Hidetaka Nambo

Feature extraction of multi-gas sensor responses using Genetic Algorithm

Abstract It is an established technique in the field of privacy protection to use a gas sensor to monitor indoor environments. In this paper, we propose a method for recognizing human activities in an indoor environment by using several kinds of high-sensitivity gas sensors. The feature of the method is estimating the signals with Genetic Algorithm (GA). The results showed that the proposed method effectively worked for the recognition of human activities. This paper demonstrates a significant result through utililization of the proposed technique in this research.

The Combination of Evolutionary Algorithm Method for Numerical Association Rule Mining Optimization

The numerical problem of association rule mining is an updated issue. Numerous authors propose some methods to solved it. A number of them are using the optimization approach by Particle Swarm Optimization (PSO). The problem is that the PSO trapped in local optima when searched the best particle in every iteration. Many researchers solved this problem by combining with Cauchy distribution because it is tremendous for searching in a large neighborhood. Hence, that combination will be implemented to accomplish the numerical association rule mining problem for some objective functions such as confidence, comprehensibility, interestingness. Based on the result the multi-objective of PSO for Numerical Association Rule Mining Problem with Cauchy Distribution (PARCD) showed the better result than the method of Multi-objective Particle Swarm Optimization for Association Rule Mining (MOPAR).

The Support System for a Visual Impaired Person Using Laser Eyewear

We construct a system for a visually impaired person. On our system, an image is input by a camera, is transformed by an image conversion system, and is provided to the visually impaired person by an eyewear like glasses. If this system is used, a person can move with the eyewear, and also can watch a scene in the position of line of sight. In the eyewear composed of our system, which is made by QDLaser Inc., the image is projected directly onto a retina. QDLaser Inc. has worked on a small-sized scanning optics to be installed inside the glass frame, and has developed the eyewear named Retinal Imaging Laser Eyewear with the trademark of RETISSA. One of remarkable features of RETISSA is focus free. Regardless of the function of a crystalline lens and eyesight, the wearer can see laser scanned full-color image captured by a digital camera installed in the center of the frame connected to a controller. One of other features of RETISSA is an attractive appearance of eyewear. RETISSA also enables universal design like sunglasses and corrective glasses and realizes complete AR. We explain the system and this eyewear.

Development of the Estimation Method of Resident’s Location Using Bioelectric Potential of Living Plants and Knowledge of Indoor Space

In Japan, an aging society is progressing rapidly, and the number of aged single-person households are also increasing. Therefore, to ensure residents’ safety, a demand of monitoring is increasing. Generally, microphones or cameras are used for monitoring, however, it is hardly accepted to residents because their privacy will be violated by such devices. In this paper, we focused a living plant as a monitoring device. Living plants are easily accepted to residents. Furthermore, they have a property at bioelectric potential of a living plant will change affected by an activity of human around the plants. Moreover, according to the change of the potential, it is able to estimate a distance or a kind of activities of human. Therefore, we have studied to develop a method to use living plants for monitoring. However, we have found that monitoring by living plants has problems of an accuracy and a measuring distance. In this paper, we develop a monitoring method using living plans and a knowledge base of indoor space. The knowledge contains a layout structure of the space. Using both living plants and the knowledge base, we propose an improved monitoring method. And then, we report a result of experiment to confirm an accuracy of the method integrated content of the abstract in your paper should include the brief introduction of research intention, method, result and conclusion.

A link selection method for web‐browser using eye‐gaze input

Methods for browsing Web pages using eye-gaze input have been proposed for severely physically handicapped people who cannot handle a computer mouse to utilize convenient services on the Internet. There are two important functions for Web-browsers, namely scroll and link selection. Since it is difficult to introduce existing link selection methods using eye-gaze input for home usage because of high costs or complexity, we propose in this paper a novel link selection method using eye-gaze input. The proposed method analyzes the HTML source in a Web page and makes a group of links by the content of each. In evaluation experiments, we compare the proposed method with a comparative method using eye-gaze input with respect to the average response time in selecting the links. The results of the experiments show that the average response is nearly twice as fast as that of the compared method.

The Translation System from Japanese into Braille by Using MeCab

When a blind student enters college, many Braille books such as textbooks and reference books are needed. Students are hoping to get them before class starts. Volunteers want to give Braille books to blind students as soon as possible. However, Braille has its own rules. In translating into Braille some mistakes will occur. In order to minimize the errors, the following system is required. In other words, it is a system that converts printed textbooks into machine-readable codes, rewrites them into Kana characters in Braille rules, makes a split, corrects braille conversion mistakes and grammatical mistakes, and translates them into Braille. I will describe Japanese complexity and Japanese Braille here. I will also describe the complexity of rules when expressing Japanese Braille and the system we build.

Application of Deep Learning to Classification of Braille Dot for Restoration of Old Braille Books

The Braille book gradually causes from ordinary printed books. Because the Braille is a reading letter and visually impaired people touch the boss with the finger’s belly, in Braille books that are frequently read, the Braille are dirty with aged use, holes open, and they collapse. In our project, we convert Braille books into machine-readable electronic data and finally restore the book. First, a Braille book is scanned by an image. Next, we detect braille by image recognition technology. And next we interpret the image, classify and identify Braille. Furthermore, we correct errors such as misdetection of braille and misidentification. Finally, we save the result as a character code. In this paper, we report techniques of elimination of gradient in the scanning images of Braille book, those of detection and identification of the Braille dot, and those of interpretation of braille dots as cells.

A System that Warns of Dangerous Environmental Sounds for the Hearing Impaired

By just a looking it, It’s hard for us to know if hearing impaired people have obstacles. Therefore, Hearing impaired people cannot understand of the circumstances, and they cannot get supports enough. They often live without telling they have obstacles people around them themselves. The purpose of this study, we will develop the system that visually notifies the dangerous sounds generated outside the visual field. We construct the database of environmental sounds, and select the dangerous sound from the environmental sound. This system checks the waveform data and dangerous sound database, and displays danger on the monitor if it meets conditions such as volume and distance from the source. This system displays characters using “realistic fonts” that we have developed from before, that is the expression method expressing the feeling of sound using manga technique. In this study, we will report outline of this research and a prototype of our system. We record the sound supposed to be dangerous sounds with variety of directions and range from sound source using microphone array. And we compare the waveform data for each microphone recorded, and verify the features. This system identifies three kinds of sounds and displays using realistic fonts reflecting volume and direction of sound source.

SARIMA Model of Bioelectic Potential Dataset

Bioelectric potential of plant produces a low electrical signal because of the plant activities like photosynthesis and respiration. Furthermore, the electrical signal will change because of the environmental factors such as temperature, humidity and human behavior. Some authors successfully used the bioelectric potential of plant for detecting the various human activities, like walking, jumping, open the door, and etc. They used decision tree (DT) J48, multi layer perceptron (MLP) and convolution neural network (CNN) as the analysis method. However, the previous accuracy was no satisfied and estimating the human position globally not specifically an exact location. This research has aim to construct a time series model for bioelectric potential dataset which is SARIMA model and to build infrastructure design of human position estimation in an exact location. For constructing the SARIMA model we use one observation location and obtained the best model is SARIMA (1,0,0) with accuracy of 80%. In addition, this research successfully designed the infrastructure of human position estimation using three locations.

Prediction of Oxygen Saturation by Pulse Oximetory from Image and Sound Data with Long Short-term Memory Recurrent Neural Network

An Internet of Things (IoT) communication function was attached to inexpensive sensors such as cameras and microphones and was used for data acquisition and analysis. In this study, the value of saturation of blood oxygen measured by pulse oximetry (SpO2), which is used for sleep apnea syndrome (SAS) detection, was estimated from the data obtained from the camera and microphone. SpO2 was recorded by a pulse oximeter worn by subjects suspected of having SAS when sleeping overnight. The camera and microphone were located on the side of the bed to record the data. The SpO2 value was learned using long short-term memory (LSTM), which is one of the deep neural network methods that have shown excellent results as a method of analyzing time series data. When evaluated by leave-one-out cross validation using the data of four persons, it was found that the amplitude of the estimated SpO2 was about half. The cause seems to be individual differences in time from apnea occurrence to SpO2 declination. By doubling the estimation result, it was confirmed that the SpO2 value was well estimated.