Soohan Kim

A soft computing approach to localization in wireless sensor networks

In this paper, we propose two intelligent localization schemes for wireless sensor networks (WSNs). The two schemes introduced in this paper exhibit range-free localization, which utilize the received signal strength (RSS) from the anchor nodes. Soft computing plays a crucial role in both schemes. In the first scheme, we consider the edge weight of each anchor node separately and combine them to compute the location of sensor nodes. The edge weights are modeled by the fuzzy logic system (FLS) and optimized by the genetic algorithm (GA). In the second scheme, we consider the localization as a single problem and approximate the entire sensor location mapping from the anchor node signals by a neural network (NN). The simulation and experimental results demonstrate the effectiveness of the proposed schemes by comparing them with the previous methods.

Human movement detection algorithm using 3-axis accelerometer sensor based on low-power management scheme for mobile health care system

Phone and PDA mobile devices that recognize a users movements and biometric information that can be utilized in a sensor system have been generating interest among users This paper proposes a low-power management scheme that uses the baseband processor installed in a portable communications device to limit the electric power consumed by the device, along with a human movement detection algorithm that records and predicts the movement of the mobile user in a low-power mode In addition, a mobile healthcare system is developed to use the proposed scheme and algorithm This system uses 3-axis accelerometer sensors on an Android platform and calculates the amount of human movement from the sensor output The users uphill, downhill, flat area, climbing stairs, going down stairs, and jogging movements were measured with accuracies of 93.2%, 97.4%, 97.6%, 98.8%, 92.2%, and 90.8%, respectively.

Design and implementation of smart driving system using context recognition system

In this paper, we develop a context recognition system that is based on multimodal biometric signals and applicable to smart driving system. The context recognition system includes a biometric analysis module that analyzes and recognizes human biometric signal patterns. The context recognition system can recognize a users emotion and level of concentration by analyzing ECG (electrocardiogram) and EEG (electroencephalogram) patterns. To predict the concentration and stress status of the user, the electroencephalogram rendering system utilizes 5 signal values: MID_BETA, THETA, ALPHA, DELTA, and GAMMA. Also, electrocardiogram analysis system utilizes 5 basic signal values: P, Q, R, S, T wave. To recognize the users electrocardiogram signal patterns, a deformation K-means-based EM algorithm was applied.

Fast reconnection of adaptive HTTP streaming in heterogeneous networks

This paper proposes the time minimizing of disconnecting contents for adaptive HTTP streaming when a mobile device moves from previous network to current network. The proposing decision rule, IM, RMIM approach and maintenance, can determine the most suitable media segment with determined data rate from the media server. This paper also verifies the effect of reducing the measurement time by the decision rule through the simulation. As a result, the contents request method proposed takes less a disconnect time than the contents request method used in ISO/TEC FCD 23001-6. The contents request method used in ISO/IEC FCD 23001-6 takes 10 sec. The contents request method proposed in this paper, however, takes just 3 sec.

Using the baseband processor power management technology for mobile devices

Recently advancement in digital equipment without regard to location and use of portable computers and mobile phones such as smart phone are being convinced and the vitalization of the smart phone in the marketplace is spreading. However, the application which works continuously consumes more power than the standby mode and unable to perform the work well are the problems. As a result, the need of a power management technology in the mobile environment is growing. In this paper, without the addition of MICOM (Micro Computer) applications using the baseband processor also offers power-saving technology. The proposed technique is based on the written application of a pedometer to measure electric power consumed and evaluate the performance of power management technology. Via the supplies of the GPIB (General Purpose Interface Bus) interface, the electric current was measured on a PC, the result of the current consumed in Sleep Mode is less than 10 mA. When the pedometer run, the average current consumption observed was 0.81089 mA.

A Classifier Algorithm Exploiting User’s Environmental Context and Bio-signal for U-Home Services

U-Home is a home-service through an interaction between human and object. Smart-home-middle-wear provides its users with services needed through interactions between users and home equipment. In this study, users’ conditions in four rooms with Smart-home-middle-wear using had been sent through EG sensor device and they were then classified by emotion-perceiving-agent-system adapting an algorithm. The emotions, which were experimented, had been divided into eight categories; Normal, Happy, Surprise, Fear, Neural, Joy, Stress(Yes) and Stress(No). In this study’s experiments, modified Decision Tree algorithm was adapted and it extracted over 90% of results totally.

Research on a Smart Input Device Using Multimodal Bio-signal Interface

This paper presents a smart input device based on multimodal analysis methods using human bio-signals. The core of the smart input device is software modules, which consist of an intelligent driving function and an analysis processing function for bio-signals. The smart input device utilizes a multimodal interface that analyzes and recognizes human bio-signal patterns. The multimodal analysis system can recognize a user’s emotional and stress status by analyzing an electrocardiogram (ECG), and it can determine the user’s level of concentration from electroencephalogram (EEG) patterns. To analyze the concentration, stress, and emotional status of the user, the EEG rendering system and ECG analysis system use five signal values, i.e., MID_BETA, THETA, ALPHA, DELTA, GAMMA, and the P, Q, R, S and T waves. A reformation of SVM and a clustering algorithm were applied to the user’s EEG and ECG signal patterns for body context recognition. In our experiment, the on/ off status of the user’s stress status controls the difficulties of the game, such as the selecting the type of race course or the number of obstacles. In addition, the speed of the car can be controlled depending on the concentration or non-concentration status of the user. The stress status of the user was predicted with an accuracy of 83.2% by the K-means algorithm, and the concentration status was predicted with an accuracy of 71.85% by the SVM algorithm. We showed that a bio-signal interface is quite useful and feasible for new games.

Extendable Simulation Framework for Virtual World Environment Based on the DEVS Formalism

A virtual world is an interactive virtual environment in which users interact with each other with computers, and can be used as a platform for virtual training activities. In order to enhance the trainee’s immersive experience, domain-specific simulation models are required for virtual world services. For this reason, we propose an extendable simulation framework for the virtual world. The simulation framework is composed of Core Simulation Framework and Virtual Level Architecture. By utilizing the Core Simulation Framework and the Virtual Level Architecture, the content creator of the virtual world can create extendable simulations for domain-specific content.

Design and implementation of intelligent physics system for mobile environment

This paper presents the architecture of an intelligent mobile physics engine. An intelligent physics engine is a software system that can be used to produce realistic physical effects in real time. The results of an analysis of its ability to automatically select proper physics effects in a general mobile game environment is also presented, along with a method for maximizing the excitement in a mobile game with a minimum amount of calculation. A technique is proposed for intelligently embodying the users expertise, and then evaluated by experiments with a mobile 3D racing car game. This study showed that the intelligent mobile physics engine should support an automatic acceleration mode controlled by the mobile 3D racing environment, collision detection model, and deceleration function by gravity.

Design and development of multimodal analysis system based on biometric signals

In this paper, we present the multimodal interface and analysis system which is based on biometric signals and applicable to contents. The multimodal interface includes a biometric analysis module that analyzes and recognizes human biometric signal patterns. The biometric multimodal interface can recognize a users emotion and concentration status by analyzing ECG(electrocardiogram) and EEG(electroencephalogram) patterns. The electroencephalogram analysis system utilizes 5 basic signal values to predict the concentration status of the user: MID_BETA, THETA, ALPHA, DELTA, and GAMMA signal. To recognize the users electrocardiogram signal patterns, K-means-based EM algorithm was applied. In emotion recognition, the neural emotion showed the highest accuracy, and three emotions were in a range of 55.8 to 75.1% accuracy. Stress recognition showed a high performance result of 83.2% accuracy.