Vital sign monitoring in daily life is very important for the early detection of hypertension, which causes cerebrovascular and cardiovascular diseases. A non-contact vital sign sensing is essential for vital sign monitoring in daily life. Our previous studies have constructed linear regression models for estimating blood pressure, using nasal skin temperature and photoplethysmogram components in the nasal region, which were obtained using a non-contact method. Feature extraction from the whole facial area is expected improve the accuracy in estimating blood pressure. In this study, feature extraction related to blood pressure levels from facial skin temperature distribution using a deep learning algorithm was performed. As the result, features at nasal and lip regions were extracted as common features related to blood pressure levels. Furthermore, a possibility for proposal of a general model for estimating blood pressure levels using the common features was shown.
Rehabilitation is important for hemiplegic patients after stroke to improve their motor deficit. For recovering and keeping their motor function following stroke, it may be effective to do rehabilitation at not only rehabilitation facility but also their home continuously. In this paper, we propose a rehabilitation assistant system at home for patients who go to rehabilitation facility regularly to perform rehabilitation in the right form which is taught by therapists at rehabilitation facility. This system implements a rehabilitation form classification engine based on decision tree algorithm applied to accelerometer data obtained during rehabilitation action and form classes defined by therapists. We evaluated the accuracy of the form classification engine with accelerometer data of simulated actions by 5 subjects without any disability. The accuracy of the proposed engine was more than 70% all over the subjects.
The central nervous system does not control multiple muscles independently, whereas it manipulates a musculoskeletal system with redundant degrees of freedom. In specific, muscle synergy that represents muscle coordination realizes an advanced motion control of human body. In the pedaling exercise, the coordinated muscle activity between lower extremities is key for high-efficient exercise when cyclists use binding pedals which fixes both pedals and shoes via cleats. Under the present circumstances, few reports focused on the muscle coordination of both legs. Thus, the relationship between the coordination of both legs and the mechanism of pedaling has not yet been clarified. In this study, we investigated the muscle synergy hidden in lower extremities by measuring surface electromyography and crank rotation angle during pedaling exercise under different mechanical constraints. In the results, this study confirmed the muscle cooperativeness formed by both the left and right leg during pedaling. Besides, the muscle cooperativeness had different roles on both legs because this coordination was asymmetrical regardless of the changes in cadence and workload. The muscle coordination formed by both legs was considered to be an evaluation index that could explain how the left and right leg were coordinated for effective pedaling with binding pedals.
In this paper, we propose a new method to update parameters of the controller implemented in the closed loop system. The proposed parameter tuning method here requires only one-shot output data obtained in the closed loop experiment together with the feedback controller used in this initial experiment and the desired tracking specification. This method is derived based on the idea of the data-driven estimation and the concept of the internal model control (IMC) structure that only appears in the virtual closed loop system. Finally, the validity and the effectiveness of the proposed method are verified through actual experiments.
In a system that suppresses power fluctuation caused by wind power and solar power generation by a storage battery (We call it wind turbine and photovoltaic power generation system, WPBHS), it is assumed that the number of constituent equipment is large. Therefore, by dividing the constituent equipment into a plurality of groups (We call it grouping), the operation plan is optimized dispersedly. In this case, it has been shown that WPBHS has fault tolerance by overlapping storage batteries in plural groups. Also, enumerating the grouping in advance allows us to switch the grouping and recover the system immediately when a system failure occurs. Motivated by this, we consider an overlapped grouping optimization problem focused on the performance deterioration and fault tolerance of the distributed management system. Enumerating the group, especially, the overlapped group requires large computational load with the more devices added. First, this paper formulates overlapped grouping optimization for WPBHS. Second, this paper proposes a graph representation method of the overlapped grouping and a near-optimal enumeration method.
Generally, electrical machines are designed using CAE (computer aided engineering) software. In addition, it is combined the design with optimization method for improving various objective functions such as magnetic path, structure and loss. However, there are various problems when multi-objective optimization is applied: modeling of design parameters is complicated, vast amounts of time are necessary, and the result is highly complex display owing to multi-objective. Therefore, it is difficult to select valid value from the complex result. In this paper, the authors have described the method that applies multi-objective function optimization for electrical machine by multi-objective genetic programming (MOGP). Moreover, its superiority and usability have been considered. Likewise, they have described the method that displays the result using dimension reduction.
In recent year, embedded systems equipped with multi-core processor have been increasing to improve performance and energy saving. When tasks are executed in parallel on multi-core of embedded systems, a mechanism for mutual exclusion that maintains real-time property between cores becomes important. In this paper, we describe a mutual exclusion method using Lock-free algorithm for multi-core processor. This method has three benefits. The first is that each core can fairly acquire the lock and it is possible to determine the worst execution time. The second is that the worst execution time when increasing the number of cores is shorter than the time proportional the number of cores. The third is high CPU utilization. This has the advantage that it is possible to real-time design on a multi-core system that requires mutual exclusion. As a result of evaluation, it was possible to satisfy three requirements by the proposal method.
Freshness evaluation of Ezoshika meat based on bioelectrical impedance analysis (BIA) using LCR meter is carried out to estimate the quality of the meat. An arc is extracted from a locus of measured impedance to evaluate the condition of the meat. It is revealed that freshness of the meat may be estimated from the position of center of the arc. The arcs obtained by two types of equivalent-circuit models and a simple least squares method (LSQ) are compared. It is also found that the arc extracted by the LSQ explains the measured result well.
Hippocampal adult neurogenesis is hypothesized to contribute spatial pattern separation. However, it is unclear whether this effect depends on the structure. In order to find structure independent functions at a cell population level, we aimed at understanding how adult neurogenesis promotes spatial pattern separation on a scale of a group of cells. Hippocampal neuronal networks with or without immature neurons were trained with two types of electrical stimulations with different spatial patterns by microelectrode array. Ratio of response times of trained stimuli to the other decreased only in the network with immature neurons, suggesting that the immature neurons promote spatial pattern learning. The result indicates that adult neurogenesis enhances the topological learning of spatial patterns.
Time-delay stimulation in rat cortical slices is considered to be a good experimental model for studying temporal learning, an intrinsic property of the cortical circuit. However, limited knowledge exists surrounding the mechanism of how stimulation with fixed-interval modifies network dynamics. Here, we aimed to elucidate network structure dependency in temporal learning by examining whether the same stimulation is affective for dissociated cultures. A neuronal network with a modular structure was formed with a micro-fabricated device. While training stimulation was effective for altering firing rate, no clear difference was shown in network dynamics, suggesting that the same stimulation used in a slice culture was insufficient for dissociated networks.
Plant factories are attracting worldwide attention as a technology for stably producing crops. One of the problems of plant factories is tipburn, which is a physiological disorder of crops. If tipburn occurs, its identification is done by eye observation and tipburn leaves are trimmed by hand, which requires much labor and cost. In this study, we aim to perform binary discrimination of tipburn occurrence and its non-occurrence in lettuce cultivated at artificial light plant factories using machine learning with convolutional neural networks. As a result of experiments, it is confirmed that binary discrimination can be performed with high accuracy.
This research is proposed to investigate an easy, fast, and effective method for automatic sleep stage detection using spectral features extraction from electrocardiography (ECG) signal alone. Sleep stage detection is the gold standard for sleep analysis. A sleep physician may suspect a treatment and diagnosis of sleep diseases through sleep stage detection. Polysomnography (PSG) method that generally used for detecting sleep stage. This method is intrusive and difficult to be implemented for in-home and portable systems. Moreover, it is involved a lot of effort, time and cost. The automatic sleep stage detection that uses only ECG signal is expected can solve these problems. The proposed method was tested and evaluated on 51 subjects that consist of 16 healthy subjects, 9 patients with insomnia, 4 patients with sleep-disordered breathing, and 22 patients with REM behavior disorder. In this research also tried to apply a minimal feature for automatic sleep stage detection. The two features applied were the normalized Low Frequency, LF (n.u.) band power and normalized High Frequency, HF (n.u.) band power that obtained from spectral features extraction. These features were then used as inputs for sleep stage classification. Mostly commonly used learning classifiers is implemented to classify sleep stage, namely KNN, NN, DT, SVM, and proposed DTB-SVM. The proposed method using DTB-SVM and spectral features extraction of ECG achieved an average classification specificity, sensitivity, and overall accuracy of 98.31%, 91.84%, 95.06%, respectively. The proposed method is able to obtain all sleep stage condition on patients and non-patients subjects. However, it is feasible to implement in-home and portable system of automatic sleep stage detection instead of using a multichannel signal.
In this paper, the stability analysis of an active noise control system using single adaptive filter with an external injection noise and an acceleration method that makes use of the property of this noise are discussed. The convergence performance of this adaptive algorithm is analyzed from the viewpoint of a property of the Hessian matrix. As a result, it is proved that the Hessian matrix is positive definite even when the power density spectrum of the noise source signal and the external injection noise are different. Furthermore, by selecting the power density spectrum of the external injection noise to be the inverse property of that of the noise source, it is proved that the system with the white noise source and the white injection noise can be realized. Finally, these theoretical considerations are verified by computer simulation.
A learning classifier system is an adaptive system that obtains a set of appropriate action rules that adapts to multi-step problems by training action rules defined in if-then form by trial and error process, in a similar framework as reinforcement learning. Because of that the input signals of the classifier system are encoded into binary values, bit strings are often lengthened when dealing with such a problem that the state of the environment continuously changes. A neural network can treat with real values as input signal, however, it cannot be applied to multi-step problems. This paper proposes a system that responds to problems such that the state of the environment continuously changes by combining a neural network and a classifier system, and actions are selected from multiple options, so that output can be defined as discrete values. In order to verify the effectiveness of the proposed system, this paper conducts several numerical experiments using benchmarks corresponding to muti-step problems defined by continuous values.
In this paper, we propose the 1+n/k frequency divider which does not depend on the duty ratio of the multi-phase clock. The proposed circuit does not have to waveform shape the multi-phase clock which is an input signal. Also, the frequency division range is not limited to n ≤ ⌊k/2⌋ (⌊x⌋ is the largest integer not exceeding x) as in the conventional circuit.
Potential therapy, using a device to create an electric field around a person, is widely used as a home-healthcare apparatus to treat ailments such as headaches, muscle stiffness, or insomnia. However, the mechanisms for its effects remain unknown. In this study, we assumed that this therapy interacted with the autonomic nerve because its effects include remission of insomnia. Although patients report an increase in skin temperature during treatment, it has not been studied whether the body temperature also increases during the treatment. Thus, this study aims to evaluate influence of potential therapy on the autonomic nerve and the axillary temperature. Twenty healthy males participated in the experiment, receiving 60 minutes of potential therapy. Before and after the therapy, electrocardiogram was recorded for five minutes and the axillary temperature was measured. The effect on the autonomic nerve was evaluated using the low frequency/high frequency (LF/HF). A result of t test revealed no significant difference in LF/HF before and after the treatment. On the other hand, the same test indicated a significant increase in the axillary temperature.
In this paper, we extend the proposed reinforcement learning with multiplex learning space to an environment that needs delay time for getting rewards. Concreatly, we prepare the multiplex learning spaces corresponding to each equal interval delay time within the predicted range. We simulated it, comparing with an ordinary one. As a result, the proposed method could get the best policy, but the ordinary method could not.