Zhiwei Luo

Age and gender differences in the control of vertical ground reaction force by the hip, knee and ankle joints

[Purpose] This study examined the relationships between joint moment and the control of the vertical ground reaction force during walking in the elderly and young male and female individuals. [Subjects and Methods] Forty elderly people, 65 years old or older (20 males and 20 females), and 40 young people, 20 to 29 years old (20 males and 20 females), participated in this study. Joint moment and vertical ground reaction force during walking were obtained using a 3D motion analysis system and force plates. Stepwise linear regression analysis determined the joint moments that predict the amplitude of the vertical ground reaction force. [Results] Knee extension moment was related to the vertical ground reaction force in the young males and females. On the other hand, in the elderly females, hip, ankle, and knee joint moments were related to the first peak and second peak forces, and the minimum value of vertical ground reaction force, respectively. [Conclusion] Our results suggest that the young males and females make use of the knee joint moment to control of the vertical ground reaction force. There were differences between the elderly and the young females with regard to the joints used for the control of the vertical ground reaction force.

Cervical spine simulation model for traction therapy analysis

This research describes the construction of a biomechanical simulation model of the cervical spine for traction therapy analysis. By building a 3D computer model of the musculoskeletal cervical spine region using a physics-based software environment, we were able to calculate and visualize the changes of the cervical vertebrae and intervertebral discs due to external traction applied during a traction therapy. Since the mechanical parameters of the simulation, such as bone sizes and masses, can be easily adjusted in the computer model, it allows us to match the simulation based on actual data from patient in the traction therapy.

An adaptive treadmill-style locomotion interface and its application in 3-D interactive virtual market system

The key issue in this paper is estimating speed of a human. Compared with previous researches on walking speed estimation, we predict the walking intention before gait action. Our proposed hypothesis is that a composite force index is linearly correlated with the intended walking speed. We did two experiments to test the hypothesis. One gives a regression test indicating the intended walking speed has strong linear correlation with the proposed force index; the other tests the linearity by statistical analysis, guaranteeing the tolerance of individual difference. According to the regression and statistics analyses, we built a treadmill-style locomotion interface. Compared with the normal cases of treadmill control, the tested subject does not have to follow the speed of treadmill, but can actively change the speed of treadmill by his/her feet. The designed locomotion interface is applied in a virtual market system. Here the subject walks in a virtual market street with the desired speed. The stereo display based on virtual reality and the ambient sounds of the environment make the subject to have an immersed sense. The layout of shops in the virtual market system is in Japanese style, making the subjects experience much more realistic.

Parametrically excited inverted double pendulum and efficient bipedal walking with an upper body

Walking locomotion involves complex movement of total center of mass. Not only the lower body behavior but also the upper body behavior affects the walking characteristics. Therefore, in this paper we derive the principle of parametrically excited inverted double pendulum to consider both lower body and upper body dynamics. We propose one approach to utilize the upper body behavior of the robot for energy efficient bipedal locomotion. In addition, we analyze the property of parametrically excited inverted double pendulum.

Dynamic simulation of cervical traction therapy: Comparison between sitting and inclined positions

This research describes the construction of a mechanical cervical traction therapy simulation using a physics software engine. The model consisted of an anatomically correct human skeleton and two types of mechanical traction devices, each representing a unique traction position. While most of the movement in the skeleton simulation model was represented by simple hinges and ball-socket joints, the cervical spine was controlled by a spring-damper model to try to mimic its actual behavior in the human body. By varying the traction force, traction angle, and traction position in the simulation, the efficacy of the sitting and inclined traction was evaluated by comparing the anterior and posterior intervertebral separations each achieved. The simulation results suggested that the inclined position creates greater intervertebral separations on both the anterior and posterior sides than the sitting position. Our findings may serve as a reference for practitioners and medical equipment designers where cervical traction therapy is applied.

Age-Related Change of the Activity of Autonomic Nervous System Measured by Wearable Heart Rate Sensor for Long Period of Time

We analyzed long period of time (more than 10 h) autonomic nervous system data of 128 subjects (78 males and 50 females in 20’s, 30’s, 40’s and 50’s respectively) by using small wearable heart rate sensors. As a result, we found that there was a significant negative correlation (p value < 0.05) between LnTP (Total-Power as an indicator of comprehensive autonomic nervous system activity) and age for both sexes (genders). Moreover, the negative correlation value for male was higher than for female. The noticeable difference from the preceding study is that our research was based on data measured by many advanced wearable heart rate sensors which enabled to accumulate long period of time data in our daily life for many subjects and that we found the similar correlation between TP and aging comparing to the preceding study.

On energy-based robust passive control of a robot manipulator

This research proposes a novel robust passive control approach for a robot manipulator with model uncertainties to interact with its dynamic environment by considering the robots mechanic energy. Here, we analyze the robots robust passivity as seen from its environment. By adjusting a scaling parameter of the robots desired velocity with respect to the robots mechanic energy, we propose our robust passive control approach. We also perform computer simulations to show the effectiveness of our approach. In the simulation, we studied three cases where the robot has no model error, or its model has errors. As the results, it is found that the previous robust position tracking control may loss passivity to the external force. However, by adjust the robots desired velocity as in our approach, we can realize the robots passivity even when the robot has mode errors. The applications of our control approach are expected be used in those robots that are required to make safe physical interaction with human.

Alleviating adversarial attacks via convolutional autoencoder

In order to defend adversarial attacks in computer vision models, the conventional approach arises on actively incorporate such samples into the training datasets. Nonetheless, the manual production of adversarial samples is painful and labor intensive. Here we propose a novel generative model: Convolutional Autoencoder Model to add unsupervised adversarial training, i.e., the production of adversarial images from the encoded feature representation, on conventional supervised convolutional neural network training. To accomplish such objective, we first provide a novel statistical understanding of convolutional neural network to translate convolution and pooling computations equivalently as a hierarchy of encoders, and sampling tricks, respectively. Then, we derive our proposed Convolutional Autoencoder Model with the ‘adversarial decoders’ to automate the generation of adversarial samples. We validated our proposed Convolutional Autoencoder Model on MNIST dataset, and achieved the clear-cut performance improvement over the normal Convolutional Neural Network.

Adaptive Generative Initialization in Transfer Learning

In spite of numerous researches on transfer learning, the consensus on the optimal method in transfer learning has not been reached. To render a unified theoretical understanding of transfer learning, we rephrase the crux of transfer learning as pursuing the optimal initialisation in facilitating the to-be-transferred task. Hence, to obtain an ideal initialisation, we propose a novel initialisation technique, i.e., adapted generative initialisation. Not limit to boost the task transfer, more importantly, the proposed initialisation can also bound the transfer benefits in defending the devastating negative transfer. At first stage in our proposed initialisation, the in-congruency between a task and its assigned learner (model) can be alleviated through feeding the knowledge of the target learner to train the source learner, whereas the later generative stage ensures the adapted initialisation can be properly produced to the target learner. The superiority of our proposed initialisation over conventional neural network based approaches was validated in our preliminary experiment on MNIST dataset.

Assess BA10 activity in slide-based and immersive virtual reality prospective memory task using functional near-infrared spectroscopy (fNIRS)

ABSTRACT By using slide-based task in a laboratory setting, previous studies have found that activation of the rostral prefrontal cortex (BA10) is related to prospective memory performance. In this present study, we used immersive virtual reality (VR) technology to measure PM performance in a real-life task in a simulated virtual environment. Functional near-infrared spectroscopy was used simultaneously to record the rostral prefrontal cortex activities of the subjects. By comparing the data against the ones from the slide-based task, the result suggested that the activation of BA10 in the VR tasks were greater than the one in the slide-based tasks, and the VR tasks have the potential to identify the particular location of BA10 that is connected to the PM performance in our daily lives.