A design method for CSD Coefficient FIR filters using PfGA has been proposed. PfGA is a method based on GA that a parameter tuning task is not required in advance and a high speed design can be realized by a search with a small number of individuals. However, the degradation of designed filter performance is confirmed by the increasing of the number of penalized individuals. In this paper, the penalized individual is replaced to the unpenalized individual for improving the diversification. The effectiveness of the method is verified through several design examples.

Muscle condition is evaluated primarily based on physical therapy; however, evaluation has not been quantitative. To quantify muscle fatigue, the authors previously derived and defined “muscle fatigue time,” which quantifies muscle fatigue using frequency analysis based on the surface-ElectroMyoGram, of the biceps brachii. The authors also constructed a muscle fatigue time estimation model based on the relationship between muscle fatigue time and muscle load for each participant. However, as model parameters were participant-dependent, versality is insufficient to broadly apply the method to unknown participants. In this study, we attempted to select physical features that influence method parameters and estimate those parameters from selected physical features using multiple regression analysis. Percent body fat and biceps skinfold thickness were selected as physical features, and parameters were determined that yielded data with an error rate of approximately 13%. These results suggest that the variation in model accuracy between individuals can be eliminated using physical features.

A distributed energy management system (EMS) is designed to improve the efficiency of energy use across multiple EMSs by coordinating their operations. We previously proposed an optimization method using the alternating direction method of multipliers (ADMM). The unit commitment problem (UCP) involves determining the start-stop states and outputs of energy conversion equipment, where the decision variables include binary (0-1) variables. Since ADMM does not guarantee convergence for problems involving binary variables, it cannot be directly applied to UCP problem. A relaxation and projection method proposed by Takapoui et al. modifies ADMM to handle optimization problems with discrete variables. In this paper, we propose a method for applying ADMM with relaxation and projection to distributed EMSs.

Recently, customer value has been changing with the times, and customers’ values are evolving as well. Customer value is expected to support actions in various situations and be related to what customers want to achieve. Therefore, companies are increasingly focusing on methods to create customer value. This paper first registers journey maps and storyboards discussed in best practices and workshops as digital twin models for each theme. Teams exploring new ideas based on the theme will create scenarios based on UX design. In this process, they will refer to the created personas and scenarios, as well as pre-registered similar models, to assess the validity of the scenarios. Additionally, the paper proposes a method to create high-value scenarios that align with customer experiences by providing an environment where the team can visit as avatars in the metaverse and easily add ideas. This study will conduct practical exercises with students using this method and discuss the results, evaluation, and insights.

The multiple sound source localization problem using two microphones is an underdetermined problem where the number of sound sources is unknown. A method based on multiple group PSO has been proposed to solve this problem. In this method, both the number of sound sources and multiple directions are estimated simultaneously under the unified framework of multiple group PSO. However, the estimation accuracy of the number of sound sources often degraded in wide-angle directions. In this paper, a reason why the degradation occurs is verified and a method to improve accuracy is proposed. The effectiveness of the method is shown from several experimental results in the actual room.

At Kindai University, the question-and-answer system on Slack faced challenges due to a high volume of similar inquiries related to class content. A question-answering assistance system using large language models (LLMs) trained on historical questions is introduced in this paper. The effectiveness of three existing Japanese LLMs using two methods: a full parameter update method and a technique called LoRa are investigated. The results show that the full parameter update method achieved lower cross-entropy loss, while the combination of both methods imploving responses.

In this paper, we propose a deep learning model that applies PointNet architecture to improve response performance for the occlusion problem beyond the wrist, which was a problem in previous studies. We also evaluate its performance through experiments to estimate finger angles. Positions, joint angles, and accelerations of the fingers are measured and utilized on immersive devices and non-contact interfaces. However, it is difficult to measure them in situations where the wrist is hidden. Therefore, we devise finger angle estimation method that uses point cloud data of forearms as input data. From the results of experiments for 22 participants, the average RMSE was 22.79 and median of R² was 0.35 when the estimation was performed using a trained model. It suggests that the proposed model can estimate finger angles from the three-dimensional shape of forearms. Moreover, the time required for processing one estimation was 3.798 ms, which indicates that the response performance was good enough.

Learning support robots have been attracting attention as a means of reducing the burden on teachers in recent years. Although learning support robots have been studied in the past, very little research has taken into account the stress levels of learners. Humans may perform better when experiencing moderate stress, in line with Yerkes-Dodson's law. However, learners often find it difficult to control their own stress while studying. In this study, we developed a learning support robot named 'Ovot' that quantitatively evaluates stress using heart rate information and provides stress feedback by changing LED colors and adjusting the position of its head. In the experiment, participants performed desk work under conditions with and without feedback from Ovot. Stroop tasks were performed before and after the desk work, and reaction times were measured and compared. As a result, learners were able to maintain appropriate stress levels by viewing Ovot's feedback. Moreover, reaction times were reduced (p < 0.05) in the condition with feedback compared to the condition without feedback. Thus, the use of Ovot helped learners maintain appropriate stress levels and improved their performance.

In this letter, a disturbance observer (DOB) which can asymptotically estimate a step or sinusoidal disturbance by the full-order state observer with only the model parameters for second-order lag systems is studied. The basic idea is to obtain both the observer and the feedback gain matrices which can assign zeros on the imaginary axis and poles in the complex left half-plane of the closed-loop transfer function from the disturbance to the disturbance estimation error. Compared with the traditional DOB, the proposed DOB can generate an observer with a lower order.