Recently, it has become a trend to construct a thermal environment with human thermal comfort. The advantage of human thermal comfort is that we could adjust the environment through the thermal sensation of human. Since human thermal comfort is influenced by several factors such as environmental temperature, environmental humidity, airflow, mean radiant heat, and so on, it is usually difficult to be directly evaluated. This paper proposes a method of estimating subjective thermal comfort and objective thermal comfort using CNN by RGB image data, thermal image data, and sensor data. We built a pipe-type booth in a room with a small air conditioner, two heaters, a humidifier, and a dehumidifier to acquired learning data, and we trained CNN under six different learning patterns to estimate thermal comfort. We found that image data is conducive to estimate subjective thermal comfort, and sensor data is conducive to estimating objective thermal comfort.

This research aims to build a system that estimates assessment expressions to others by machine learning, collects them in real time, and feeds them back to the participants, and this study proposes two techniques that are necessary for this purpose. First, we propose a method to estimate the evaluative behavior to others, which is simplified to positive, negative, and neutral, from the facial images and head acceleration of participants in the conversation. The supervised machine learning algorithm requires ground-truth labeling. To analyze the expression of a large number of people, such as in a classroom, ground-truth labeling becomes a challenge in analysis and system construction. Second, to ease the labeling process we propose and implement a method to recommend the estimated labels on the labeling software. The accuracy of the proposed method is approximately 0.45 in F-value. Considering the time-series information, the proposed method recognizes major changes in positive evaluation behaviors. Presenting recommended labels on our labeling support system increase with time for the ground-truth labeling. However, the accuracy of labeling positive assessments improve in five out of six subjects.

Hundreds of thousands of Chinese characters are used in the Japanese and Chinese writing systems. Memorizing all the Chinese characters requires great effort because learners have to memorize not only the shape of the character but also its meaning and usage. Research to find a method to present the shape of the figures using sound has been conducted. Consequently, it is possible to generate a song for learning Chinese characters comprehensively using sounds that represent the shapes of the characters and lyrics that represent the meaning and usage. We propose a method to generate a song based on the structures, strokes, and usage of Chinese characters and a method to memorize Chinese characters using the generated song.

There have been several studies on object detection and activity recognition on a table conducted thus far. Most of these studies use image processing with cameras or a specially configured table with electrodes and an RFID reader. In private homes, methods using cameras are not preferable since cameras might invade the privacy of inhabitants and give them the impression of being monitored. In addition, it is difficult to apply the specially configured system to off-the-shelf tables. In this work, we propose a system that recognizes activities conducted on a table and identifies which user conducted the activities with load cells only. The proposed system uses four load cells installed on the four corners of the table or under the four legs of the table. User privacy is protected because only the data on actions through the load cells is obtained. Load cells are easily installed on off-the-shelf tables with four legs and installing our system does not change the appearance of the table. The results of experiments using a table we manufactured revealed that the weight error was 38 g, the position error was 6.8 cm, the average recall of recognition for four activities was 0.96, and the average recalls of user identification were 0.65 for ten users and 0.89 for four users.

In recent years, due to the increasing popularization of various broadcast services, there has been an increasing demand for information filtering techniques. Although filtering methods that have been proposed filter data in their own ways, mathematical representation of these methods does not exist. Consequently, it is not possible to qualitatively evaluate various filtering methods, to optimize processing methods in filtering, or to design a declarative language for filtering processes. In our previous work we have defined a filtering function that represents filtering as a function, and clarified the properties of different filtering methods. In practice, current filtering methods actually consist of multiple methods. In this paper, we make it possible to qualitatively represent those filtering methods by introducing the concept of composition into the framework of filtering function. Moreover, we reveal the characteristics of those combined filtering methods by clarifying the properties of composite filtering functions.

Because of the recent development of network technologies such as ATM (Asynchronous Transfer Mode), broader channel bandwidth is becoming available everywhere in the world wide networks. As one of the new technologies to make good use of such broadband channel, dynamic relocation of databases through networks, which we call database migration, will soon become a powerful and basic database operation of practical use. We discuss our proposal of a distributed database system, DB-MAN (distributed database system based on DataBase Migration in ATM Networks), which takes advantage of database migration in virtual LANs (local area networks) of ATM networks. DB-MAN has two notable mechanisms: a mechanism for selecting the transaction processing method and a mechanism for concurrency control with database migration. The former, is a mechanism which chooses the more efficient method between two transaction processing methods: the conventional method based on the two phase commit protocol and our method employing database migration. The latter is a mechanism to prevent the transaction processing throughput from deteriorating in environments where data contention is a significant factor. Then we show simulation results regarding performance comparison between our proposed system and the conventional distributed database system based on the two phase commit protocol. The obtained results demonstrate that effective use of database migration gives higher performance than that of the conventional system.

Due to recent developments in network technologies, broader channel bandwidth is becoming prevalent in worldwide networks. As one of the new technologies making good use of such broadband channels, dynamic relocation of databases through networks, database migration, will soon be used in practice as a powerful and basic database operation. We propose two transaction processing methods to take advantage of database migration in broadband networks. These methods choose the most efficient transaction processing method between the conventional method, based on the two-phase commit protocol, and our method, using database migration. We also propose a concurrency control mechanism and a recovery mechanism for our proposed methods. Simulation results are presented comparing the performance of our proposed methods and the conventional transaction processing method based on the two-phase commit protocol. The results demonstrate that the effective use of database migration produces better performance than the conventional method.

Deductive databases have the ability to deduce new facts from a set of existing facts by using a set of rules. They are also useful in the integration of artificial intelligence and databases. However, when recursive rules are involved, the number of deduced facts can become too large to be practically stored, viewed or analyzed. This seriously hinders the usefulness of deductive databases. In order to overcome this problem, we propose four methods to discover characteristic rules from a large number of deduction results without actually having to store all the deduction results. This paper presents the first step in the application of knowledge discovery techniques to deductive databases with large numbers of deduction results.