Haruhiko Kimura

Dynamic model to estimate the dependence of gas sensor characteristics on temperature and humidity in environment

Abstract A gas sensor using tin oxide changes its electrical resistance ( R s ) with a change in the concentration of a reducing gas ( C ). However, the R s vs. C correlation is affected significantly by the temperature ( t ) and relative humidity ( h ) of the surrounding atmosphere. We propose a dynamic model to describe the influences of t and h . As is well-known, the R s vs. C correlation was expressed satisfactorily by an equation, log 10 R s = α log 10 C + β ( α and β constant), under fixed conditions of t and h . Analysis of the observed data indicated that the values of α and β under various environmental conditions were distributed along a plane when plotted three-dimensionally against t and h . This fact allows to propose a simple model in which both α and β are linear functions of t and h , respectively. The three constants included in each function were determined experimentally. The resulting equation of α and β were found to describe the R s vs. C correlation well under various conditions of t and h .

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.

Detection of human activities by inverse filtration of gas sensor response

Abstract In an approach to identifying human activities in an indoor space from the response signals of a combustible gas sensor, an indoor space model was presented in which the response signal and the gas generation due to the human activities were correlated linearly through an indoor gas transfer function. Analysis indicated that: (1) the gas transfer by diffusion can describe essential features of the transfer function in the indoor space, and (2) the inverse filtration based on the space transfer characteristics thus estimated enables one to restore or extract the gas-generation function from the sensor signal output. The model was applied to identify daily human activities in a domestic house. The sensor response output consisted of heavily overlapping signals due to human actions taking place in succession. However, the inverse filtration could eliminate the overlapping (residual) components from the overall output to single out each human action associated with causing gas generation in the indoor space.

Proposition of sensor agent for estimation of air-pollution direction and its experimental simulation

Air pollutants appear at various areas on the earth due to economic growth, and harm the health of the residents of those regions. Various countries around those areas are also polluted. Many kinds of air-polluting factors must be grasped, for example, the place of origin, kind of pollutant, its direction and speed. These factors are also grasped quickly over a wide area. Various kinds of chemical sensors have been investigated and developed to detect air pollutants. In this study, metal oxide odor sensors are installed at 12 points in an experimental residence and experimental simulations are carried out. Each sensor with a microprocessor has the ability of an agent, which can autonomously make many advanced decisions and estimations. This sensor system is named sensor agent. The sensor agents can derive pollution direction by cooperation and collaboration with other sensor agents. Derived directions are compared with daily resident behaviors and reliability is examined. Results show that polluting direction is estimated with good reliability and a clue to practical use is achieved.

High‐speed production system using dynamic two‐way switching of match algorithm

Expert systems support reasoning based on experts knowledge and data represented through a set of rules, thus substituting for human experts. Production systems are a popular tool for expert system building. However, production systems suffer from slow processing speed, and extensive research has been conducted to improve their performance. In particular, processing of high-cost rules takes an extremely long time. This study offers dynamical switching between direct match algorithm, which was previously proposed for high-cost rule processing, and conventional Rete algorithm; it is found that the two algorithms prove to make a complementary pair which ensures efficient processing. In addition, rules for which this dynamical switching is appropriate are defined, and efficiency of the proposed approach is confirmed through experiments with sample knowledge bases.

A proposal to reduce cumulative reasoning time in hypothetical reasoning

Considered in this paper is a method to reduce cumulative execution time in hypothetical reasoning using inference-path network. Logic-based hypothetical reasoning provides an efficient framework that can be applied to such areas as diagnostics and design, but in so doing, the reasoning speed may prove insufficient. This problem seems to be solvable by using partial hypothetical knowledge. In this case, more than one iteration with partial hypothetical knowledge may be required to find a solution. Some existing techniques involve iterative reasoning within the same knowledge base while varying the hypothetical knowledge; networks are generated as necessary, and hypotheses are synthesized at each iteration. On the contrary, the method proposed in this study involves reduction of execution time by reusing the results of previous reasoning. Specifically, the cumulative execution time is reduced by applying the concept of stage, by partially deterring inference-path generation, and by keeping the results of hypothetical reasoning.

A modeling of tin-oxide gas sensor responses for temperature and humidity changes

Output responses of an tin-oxide gas sensor depends on indoor temperature/humidity, but not on the reacted gas concentrations. The modeling of a gas sensor characteristics curve between gas concentrations and voltage was attempted with respect to a temperature/humidity change, to remove the influence of indoor environment from sensor responses. A characteristics curve was derived as an inverse proportion function and approximated by undecided variables with a plane approximation of temperature/humidity coefficients. The results of this study in a home indicate that each temperature/humidity coefficient is approximate to linear equation in error ranges of 5% and approximate functions of a characteristics curve are in a mean error range of about 1% for a reserved region.