Search

In recent years, sensor-based human activity recognition (HAR) has gained tremendous attention around the world with a range of applications. Instead of using body sensor network-based recognition systems which are intrusive and increase equipment cost, we focus on the development of efficient HAR approach based on a single triaxial accelerometer. In order to improve the recognition accuracy of the system, a novel recognition approach based on kernel discriminant analysis (KDA) and quantum-behaved particle swarm optimization-based kernel extreme learning machine (QPSO-KELM) is proposed. KDA is utilized to extract more meaningful features and enhance the discrimination between different activities. To verify the effectiveness of KDA, three kinds of features including original features, linear discriminant analysis (LDA) features and KDA features are extracted and compared for activity recognition. In addition, QPSO-KELM is compared with two existing classification methods: support vector machine (SVM) and extreme learning machine (ELM), which are commonly utilized in activity recognition. Meanwhile, two comparative optimization methods for KELM are also discussed in the experiment. The experimental results demonstrate the superiority of the proposed approach.

Sensor-based human activity recognition (HAR) has attracted interest both in academic and applied fields, and can be utilized in health-related areas, fitness, sports training, etc. With a view to improving the performance of sensor-based HAR and optimizing the generalizability and diversity of the base classifier of the ensemble system, a novel HAR approach (pairwise diversity measure and glowworm swarm optimization-based selective ensemble learning, DMGSOSEN) that utilizes ensemble learning with differentiated extreme learning machines (ELMs) is proposed in this paper. Firstly, the bootstrap sampling method is utilized to independently train multiple base ELMs which make up the initial base classifier pool. Secondly, the initial pool is pre-pruned by calculating the pairwise diversity measure of each base ELM, which can eliminate similar base ELMs and enhance the performance of HAR system by balancing diversity and accuracy. Then, glowworm swarm optimization (GSO) is utilized to search for the optimal sub-ensemble from the base ELMs after pre-pruning. Finally, majority voting is utilized to combine the results of the selected base ELMs. For the evaluation of our proposed method, we collected a dataset from different locations on the body, including chest, waist, left wrist, left ankle and right arm. The experimental results show that, compared with traditional ensemble algorithms such as Bagging, Adaboost, and other state-of-the-art pruning algorithms, the proposed approach is able to achieve better performance (96.7% accuracy and F1 from wrist) with fewer base classifiers.

Sensor-based human activity recognition can benefit a variety of applications such as health care, fitness, smart homes, rehabilitation training, and so forth. In this paper, we propose a novel two-layer diversity-enhanced multiclassifier recognition method for single wearable accelerometer-based human activity recognition, which contains data-based and classifier-based diversity enhancement. Firstly, we introduce the kernel Fisher discriminant analysis (KFDA) technique to spatially transform the training samples and enhance the discrimination between activities. In addition, bootstrap resampling is utilized to increase the diversities of the dataset for training the base classifiers in the multiclassifier system. Secondly, a combined diversity measure for selecting the base classifiers with excellent performance and large diversity is proposed to optimize the performance of the multiclassifier system. Lastly, majority voting is utilized to combine the preferred base classifiers. Experiments showed that the data-based diversity enhancement can improve the discriminance of different activity samples and promote the generation of base classifiers with different structures and performances. Compared with random selection and traditional ensemble methods, including Bagging and Adaboost, the proposed method achieved 92.3% accuracy and 90.7% recall, which demonstrates better performance in activity recognition.

Purpose This paper aims to find a new method that could be applied to the high and mid-grade prosthesis knee joint. Design/methodology/approach Based on analysis, calculation, modeling, simulation and experimental study of the motion law of knee joint, this paper not only determines the structure and parameters of the knee joint and calculates the instantaneous center but also analyzes the stance stability and completes the optimization. With the help of experimental tests (fatigue test and gait curve test), the quality and performance of the designed knee joint is verified. Findings The experimental results show that the gait curve of the designed knee joint is much closer to health people. The designed prosthesis knee joint, with adjustable swing speed and gait curve which are close to health limb, has a better performance when compared to the ordinary knee joint with four-bar linkage structure. Originality/value This paper developed a prosthesis knee joint based on a novel design method that could be applied to the “high and mid” grade prosthesis knee joint and verified its function on an amputee performed the lower amputation, which could provide theoretical support for researches and designs related to prosthesis knee joint in future.

The automatic classification of epileptic electroencephalogram (EEG) is important for the diagnosis and treatment of epilepsy. In this paper, an epileptic EEG classification method based on wavelet multi-scale analysis and particle swarm optimization is proposed. Firstly, the multi-scale is carried out to the original EEG to extract its sub-bands of different frequency. Secondly, the Hurst exponent and the sample entropy are used to extract the EEG signals and its sub-bands. Finally, the particle swarm optimization (PSO) algorithm is used to optimize the parameters of the extreme learning machine (ELM), and the obtained eigenvector is put to PSO-ELM to realize the purpose of classification of epileptic EEG. The proposed method in this paper achieved 99.7% classification accuracy for the discrimination between epileptic ictal and interictal EEG, which is superior to those methods in other studies.

In order to improve the performance of the base classifier in the process of AdaBoost algorithm and simplify the complexity of the whole ensemble learning system, this paper presents a SVM ensemble method based on an improved iteration process of Adaboost algorithm. The improved Adaboost algorithm is added with methods of adding sample selection and feature selection in its iterative process in order to solve the problem that Adaboost is susceptible to noise and has long training time. First of all, the samples subsets are selected by means of mean nearest neighbor algorithm. Secondly, the feature subset are obtained using the method of relative entropy. Lastly, the individual SVM classifiers are trained by the resulting optimal feature samples subset in each cycle and combined via majority vote to generate the final decision system. The simulation results of UCI datasets show that this algorithm can achieve a higher recognition accuracy on the basis of fewer samples and features compared with the traditional Adaboost support vector machine ensemble algorithm.

Prosthetic knee joint (PKJ) is an important apparatus for trans-femoral amputees to regain walking ability. This study has two objectives: (1) to design a high performance and low-cost passive PKJ and (2) to evaluate the performance of the PKJ. In the proposed PKJ design, a four-bar linkage was employed as the mechanical structure, and parallel spring and damper were used as the two connecting rods of the four-bar linkage. With the parallel spring, the length of the connecting rod is variable and a buffer flexion angle can be generated, which was consistent with that of the human knee joint. The damper was used to regulate the swing speed of the shank. Through theoretical analysis, modeling, and simulation, key parameters of the mechanical structure were optimized. Finally, experimental studies were conducted to test the performance of the PKJ, including fatigue test and gait analysis. The results showed that the designed PKJ is reliable and the gait of PKJ is close to the healthy subject. Moreover, it is comfortable and showed no adverse effects on the amputee during the walking experiment.

Inspired by the indoor localization technology of mobile robots, a novel Electronic Travel Aids (ETA) is developed based on multi-sensor fusion using an extended Kalman filter (EKF). A wearable sensor system is designed to estimate the real-time posture of user's lower limb. This system can be used both for monitoring the user's walking status and for realizing the relative localization. A belt of multiple sonar sensors worn by the user can detect the obstacle of unknown environment. Simultaneously, this sonar belt can also be used as the absolute localization device in a similar way as the mobile robot. An EKF is applied to realize the data fusion of different sensor systems. The abilities of obstacle avoidance and indoor localization of the proposed ETA are both verified through experiments.

A pneumatic drive device for testing and evaluating lower-limb prostheses was designed, which is based on the ISO 10328 requirements for structural testing of lower-limb prosthetic. It can measure the structural strength and fatigue life of the lower-limb prosthesis at both heel-strike and toe-off moments, during which the knee joint bears the largest stress. The device well meets the testing requirements according to the current operation.

This paper present a low-cost wearable ultrasound bladder volume measurement and alarm system based on ARM9 and DSP. The system used a new volume computing method to estimate bladder volume based on the ultrasound echo obtained by only one phased-array ultrasonic transducer. The main signal processing task of ultrasound echo are performed using a DSP chip. The estimated value of bladder volume was transmitted to an alarm system based on Zigbee radio link. The system can be used for continuous monitoring of bladder volume and nursing on unconsciousness elders, handicapped with spinal cord injury, other urological patients, and children with nocturnal enuresis through setting alarm threshold.

A passive above knee prosthesis being currently in use does not respond adequately to the needs of daily living activities of an amputee due to its limited functions. Therefore, this study deals with design of an active above knee prosthesis. The essential part of this prosthesis is the knee joint actuator based on active control of basic flexion and extension functions of the knee joint with the DC motor. The failure of a single sensor has catastrophic consequences on control. Small inertial measurement unit adhereing to the sound leg and two electrodes on stump detect the sound leg gesture and stump SEMG. Using echo control scheme for gait control, which modified knee trajectory from sound leg attitude value and SEMG switch signal.

Environmental control system for disabled people has been growing rapidly over the past few years. Environmental control systems not only give disabled persons greater independence, but also provide a way to participate in their life activities and the same time increase the quality of their life. In this paper, we combine different control technologies together to design an environmental control system which is easy to use, cost effective, and can be used by many kinds of disabled people.

For the optional control to artificial lower limb and natural gait, an artificial limb model with myoelectrical control was presented, and the recognition method based on support vector machine was discussed. The electromyography signal after pretreating and wavelet packet analyzing was stored in company with the motion of lower limb. Then support vector machine was used to build a model and found the relationship between electromyography signal and motion of leg. Finally, continuous angles of knee joint were used to control artificial limb device. Simulation results show that electromyography signal and angles of knee joint have strong relationship, and this algorithm obtains preferable forecast result

This paper proposes a new-type prosthesis model and control mechanism, represents an ongoing investigation of voluntary and natural control of lower limb prosthesis using the myoelectric signals. First of all, myoelectric signals, which are sampled from the lower limb during the subjects walked normally on the three kinds terrains, such as flat, slop and stair, are extracted the motion features with Wavelet packet decomposition. And then these features are sent into LVQ neural network to classify. In the end sending the new signals into the network, the high identification ratio was obtained. The greatest characteristic of the Robot-Prosthesis is that it can be controled at will, when the active control should offered corresponding parameter to the execution part for different terrain.