Three-beat sounds are widely used in cheering. Five- or seven-syllable phrases are used in Japanese and Chinese traditional poems. Odd numbers of short sounds seem to relate to an affective phenomenon. Our experiment shows that a three-beat sound is the most exciting beat sound for people. We propose a neural-network model that relates to such a phenomenon. It is realized by the combination of neural networks with reciprocal-inhibition and cascade structures. We additionally take into account the effect of attention fluctuation. We simulate the neural-network operation, indicating that degree of excitement defined for the model takes the highest level for three beats. This result is consistent with our experimental result. When the attention fluctuation is suppressed, the degree of excitement takes higher levels for odd numbers of beats than even numbers of beats. This relates to the ground of five- or seven-syllable phrases used in Japanese and Chinese traditional poems.
As an efficient approach to create garments with a designer’s Kansei, we developed a method for three-dimensional (3D) garment modeling with sleeves by deforming a 3D reference clothing model according to the contours of garment images with the aim to apply it to garment patternmaking. We determined an appropriate reference bodice model from various dress forms. We made the reference sleeve model by sweeping the armhole of a dress whose armhole was on a plane. Using this method, we obtained models of two jacket bodices and sleeves by deforming each reference clothing model using the contours of jacket images from the front and side views. After adjusting the positions of the 3D models of the bodice and sleeve, we obtained jacket models with a sleeve. We obtained the jacket armhole from the intersection of the surfaces of the bodice and sleeve models. We successfully made 3D garment models and armholes using the proposed method. We also obtained bodice and sleeve patterns that reflected the armhole. The simulated patterns exhibited a similar clothing shape to the images with curved shapes along the arm. Thus, we demonstrated that the proposed method efficiently reflected the designer’s Kansei clothing model.
Recently, computational studies of analysis of facial attractiveness features have attracted much attention because of the comprehensive understanding that is difficult to achieve using only experimental approaches. However, the differences in results between models and methods have not been examined in detail. In this study, a tuned convolutional neural network (CNN) model was constructed, and the results were confirmed using several methods to visualize features important for prediction. Results using gradient-weighted class activation mapping (Grad-CAM), Grad-CAM++, and Score-CAM methods showed that the eye area tended to be activated in highly attractive female images, consistent with findings in psychology. In contrast, some features showed different results depending on the method and training times. It was suggested that the model learns the highly universal and method-dependent features of facial attractiveness. This affective engineering approach contributes to understanding perceptual psychology and various engineering applications.
With the explosive growth of online discussions nowadays, fostering interesting and satisfying group discussions for all group members has become challenging. In this work, we particularly seek to address the issue of online group formation where diverse participants with various topic interest levels gather and carry-on open-ended synchronous discussions in small groups. In these groups, members often encounter various difficulties, especially when their degree of interest in the discussed topic decreases drastically. Our proposed method is a boids-model inspired algorithm that captures group discussion dynamics in terms of the evolution of discussed topics over time, and variations in group members’ degree of interest for the discussed topics. Discussion topics are modeled as multidimensional vectors where dimensions correspond to factors that are associated with group members’ interest vectors. In this paper, we present the proposed method and discuss its potential for achieving dynamic tracking of variations in individuals’ interests and detection of left-out members. We confirmed the feasibility as well as the meaningfulness of the proposed approach through numerical simulations. In addition, we outline our future plans to investigate the meaningfulness of our approach through more complex simulations and interactions involving actual users.
In this study, we used canonical correlation analysis (CCA), frequency component correlation method (FCCM) and ensemble model to develop a steady state visual evoked potential brain-computer interface (SSVEP-BCI) with fifty-selective. For the proposed fifty-selective SSVEP-BCI with only CCA, it was not possible to obtain sufficient SSVEP induction. In a previous study where a similar problem occurred, the maximum accuracy was 79.53%, and the information transfer rate (ITR) was 45.16 bits/min. Therefore, we proposed FCCM and ensemble model in CCA to improve the accuracy even when the SSVEP induction was not sufficient. We used the proposed method to achieve the highest accuracy of 93.23% and the highest ITR of 58.88 bits/min. The system also achieved an average accuracy of 71.01% and an average ITR of 40.79 bits/min, demonstrating the usefulness of the system. Also, the maximum accuracy and ITR in additional experiments were 98.53% and 65.41 bits/min.
The term “Highly Sensitive Person” (HSP) refers to an individual’s temperament. They have common characteristics with attention-deficit/hyperactivity disorder and other, and their central feature is a high empathy. However, the questionnaires may be inaccurate because they depend on the examinee’s self-perception. Conversely, emotional empathy and activation of mirror neurons are as characteristics of HSP. We thought a quantitative evaluation of people with HSP would be possible if we could physiologically measure these characteristics. In this study, we investigated the correlation between the Highly Sensitive Person Scale (HSPS) and changes in alpha wave band power induced by facial expression stimulation and the event-related desynchronization (ERD) induced by the video of motor imagery as physiological indicators. The results showed HSPS scores of above 100 had an ERD of above 50%, indicating mirror system activity. Furthermore, HSPS scores of above 100 had lower alpha wave band power values when presented sadness face images.
Kawaii is a Japanese cultural uniqueness that attracts attention around the world. This study focused on pink as a typical kawaii color. The goal is to clarify the relation of preferences in pink colors and fashion taste. Four pink colors were selected and used in our questionnaire to collect data about most kawaii and most favorite pink colors as well as the behavior in using pink items in 2020 and 2021. From the questionnaire results, we obtained tendencies of the pink colors for the most kawaii and the most favorite as well as their relationship with fashion trend. Finally, we clarified the pink color that have constant trend over time and the one that tends to be influenced by annual trend. The results suggest that different pink colors may give different impressions according to the annual trend. The results from this study contribute to the fashion industry.
Practicing English pronunciation is difficult for non-native speakers because of the differences in vowels and consonants. There are several ways to practice them such as Shadowing, however, if the voice’s features greatly differ from the learner’s voice, it should be difficult for learners to reproduce. To solve this problem, we propose a method to make the pronunciation data of the model pronunciation similar to the learner’s voice by using UTAU and Interactive Differential Evolution. A listening experiment was conducted with the concrete system of IDE and UTAU. Twelve examinees participated in the experiment through ten generations based on paired comparisons for making the voices similar to their own voices inside their heads. As a result, we could successfully make the voices similar to the examinees’ voices. Since it has paired comparison, we believe that the paired comparison-based IDE is a better method than the general Interactive Genetic Algorithm with scoring.
Today, almost all aspects of life are influenced by technology to simplify these aspects of life. Researching how humans and computers interact is very important to build a good model for better human-computer interaction in the future. Using technology such as computer vision, we can now collect the information we are looking for directly from a human. Humans can use many kinds of modalities to interact with computers. Hands are perhaps the largest source of body language information after the face. To understand the gesture’s meaning, we can use MediaPipe Hands, developed at Google LLC, as a method to track and recognize human hands. However, if we want to understand some kinds of hand gestures using MediaPipe Hands, we need to create a condition using if-else manners. This research tried to collect the wide variety of each hand gesture using the 21 key points in x, y, and z coordinates as a feature. We chose Support Vector Machine (SVM) and Artificial Neural Network (ANN) as classifiers to validate the impact of large sample datasets. This research found that ANN is the best among all the methods we used as a classifier method for the 3D value of 21 key points from the hand skeleton. The accuracy and F1-score from ANN were 98.4% accuracy and 98.2% F1-Score, respectively, representing the best performance for each class of all the methods we used in this research.