The Reaserch Of Fall Recognition Based On Surface EMG

Fall protection and early warning is a hot topic in today’s domestic andinternational research. This study can reduce the injuries from falls in the elderly. So itis a significant work. The study is aim to identify fall behavior and activities of dailylife, so that alarm or activate protective devices to avoid or reduce the damage causedby falls.Surface EMG (Surface electromyography, sEMG) which is recorded from thesurface of skeleton muscle by electrodes is the bioelectricity discharged byneuromuscular activities. It is a nonlinear, non-stationary signals, to some extent itreflects the functional state of nerve and muscle, and has been widely used in manyareas such as clinical diagnosis, sports medicine and so on. Although the fallidentification based on EMG is immature, as an attempt surface EMG is used for thedouble signal sources.Some research on the identification of EMG falls was done based on EMG andEEG. Many theory exploration and practice were done on the acquisition and featureextraction of EMG, pattern classification based on EMG. In the paper, the followingwork and innovations were made.(1) From the generation mechanism of EMG, the paper summarizes thecharacteristics of EMG, common interference in the acquisition of EMG, and Studiesthe methods of EMG pickup. In addition, studies the lead methods of electrodes andacquisition of EEG, analysis the human lower limb muscle structure andcharacteristics, explains how to select a typical muscles as the signal sourceidentification fall, as well as acquisition and preliminary analysis of lower limbsurface EMG, surface EMG signal de-noising, feature extraction and patternrecognition classification for reference.(2) De-noising method based on ensemble empirical mode decomposition(EEMD) and second generation wavelet transform is applied in EMG signalde-noising in the paper for the first time. The method combines the advantages of bothensemble empirical mode decomposition and second generation wavelet transformand discusses in detail the feasibility of this method for the EMG signal. Theexperimental results show that this method not only preserves the active ingredients ofa good signal, but also removes most of the EMG noise, get a more ideal noise-canceling effect, creates a good condition for feature extraction of surfaceEMG.(3) To extract the valid feature of EMG, feature extraction methods through thepermutation entropy based on EMG and EEG are studied. The concrete methodsinclude the feature extraction based on empirical mode decomposition permutationentropy, for six motion patterns (palmar dorsiflexion and flexion, hand extension andgrasp), four-channel EMG on tibialis anterior muscle, rectus femoris, vastus andmedial gastrocnemius. It decomposes the sEMG signal into a set of intrinsic modefunctions (IMF), then combines some of the IMF which contains the usefulinformation according to frequency effectiveness, and calculates the permutationentropy of the combination input support vector machine for classification, themethod achieved high recognition rate.(4) In order to better identify the fall, the study input EMG feature vectors toK-means clustering classifier, conducted nuclear spindle falls recognition basedclustering and support vector machine classifier, and the results show that supportvector machines can better recognition falls, has a higher recognition rate, whichsensitivity reached87.5%, and specificity reached96.5%.