Muye Pang

Comparison of sEMG-Based Feature Extraction and Motion Classification Methods for Upper-Limb Movement

The surface electromyography (sEMG) technique is proposed for muscle activation detection and intuitive control of prostheses or robot arms. Motion recognition is widely used to map sEMG signals to the target motions. One of the main factors preventing the implementation of this kind of method for real-time applications is the unsatisfactory motion recognition rate and time consumption. The purpose of this paper is to compare eight combinations of four feature extraction methods (Root Mean Square (RMS), Detrended Fluctuation Analysis (DFA), Weight Peaks (WP), and Muscular Model (MM)) and two classifiers (Neural Networks (NN) and Support Vector Machine (SVM)), for the task of mapping sEMG signals to eight upper-limb motions, to find out the relation between these methods and propose a proper combination to solve this issue. Seven subjects participated in the experiment and six muscles of the upper-limb were selected to record sEMG signals. The experimental results showed that NN classifier obtained the highest recognition accuracy rate (88.7%) during the training process while SVM performed better in real-time experiments (85.9%). For time consumption, SVM took less time than NN during the training process but needed more time for real-time computation. Among the four feature extraction methods, WP had the highest recognition rate for the training process (97.7%) while MM performed the best during real-time tests (94.3%). The combination of MM and NN is recommended for strict real-time applications while a combination of MM and SVM will be more suitable when time consumption is not a key requirement.

Electromyography-Based Quantitative Representation Method for Upper-Limb Elbow Joint Angle in Sagittal Plane.

This paper presents a quantitative representation method for the upper-limb elbow joint angle using only electromyography (EMG) signals for continuous elbow joint voluntary flexion and extension in the sagittal plane. The dynamics relation between the musculotendon force exerted by the biceps brachii muscle and the elbow joint angle is developed for a modified musculoskeletal model. Based on the dynamics model, a quadratic-like quantitative relationship between EMG signals and the elbow joint angle is built using a Hill-type-based muscular model. Furthermore, a state switching model is designed to stabilize the transition of EMG signals between different muscle contraction motions during the whole movement. To evaluate the efficiency of the method, ten subjects performed continuous experiments during a 4-day period and five of them performed a subsequent consecutive stepping test. The results were calculated in real-time and used as control reference to drive an exoskeleton device bilaterally. The experimental results indicate that the proposed method can provide suitable prediction results with root-mean-square (RMS) errors of below 10° in continuous motion and RMS errors of below 10° in stepping motion with 20° and 30° increments. It is also easier to calibrate and implement.

A surface EMG signals-based real-time continuous recognition for the upper limb multi-motion

This paper was aimed at the continuous recognition of the upper limb multi-motion during the upper limb movement for rehabilitation training. The amplitude of the surface electromyographic (sEMG) signals change during movement of the upper limb and the features of sEMG signals are different with the changes. These variances in the features represent the different statuses of the upper limb. Recognizing the variances will lead to recognition of the upper limb motion. In this study, sEMG signals were recorded through five noninvasive electrodes attached on the anatomy points of the upper limb and an autoregressive model was used to extract the features of the detected sEMG signals. After that the Back-propagation Neural Networks was applied to recognize the patterns of the upper arm motion using the variant features as the training and input data. Three volunteers participated in the real-time experiment and the results stated that this method is effective for a real-time continuous recognition of the upper limb multi-motions.

Recognition of motion of human upper limb using sEMG in real time: Towards bilateral rehabilitation

The surface electromyographic (sEMG) signal has been researched in many fields, such as medical diagnoses and prostheses control. In this paper, recognition of motion of human upper limb by processing sEMG signal in real time was proposed for application in bilateral rehabilitation, in which hemiplegia patients trained their impaired limbs by rehabilitation device based on motion of the intact limbs. In the processing of feature exaction of sEMG, Wavelet packet transform (WPT) and autoregressive (AR) model were used. The effect of feature exaction with both methods was discussed through the processing of classification where Back-propagation Neural Networks were trained. The experimental results show both methods can obtain reliable accuracy of motion pattern recognition. Moreover, on the experimental condition, the recognized accuracy of WPT is higher than that of AR model.

Design of a master-slave rehabilitation system using self-tuning fuzzy PI controller

Many robotic devices have been developed for stroke patients to recover their upper limb motor function. Among them, master-slave type rehabilitation systems provide surveillance of the therapist to the patient who is performing home-rehabilitation. In this study, we proposed a wearable and light exoskeleton device for upper limb rehabilitation and designed a master-slave rehabilitation system using the exoskeleton device as slave device and a haptic device (Phantom Premium) as master device. To convey therapists experience to patients using this system, the slave device is driven to track the motion of the master device manipulated by the therapist. In order to improve the tracking efficacy of traditional PI control, a self-tuning fuzzy PI control was proposed. Results of simulation indicated the proposed control method is more effective than the traditional PI control, particularly in tracking accuracy and response speed.

Study on recognition of upper limb motion pattern using surface EMG signals for bilateral rehabilitation

Surface electromyographic signal (sEMG) is deep related with the activation of motor muscle and motion of human body, which can be used to estimate the intention of the human movement. So it is advantaged in the application of bilateral rehabilitation, where hemiplegic patients can perform rehabilitation training to their impaired limbs following the motion of intact limbs by using a certain training tool. In this paper, we discussed the motion pattern recognition of human upper limb based on the sEMG signals. The main features of motion patterns based on sEMG signals are extracted via wavelet packet transform. Because the sEMG signal is a kind of non-stationary signal and there are many factors which can affect it like inherent noise, cross talk and so on. Therefore, a simple new method to obtain the trend of sEMG with weighted peaks as features was proposed and support vector machine (SVM) is utilized as the classifier. The contrastive experimental results show that the proposed method improved the recognition rate.

ULERD-based active training for upper limb rehabilitation

In this paper, we proposed a control method to implement the upper-limb active training which is performed with the proposed exoskeleton device. It provides a wide approach for Human Machine Interface (HMI) in which the device is of high inertia, high friction and non-backdrivability and it is difficult to obtain the contact force between human and the device directly. The main idea of this method is to measure the motion of human body rather than the motion of device. This method is more suitable to the HMI in which the contact between human and device can be assumed as a spring-damper model. According to two kinds of experiments designed, different contact resistance was exerted to the forearm of the user. The sEMG signals detected from biceps brachii and triceps brachii were processed and the two kinds of resistance exerted to human forearm were confirmed.

Development of a Bilateral Rehabilitation Training System Using the Haptic Device and Inertia Sensors

According to the neuro-rehabilitation theory, passive, resistance and bilateral training are commonly applied for recovering the motor-function of stroke patient. Among them, bilateral training is proved to be an effective method for the hemiparesis that occupies most part of stroke patients. In this article, a novel system is proposed for providing the bilateral training with coordinative motion of two limbs. This system is developed for the elbow function recovery and the motion of two limbs is detected with two inertia sensors. A commercial haptic device (Phantom Premium) is adopted for providing a feedback with information of errors and how to correct them. Combined with a graphic interface which provides a visual feedback, the patient can adjust the two limbs to a coordinative motion. This system can perform the training to those patients with some muscle strength. However, usually the rehabilitation training is hierarchical and those patients with little muscle strength can even not lift their own limbs. Therefore, a light-weight exoskeleton device is applied and this device could provide partial assisting force, thus the patient can gradually adapt to the training. In this article, an issue about the effectiveness of feedback is discussed and verified with several contrast experiments.

Finger joint continuous interpretation based on sEMG signals and muscular model

The human hand is very dexterous and can perform various of gestures in activities of daily living. Only dividing the motions of hand into several types and applying pattern recognition method for implementation of manipulation control may result in low dexterity and delicacy. In this paper, a novel finger joint interpretation method based on sEMG signals and muscular model is presented. The motion of finger is flexion and extension without any external resistant force and at a natural movement velocity. sEMG signals are recorded from flexor digitorum superficialis and extensor digitorum of the forearm. The Hill-based muscular model is used to calculate the force of muscles according to sEMG signals. In this paper, we assume that the changing of force corresponds directly to the motions of fingers given the circumstance that the subjects hold nothing in their hand and keep the movement velocity. The curve fitting method and Kalman filter are implemented to calculate the relation between force and basic movements of digits. Five subjects participated in the experiment to evaluate the efficiency of this method.

Preliminary study on upper limb movement identification based on sEMG signal

Stroke has become a very prevalent disease, especially in elder people. Many researches have focused on developing advanced and intelligent robotic system to assist the treatment of patients. For this field, Electromyography (EMG) is widely used for its benefit to get valuable information about the neuromuscular activity of a muscle. In this paper, wavelet packet decomposition method which is a kind of time-frequency domain is used for movement identification. Appropriate coefficients between three important movements for ADLs and sEMG signal will be extracted with wavelet packet decomposition method. These coefficients could be used as the input of BP neural network for movement identification. Experimental results proved that this method is effective off-line. Whereas the on-line identification rate should be improved in the future works.