The method and results of recognizing single Japanese sounds using surface electromyography (sEMG) signals generated from muscles around the mouth are presented. We determined six features of the waveforms of four muscles (a total of 24 indexes) and recognized 45 single Japanese sounds. We used machine learning with a neural network to improve sound recognition. The neural network has a 24 node input layer, a 100 node intermediate layer, and a 45 node output layer. Each index of a sound was entered into the input layer; the probabilities of the sound were output to the output layer. They were compared, and the output (sound) with the highest probability was determined to be a recognition sound. We used the cross-entropy loss as the loss function and gradient descent as the machine learning method. Machine learning, which built a neural network, has dramatically increased recognition; it stands at approximately 94%.
In this study, we developed a capacitive seating analysis sensor using a conductive textile for evaluating a Seating Motion. Physical burden on caregivers is a common problem at nursing care sites. Developing long-term caregiving skills can help reduce the physical burden on amateur caregivers. Several quantitative evaluation studies have focused on the skills of caregivers. However, such evaluation is highly expensive and require huge area because it involves the use of motion capture systems with several 3D cameras for dynamic motion analysis. In our previous study, we developed a conductive textile sensor for measuring the seating position of a care recipient in a wheelchair. The sensor could measure either seating body pressure or distance between the buttocks and the seat. The system can measure the care recipient's motion without requiring any motion capture system. We recommend using the sensor to evaluate the skills of caregivers; the sensor output can be used to determine the seating speed of the care recipient on the chair. Herein, we report the validity of the measurement system comprising a conductive textile-based capacitive seating analysis sensor for evaluating seating in nursing care.
The purpose of this study was to develop a computerized classification method for molecular subtypes in low-grade gliomas (LGGs) with multi-scale 3D-at-tention branch networks analyzing multi-sequence brain MRI images. Our dataset consisted of brain T1-weighted and T2-weighted MRI im-ages for 217 patients (58 Astrocytoma IDH-mutant, 49 Astrocytoma IDH-wildtype, and 110 Oligodendroglioma). The proposed method was constructed from a feature extractor, an attention branch, and a perception branch. In the feature extractor, the feature maps were extracted from brain T1-weighted and T2-weighted MRI images, respectively. The attention branch focused on a tumor region and generated the attention maps normalized to 0.0 - 1.0. The feature maps were then multiplied by the attention maps to weight features on LGG in the feature maps. The molecular subtype in LGG was evaluated in the perception branch. The classification accuracy for the proposed method was 63.6%, showing an improvement when compared with the conven-tional method using only single sequence (T2-weighted) MRI images (59.9%).
In recent years, the lowing supply voltage and the A/D conversion accuracy are required for ΔΣA/D Conversion Circuits. We proposed MASH structure using the 2nd order swing-suppression A/D conversion circuit at previous research. We showed that this circuit is high accuracy and suitable at low supply voltage. However, the accuracy of this circuit is degraded by the analog device value deviation and the OP-AMP DC gain degradation. In this paper, we proposed the low voltage ADC using Sturdy-MASH (SMASH) constructure and the swing-suppression A/D conversion circuit. We realized the SNR of 86 dB with ±1% device value deviation and the OP-AMP DC gain of 40 dB.
We propose a total optimization model of electric power networks for EV taxi services. Planning of charging and discharging of the EV batteries are optimized in conditions of not only the transportation demand but also the PV generation. This optimization problem is modeled into a mixed integer programming model, in which each EV movement in a day is expressed. The optimal solution obtained by this model can provide performance limits for the assumed system performance and will provide important information for the design policy of the system designer or relevant stakeholders. Through some numerical experiments using actual demand data, we show the potential of the target transportation services.
This paper proposed a guide dog robot that can express visual environment by voice. The guide dog robot uses NIC, a combination of CNN and LSTM methods, to recognize the visual scene and generate captions. The generated sentences are then converted to speech using speech synthesis to guide the visually impaired. The guide dog robot can not only guide the visually impaired person safely to their destination, but also providing explanation services of street scenes, enabling the visually impaired person to enjoy visual scenes. The system is consisted of visual scene recognition, caption, and speech synthesis. The effectiveness of the system is confirmed through experiments using the proposed guide dog robot.
In this paper, we propose a method of estimating the gaze point of table-meeting participants by a spherical camera. Assuming participants are sitting around a table and observed by a spherical camera, multiple persons and their gaze points can be observed by a single camera simultaneously. In contrast with the previous research on estimating gaze points using a spherical camera based on a spherical model, we develop a deeper neural network and add a normalization term to the cost function which constrains estimated gaze points onto a unit sphere. As shown in the results of the comparative experiments, the proposed method improves the accuracy of gaze point estimation significantly.
This paper presents a method for forecasting the change direction of nonstationary time series using the improved leading indicator. The leading indicator is a method developed by Ehlers that translates a time series mathematically in the future direction with respect to the time axis and calculates the leading value of the time series. However, this method has the problem that the leading value can be calculated only in the low frequency region with a normalized frequency of 0.06 or more (normalized period of 17 or more) at the maximum. In order to solve this problem, by gentle slope the amplitude characteristics in the low frequency region of the leading indicator, it is possible to calculate the leading value in the frequency domain with a normalized frequency of 0.25 or more (normalized period of 4 or more) at the maximum. By applying the preceding value to the instantaneous periodic time series by the improved sine wave indicator developed by Ehlers, it is possible to forecast the change direction of the non-stationary time series in the short term.