Ryosuke Yamanishi

Relationships between emotional evaluation of music and acoustic fluctuation properties

We studied the relationships between the acoustic fluctuation properties of music and the emotional evaluations of music as a component technology for an automated song selecting system based on instinct and human emotion. When people listen to music, they select songs reflect their own feelings at the time, and then they emotionally evaluate the song. We believe that our emotional evaluation of songs are influenced by the fluctuation properties of both the volume and pitch of songs. Thus, we focused on the fluctuation properties containing dynamic ideas concerning music, and extracted thirty six fluctuation properties concerning both the volume and pitch from each song using the Fast Fourier Transform. We also prepared a subjective evaluation experiment for plural songs using the Semantic Differential method, and obtained an emotional evaluation of the songs. Then, we studied the relationships between the extracted properties and the emotional evaluations of the songs by conducting a multiple discriminant analysis. As a result, a high percentage of the questions were answered correctly and low discriminant errors were shown, and therefore, we suggested that the fluctuation properties of the songs influenced the emotional evaluations of them. Furthermore, we confirmed the especial properties related with the emotional evaluation of the music by taking into consideration the coefficients of the liner discriminants of the canonical variates that describe the discriminant spaces.

Interactive music recommendation system for adapting personal affection: IMRAPA

We have so various types of entertainment, and music is one of the most popular one. In this paper, we proposed music recommendation system that interactively adapts a users personal affection with only a simple operation, in which both acoustic and meta features are used. The more a user uses the proposed system, the better the system adapts the users personal affection and recommends the suitable songs. Through the evaluational experiment, we confirmed that the proposed system could recommend songs adapting users personal affection even if the personal affection variated.

Recommendation System for Alternative-Ingredients Based on Co-occurrence Relation on Recipe Database and the Ingredient Category

Many people often cook a dish with a cooking recipe on Websites and magazines. The listed ingredients in the recipe sometimes can not be prepared. This paper proposes a recommendation system for alternative ingredients. The recommendation ingredients based on co-occurrence frequency of ingredients on recipe database and ingredient category stored in a cooking ontology. The result of the subjective evaluation experiments showed 88% appropriateness for alternative-ingredients recommendation.

Automatic Throwing of Questions Generated from Posted Comments to Activate Text Based Discussions

It is more difficult to activate text based discussions than face-to-face discussions because participants tend to hesitate give their opinions. To activate text based discussions, it is necessary for participants to give their opinions many times. We suppose that participants will give their opinions if questions are thrown to discussions. This paper proposes an agent that throws questions automatically in text based discussions. The questions are generated from comments posted by discussion participants previously in a discussion. The agent chooses a comment including a word that appears frequently in previous posted comments. The agent adds “why” or “how” to the beginning of the chosen comment, then the agent throws the comment as a question to the discussion participants. We experimentally verified that the agent could activate discussions by throwing questions.

Hybrid Approach of Ontology and Image Clustering for Automatic Generation of Hierarchic Image Database

This paper proposes a hybrid approach of ontology and image clustering to automatically generate hierarchic image database. In the field of computer vision, ”generic object recognition” is one of the most important topics. Generic object recognition needs three types of research: feature extraction, pattern recognition, and database preparation; this paper targets at database preparation. The proposed approach considers both object semantic and visual features in images. In the proposed approach, the semantic is covered by ontology framework, and the visual similarity is covered by image clustering based on Gaussian Mixture Model. The image database generated by the proposed approach covered over 4,800 concepts (where 152 concepts have more than 100 images) and its structure was hierarchic. Through the subjective evaluation experiment, whether images in the database were correctly mapped or not was examined. The results of the experiment showed over 84% precision in average. It was suggested that the generated image database was sufficiently practicable as learning database for generic object recognition.

Estimation of Dialogue Moods Using the Utterance Intervals Features

Many recent studies have focused on dialogue communication. In this paper, our target is to make a robot support a communication between humans. To support a communication between humans, we believe that there are two important functions: estimating dialogue moods and behaving suitably. In this paper, we propose dialogue mood estimation model using the utterance intervals. The proposed estimation model is composed by relating the subjective evaluations for several adjectives with the utterance intervals features. Through the estimation experiments, we confirmed that the proposed system could estimate the dialogue moods with a high degree of accuracy, especially for “excitement,” “seriousness,” and “closeness.” And we suggested that the utterance intervals features had a high potential for the dialogue mood estimation.

Automated song selection system complying with emotional requests

Recently, we have a lot of musical pieces due to a large capacity of storage. However, it would be difficult to select the song with bibliographic data as the capacity of the music database increases. Therefore, we proposed an emotional song selection system. In this study, the Acoustic - Emotion model was composed by relating the acoustic fluctuation features that can explain the time variation of music with the emotional evaluations of music obtained through the subjective evaluation experiments. Based on the model, the emotional evaluations of music were calculated from their acoustic features. Using the proposed system, user can select the song with the adjective words and their degrees.

Dance Dance Gradation: A Generation of Fine-Tuned Dance Charts

This paper proposes a system to automatically generate dance charts with fine-tuned difficulty levels: Dance Dance Gradation (DDG). The system learns the relationships between difficult and easy charts based on the deep neural network using a dataset of dance charts with different difficulty levels as the training data. The difficulty chart automatically would be adapted to easier charts through the learned model. As mixing multiple difficulty levels for the training data, the generated charts should have each characteristic of difficulty level. The user can obtain the charts with intermediate difficulty level between two different levels. Through the objective evaluation and the discussions for the output results, it was suggested that the proposed system generated the charts with each characteristic of the difficulty level in the training dataset.

Enjoy Watching Japanese Chess Games like Football: an Evaluation Method of Game Positions for Beginners

Researches for computer Shogi (Japanese chess) have been widely conducted. Most of the previous researches have tried to create a strong program that can defeat the professional players. They use the estimated information of pieces arrangement to evaluate a current game position. The evaluation is less comprehensible for beginners because it seems not to correspond to the current arrangement. We propose a novel method for the beginners to evaluate a current game position of Japanese chess game. Our method uses only a current arrangement of pieces to evaluate the game position. We developed a prototype interface to visualize the evaluated game position, and conducted a small experiment. We asked participants to watch Japanese chess games with the interface, and speak what they thought while watching the games. The participants gave more questions and interpretations that were related to understanding of game positions. The results indicated that our system enhanced the beginners enjoyment of watching Japanese chess games with questions and understandings.

Detection of Words Accepted to Dynamic Abstracts Focusing on Local Variation of Word Frequency

Abstract Through the widely-spread of digital devices such as smartphone, the digital books have become more popular. We are aiming to develop a system to support reading novel taking an advantage of digital books. In this paper, we propose an elemental method to generate dynamic abstracts for each reading progress. Generating dynamic abstract can be assumed as a topic of summarization task in the field of natural language processing. The proposed method focuses on the local variation of word importance, though some existing criterions for summarization focus on the overall word importance. We prepared four types of local variation and compared the effectiveness of those with each other. We conducted the experiment to detect words accepted to manually-generated dynamic abstracts with each types of the proposed method while the general word importance criterion (tf-idf) is used as the comparative method. Through the discussions of the results, it was confirmed that some types of the proposed method were more effective to detect the words accepted to dynamic abstracts than the comparative method.