Nan Bu

A Hybrid Motion Classification Approach for EMG-Based Human–Robot Interfaces Using Bayesian and Neural Networks

In a human-robot interface, the prediction of motion, which is based on context information of a task, has the potential to improve the robustness and reliability of motion classification to control prosthetic devices or human-assisting manipulators. This paper proposes a task model using a Bayesian network (BN) for motion prediction. Given information of the previous motion, this task model is able to predict occurrence probabilities of the motions concerned in the task. Furthermore, a hybrid motion classification framework has been developed based on the BN motion prediction. Besides the motion prediction, electromyogram (EMG) signals are simultaneously classified by a probabilistic neural network (NN). Then, the motion occurrence probabilities are combined with the NN classifiers outputs to generate motion commands for control. With the proposed motion classification framework, it is expected that classification performance can be enhanced so that motion commands can be more robust and reliable. Experiments have been conducted with four subjects to demonstrate the feasibility of the proposed methods. In these experiments, forearm motions are classified with EMG signals considering a cooking task. Finally, robot manipulation experiments were carried out to verify the proposed human interface system with a task of taking meal. The experimental results indicate that the proposed methods improved the robustness and stability of motion classification.

Monitoring of Respiration and Heartbeat during Sleep using a Flexible Piezoelectric Film Sensor and Empirical Mode Decomposition

Cardiorespiratory monitoring during sleep is one of the basic means for assessment of personal health, and has been widely used in diagnosis of sleep disorders. This paper proposes a novel method for non-invasive and unconstrained measurement of respiration and heartbeat during sleep. A flexible piezoelectric film sensor made of aluminum nitride (AlN) material is used in this study. This sensor measures pressure fluctuation due to respiration and heartbeat on the contact surface when a subject is lying on it. Since the AlN film sensor has good sensitivity, the pressure fluctuation measured can be further separated into signals corresponding to respiration and heartbeat, respectively. In the proposed method, the signal separation is achieved using an algorithm based on empirical mode decomposition (EMD). Experiments have been conducted with three subjects. The experimental results show that respiration and heartbeat signals can be successfully obtained with the proposed method.

EMG-Based Motion Discrimination Using a Novel Recurrent Neural Network

This paper presents a pattern discrimination method for electromyogram (EMG) signals for application in the field of prosthetic control. The method uses a novel recurrent neural network based on the hidden Markov model. This network includes recurrent connections, which enable modeling time series, such as EMG signals. Weight coefficients in the network can be learned using a well-known back-propagation through time algorithm. Pattern discrimination experiments were conducted to demonstrate the feasibility and performance of the proposed method. We were able to successfully discriminate forearm motions using the EMG signals, and achieved considerably high discrimination performance compared with other discrimination methods.

LMI based neurocontroller for guaranteed cost control of discrete-time uncertain system

This paper investigates the application of neural networks to the guaranteed cost control problem of discrete time uncertain system. Based on the linear matrix inequality (LMI) design approach, a class of a state feedback controller is established, and the sufficient conditions for the existence of guaranteed cost controller are derived by making use of the LMI. The novel contribution is that the neurocontroller is substituted for the additive gain perturbations. It is newly shown that although the neurocontroller is included in the discrete-time uncertain system, the robust stability for the closed-loop system and the reduction of the cost performance are attained.

FPGA implementation of a probabilistic neural network for a bioelectric human interface

Since a probabilistic neural network (PNN) provides a stochastic perspective of pattern discrimination, it has been proven to be efficient for complicated data such as bioelectric signals. As for practical implementation, however, a general-purpose computer is usually necessary, so that a compact design of an application system is difficult to be realized. This paper describes a field programmable gate array (FPGA) implementation of a PNN, with which system on chip (SoC) design of a bioelectric human interface device becomes possible. Its effectiveness is then verified with a practical application, and it is shown that the hardware implementation provides comparable performance with the software solution on a general-purpose computer.

Measuring muscle movements for human interfaces using a flexible piezoelectric thin film sensor

This paper proposes a novel method to measure muscle movements for human interfaces. During muscle movements, cross-sectional muscle area changes, and this can be detected at the skin surface. In this study, a flexible piezoelectric thin film sensor is used to measure the morphological change of the skin surface. This sensor is made of oriented aluminum nitride (AlN) thin film, and the total thickness is less than 40 μm. Since the AlN film sensor has good sensitivity, small strain of the skin surface can be measured. Furthermore, a motion classification method is developed to investigate the potential of the proposed sensor for its use in human interfaces. Response characteristics of the AlN sensor were tested with experiments using a cantilever beam. In addition, motion classification experiments were conducted with five subjects, including a patient with cervical spine injury. The experimental results validate the effectiveness of the proposed method.

Phoneme Classification for Speech Synthesiser using Differential EMG Signals between Muscles

This paper proposes the use of differential electromyography (EMG) signals between muscles for phoneme classification, with which a Japanese speech synthesiser system can be constructed using fewer electrodes. In distinction from traditional methods using differential EMG signals between bipolar electrodes on the same muscle, an EMG signal is derived as differential between monopolar signals on two different muscles in the proposed method. Then, frequency-based feature patterns are extracted with filter banks, and classification of phonemes is realized by using a probabilistic neural network, which combines feature reduction and pattern classification processes in a single network structure. Experimental results show that the proposed method can achieve considerably high classification performance with fewer electrodes

Biomimetic Impedance Control of an EMG-Based Robotic Hand

The number of extremity amputations resulting from workplace mishaps, traffic accidents and other incidents has shown an increasing trend over time, although the importance of safety management and the prevention of such accidents is fully recognized. Since precise and complex motion may be very difficult in the daily activities of amputees, the development of prosthetic systems is necessary to support their lives and enable social integration. In particular, there is a mandatory requirement for the development of externally powered prosthetic hands with a natural feeling of control, since the role played by this part of the body is very important. However, the control of such hands is problematic, and they must be carefully designed in line with the amputee’s remaining functions. Many researchers have designed prosthetic limbs for amputees since the concept was proposed by N. Wiener in Cybernetics [1]. In previous research, electromyograms (EMGs) have been widely used as an interface tool for prosthetic hands because EMG signals contain information about the operator’s intended motion [2] – [8]. For example, an EMGprosthetic hand made in the USSR [2], the Waseda hand developed by Kato et al. [3], the Boston arm by MIT [4] and the Utha artificial arm by Jacobson et al. [5] were all pioneering steps in the field. Since EMG signals also include information on the force level and mechanical impedance properties of limb motion, Akazawa et al. [6] estimated the force of flexors and extensors from these signals and proposed a scheme to use them in controlling a prosthetic hand. Abul-haj and Hogan [8] also proposed prosthetic control based on an impedance model and analyzed its control characteristics. Most previous research, however, dealt only with on/off control for prosthetic arms depending on the results of EMG pattern discrimination [2], [3], [7], or controlled only a particular joint depending on the torque estimated from EMG signals [4], [5], [6], [8]. Multijoint control of prosthetic arms considering the variable viscoelasticity of flexors and extensors has not yet been realized. This chapter introduces a biomimetic control for an externally powered multi-joint prosthetic hand that considers the muscular contraction levels of flexors and extensors using 9

LMI based neurocontroller for output-feedback guaranteed cost control of discrete-time uncertain system

This paper presents a nonlinear output feedback controller design method that integrates the guaranteed cost control approach for a class of discrete-time system with parametric uncertainties and neural networks (NNs). Based on the linear matrix inequality (LMI) design approach, a class of output feedback controller is established, and some sufficient conditions for the existence of guaranteed cost controller is derived. The novel contribution is that the neurocontroller is substituted for the additive gain perturbations. Although the neurocontroller is included in the uncertain system, the closed-loop system is asymptotically stable and the closed-loop cost function value is not more than specified upper bound for all admissible uncertainty. A numerical example is given to illustrate the computational efficiency of the proposed method.

A preliminary study on detection of muscle activity using a flexible AlN piezoelectric thin film sensor

Muscle activity can be used to generate control signals for prosthetic devices and human-assisting manipulators. Recently, a novel method has been developed to measure muscle movement using a flexible piezoelectric thin film sensor, which is made of oriented aluminum nitride (AlN) film. However, rigorous relationship between muscle activity and the sensors output has not been clarified. As a preliminary study, this paper proposes a simple model to illustrate the sensor output according to limbs cross-sectional dimensional change. Then, experiments have been carried out for comparison between muscle movement measurement methods using the AlN film sensor and ultrasound imaging. In the experiments, simultaneous measurement of signals of the AlN film sensor and ultrasonography was conducted during muscle actions. Both isokinetic and isometric muscle contractions were considered in the experiments. Since Youngs modulus of the AlN film sensor is significantly larger than those of skin, panniculus adiposus, and muscle layer, precise dimensional changes in muscles cannot be obtained from the proposed AlN film sensor based on a direct sensor attachment method. From the experimental results, however, it is found that times of contraction and relaxation indicated by the AlN film sensor agree well with the results obtained from ultrasonography. It is considered that the timing information of muscle activity can be detected with the AlN film sensor.