多指标驱动下自适应小波变换的下肢假肢运动意图识别OA
Lower Limb Prosthesis Motion Intention Recognition Based on Adaptive Wavelet Transform Driven by Multiple Indicators
在基于小波变换的下肢假肢意图识别中,小波分解层数通常依赖人工经验选择进行设定,而不同动作类型的特征复杂度之间差异较大,人工经验选择的固定分解层数在提取所有运动特征方面存在局限性.因此,文章提出了一种多指标驱动下的自适应小波变换算法,结合能量变化、小波熵、峰值比等多个指标对运动信号特性进行综合评估.针对不同动作下的最优分解层数不同从而引发的特征向量长度维度不一致的问题,本研究引入基于三次样条插值的特征对齐机制,保留信号关键趋势的同时统一特征维度.最终,使用支持向量机进行分类.实验结果表明,投票策略增强了最优小波分解层数算法决策的鲁棒性,有效抑制了不同动作和异常数据对分解层数的影响.该方法对自采集数据集中的13种综合运动模式的识别率可以达到96.82%,提高了模型整体的识别率和鲁棒性.
In lower limb prostheses intention recognition based on wavelet transform,the number of wavelet decomposi-tion layers is usually empirically.However,due to the significant differences in feature complexity among different action types,the fixed number of decomposition layer selected empirically is limited in extracting all motion features.Therefore,this paper proposes an adaptive wavelet transform algorithm driven by multiple indicators,which combines multiple indicators such as energy change,wavelet entropy,and peak ratio to comprehensively evaluate the characteristics of the motion signal.To address the inconsistency in feature vector dimensions caused by varying optimal numbers of decomposition layers for differ-ent actions,a feature alignment mechanism based on cubic spline interpolation is introduced,which retains the key trends of the signal while unifying the feature dimensions.Finally,support vector machines are used for classification.Experimental re-sults show that the voting strategy enhances the robustness of the decision-making of the optimal wavelet decomposition layer algorithm and effectively suppresses the influence of different actions and abnormal data on the number of decomposition lay-ers.This method can achieve a recognition rate of 96.82%in 13 comprehensive motion modes of self-collected data sets,im-proving the overall recognition rate and robustness of the model.
宗文杰;陈欣;刘文瑶;罗彪;苏本跃
安庆师范大学 人工智能与计算机学院,安徽 安庆 246133||铜陵学院 数学与计算机学院,安徽 铜陵 244061安庆师范大学 数学与统计学院,安徽 安庆 246133安庆师范大学 人工智能与计算机学院,安徽 安庆 246133||铜陵学院 数学与计算机学院,安徽 铜陵 244061安庆师范大学 数学与统计学院,安徽 安庆 246133铜陵学院 数学与计算机学院,安徽 铜陵 244061
信息技术与安全科学
下肢假肢意图识别自适应小波变换层数多指标决策特征对齐三次样条插值
lower limb prosthesis intention recognitionadaptive wavelet transform decomposition levelmulti-index
《安庆师范大学学报(自然科学版)》 2026 (1)
81-88,8
安徽省高校优秀科研创新团队项目(2023AH010056)和铜陵学院联合培养研究生创新基金项目(23tlcx01,24tlcb01)
评论