年人生活质量.

The application of support vector machine （SVM） for evaluation of the symmetry of elderly gait is investigated following the assumption that the determination of similarity between the lower limbs can be considered as a binary classification task. Adopting SVM to classify the right and left sides gait patterns of the elderly lower limbs during normal walking, the difference between the corresponding gait parameter curves can be determined effectively to assess the gait pattern symmetry. The bilateral kinetic gait data of 24 elderly participants are recorded by a force platform during normal walking. The cross-validation method is employed to evaluate the generalization ability of SVM, and three types of kernel functions （linear, polynomial and radial basis function） are validated. The experimental results demonstrate that SVM with polynomial kernel outperforms SVM with linear kernel and radial basis function in classification performance to identify the right and left sides gait patterns with an accuracy up to 88%. It is suggested that SVM serves as an efficient tool for evaluation of the symmetry of elderly gait to identify at-risk gaits in older population.