传感技术学报
傳感技術學報
전감기술학보
Chinese Journal of Sensors and Actuators
2015年
11期
1586-1590
,共5页
表面肌电信号%小波包分解%排列组合熵%支持向量机%粒子群算法
錶麵肌電信號%小波包分解%排列組閤熵%支持嚮量機%粒子群算法
표면기전신호%소파포분해%배렬조합적%지지향량궤%입자군산법
surface electromyography%wavelet packet decomposition%permutation entropy%support vector machine%particle swarm optimization
跌倒对老年人的健康构成严重危害，设计了一种基于肌电信号的跌倒识别方法，可用于跌倒检测报警。该方法首先对表面肌电动作信号进行小波包分解，再依据信号特征选取信号的低频分量并重构，计算其排列组合熵，最后以4路肌电信号对应的排列组合熵组成的特征向量输入SVM进行模式识别并采用粒子群算法对SVM中惩罚参数c和核函数参数g进行优化，对8种动作进行识别实验，跌倒识别率88%，特异度98.3%，平均识别率97.0%，优于网格法和遗传算法支持向量机（GA-SVM）的参数优化，具有较强的鲁棒性和抗干扰能力。
跌倒對老年人的健康構成嚴重危害，設計瞭一種基于肌電信號的跌倒識彆方法，可用于跌倒檢測報警。該方法首先對錶麵肌電動作信號進行小波包分解，再依據信號特徵選取信號的低頻分量併重構，計算其排列組閤熵，最後以4路肌電信號對應的排列組閤熵組成的特徵嚮量輸入SVM進行模式識彆併採用粒子群算法對SVM中懲罰參數c和覈函數參數g進行優化，對8種動作進行識彆實驗，跌倒識彆率88%，特異度98.3%，平均識彆率97.0%，優于網格法和遺傳算法支持嚮量機（GA-SVM）的參數優化，具有較彊的魯棒性和抗榦擾能力。
질도대노년인적건강구성엄중위해，설계료일충기우기전신호적질도식별방법，가용우질도검측보경。해방법수선대표면기전동작신호진행소파포분해，재의거신호특정선취신호적저빈분량병중구，계산기배렬조합적，최후이4로기전신호대응적배렬조합적조성적특정향량수입SVM진행모식식별병채용입자군산법대SVM중징벌삼수c화핵함수삼수g진행우화，대8충동작진행식별실험，질도식별솔88%，특이도98.3%，평균식별솔97.0%，우우망격법화유전산법지지향량궤（GA-SVM）적삼수우화，구유교강적로봉성화항간우능력。
A new fall detection method was designed for fall alarm based on sEMG. Firstly,the sEMG signals are de?composed into subspaces with wavelet packet. Then,depending on the signal characteristics,signals of low-frequen?cy component were recombined to calculate the permutation entropy. Finally,the SVM method was used to recog?nize eight actions according to the permutation entropy of four sEMG signals,and the particle swarm optimization was used to optimize punishment parameter c and nuclear parameter g . The result shows fall sensitivity,fall spec?ificity,the average recognition rate were 88%,98.3%,97.0%,better than the gird method and genetic algorithm pa?rameters optimization. The method has strong robustness and noise immunity.
