Yunyoung Nam

A similarity-based leaf image retrieval scheme: Joining shape and venation features

In this paper, we propose a new scheme for similarity-based leaf image retrieval. For the effective measurement of leaf similarity, we have considered shape and venation features together. In the shape domain, we construct a matrix of interest points to model the similarity between two leaf images. In order to improve the retrieval performance, we implemented an adaptive grid-based matching algorithm. Based on the Nearest Neighbor (NN) search scheme, this algorithm computes a minimum weight from the constructed matrix and uses it as similarity degree between two leaf images. This reduces necessary search space for matching. In the venation domain, we construct an adjacency matrix from the intersection and end points of a venation to model similarity between two leaf images. Based on these features, we implemented a prototype mobile leaf image retrieval system and carried out various experiments for a database with 1,032 leaf images. Experimental result shows that our scheme achieves a great performance enhancement compared to other existing methods.

Utilizing venation features for efficient leaf image retrieval

Most Content-Based Image Retrieval systems use image features such as textures, colors, and shapes. However, in the case of a leaf image, it is not appropriate to rely on color or texture features only as such features are very similar in most leaves. In this paper, we propose a new and effective leaf image retrieval scheme. In this scheme, we first analyze leaf venation which we use for leaf categorization. We then extract and utilize leaf shape features to find similar leaves from the already categorized group in a leaf database. The venation of a leaf corresponds to the blood vessels in organisms. Leaf venations are represented using points selected by a curvature scale scope corner detection method on the venation image. The selected points are then categorized by calculating the density of feature points using a non-parametric estimation density. We show this techniques effectiveness by performing several experiments on a prototype system.

SmartBuckle: human activity recognition using a 3-axis accelerometer and a wearable camera

Recognizing human activity is one of the most important concerns in many ubiquitous computing systems. In this paper, we present a wearable intelligence device for medical monitoring applications. We called the SmartBuckle that is designed to recognize human activity and to monitor vitality. We developed human activity recognition algorithms and evaluated them by using data acquired from a 3-axis accelerometer with embedded one image sensor in a belt. In order to evaluate, acceleration data was collected from 9 activity labels. In the image sensor, we extracted activity features based on grid-based optical flow method. In the 3-axis accelerometer sensor, we used the correlation between axes and the magnitude of the FFT for feature extraction. In the experiments, our classifiers showed the excellent performance in recognizing activities with an overall accuracy rate of 93%.

Child Activity Recognition Based on Cooperative Fusion Model of a Triaxial Accelerometer and a Barometric Pressure Sensor

This paper presents a child activity recognition approach using a single 3-axis accelerometer and a barometric pressure sensor worn on a waist of the body to prevent child accidents such as unintentional injuries at home. Labeled accelerometer data are collected from children of both sexes up to the age of 16 to 29 months. To recognize daily activities, mean, standard deviation, and slope of time-domain features are calculated over sliding windows. In addition, the FFT analysis is adopted to extract frequency-domain features of the aggregated data, and then energy and correlation of acceleration data are calculated. Child activities are classified into 11 daily activities which are wiggling, rolling, standing still, standing up, sitting down, walking, toddling, crawling, climbing up, climbing down, and stopping. The overall accuracy of activity recognition was 98.43% using only a single- wearable triaxial accelerometer sensor and a barometric pressure sensor with a support vector machine.

Time-Varying Coherence Function for Atrial Fibrillation Detection

We introduce a novel method for the automatic detection of atrial fibrillation (AF) using time-varying coherence functions (TVCF). The TVCF is estimated by the multiplication of two time-varying transfer functions (TVTFs). The two TVTFs are obtained using two adjacent data segments with one data segment as the input signal and the other data segment as the output to produce the first TVTF; the second TVTF is produced by reversing the input and output signals. We found that the resultant TVCF between two adjacent normal sinus rhythm (NSR) segments shows high coherence values (near 1) throughout the entire frequency range. However, if either or both segments partially or fully contain AF, the resultant TVCF is significantly lower than 1. When TVCF was combined with Shannon entropy (SE), we obtained even more accurate AF detection rate of 97.9% for the MIT-BIH atrial fibrillation (AF) database (n = 23) with 128 beat segments. The detection algorithm was tested on four databases using 128 beat segments: the MIT-BIH AF database, the MIT-BIH NSR database ( n = 18), the MIT-BIH Arrhythmia database ( n = 48), and a clinical 24-h Holter AF database ( n = 25). Using the receiver operating characteristic curves from the combination of TVCF and SE, we obtained a sensitivity of 98.2% and specificity of 97.7% for the MIT-BIH AF database. For the MIT-BIH NSR database, we found a specificity of 99.7%. For the MIT-BIH Arrhythmia database, the sensitivity and specificity were 91.1% and 89.7%, respectively. For the clinical database (24-h Holter data), the sensitivity and specificity were 92.3% and 93.6%, respectively. We also found that a short segment (12 beats) also provided accurate AF detection for all databases: sensitivity of 94.7% and specificity of 90.4% for the MIT-BIH AF, specificity of 94.4% for the MIT-BIH-NSR, the sensitivity of 92.4% and specificity of 84.1% for the MIT-BIH arrhythmia, and sensitivity of 93.9% and specificity of 84.4% for the clinical database. The advantage of using a short segment is more accurate AF burden calculation as the timing of transitions between NSR and AF are more accurately detected.

Intelligent video surveillance system: 3-tier context-aware surveillance system with metadata

This paper presents an intelligent video surveillance system with the metadata rule for the exchange of analyzed information. We define the metadata rule for the exchange of analyzed information between intelligent video surveillance systems that automatically analyzes video data acquired from cameras. The metadata rule is to effectively index very large video surveillance databases and to unify searches and management between distributed or heterogeneous surveillance systems more efficiently. The system consists of low-level context-aware, high-level context-aware and intelligent services to generate metadata for the surveillance systems. Various contexts are acquired from physical sensors in monitoring areas for the low-level context-aware system. The situation is recognized in the high-level context-aware system by analyzing the context data collected in the low-level system. The system provides intelligent services to track moving objects in Fields Of View (FOVs) and to recognize human activities. Furthermore, the system supports real-time moving objects tracking with Panning, Tilting and Zooming (PTZ) cameras in overlapping and non-overlapping FOVs.

m CLOVER: mobile content-based leaf image retrieval system

This demonstration presents a content-based leaf image retrieval system that supports wired/wireless access. For example, if we want to know about a plant that we encounter in a mountain or field, we might look it up in an illustrated book. But, it will take a long time to search due to the lack of appropriate indexing or search clues and huge amounts of similar plants. In order to solve this problem, we developed a content-based leaf image retrieval system called mCLOVER that supports both wired and wireless access and includes a set of novel features for easy querying and efficient retrieval.

Physical activity recognition using multiple sensors embedded in a wearable device

In this article, we present a wearable intelligence device for activity monitoring applications. We developed and evaluated algorithms to recognize physical activities from data acquired using a 3-axis accelerometer with a single camera worn on a body. The recognition process is performed in two steps: at first the features for defining a human activity are measured by the 3-axis accelerometer sensor and the image sensor embedded in a wearable device. Then, the physical activity corresponding to the measured features is determined by applying the SVM classifier. The 3-axis accelerometer sensor computes the correlation between axes and the magnitude of the FFT for other features of an activity. Acceleration data is classified into nine activity labels. Through the image sensor, multiple optical flow vectors computed on each grid image patch are extracted as features for defining an activity. In the experiments, we showed that an overall accuracy rate of activity recognition based our method was 92.78%.

Sleep Monitoring Based on a Tri-Axial Accelerometer and a Pressure Sensor

Sleep disorders are a common affliction for many people even though sleep is one of the most important factors in maintaining good physiological and emotional health. Numerous researchers have proposed various approaches to monitor sleep, such as polysomnography and actigraphy. However, such approaches are costly and often require overnight treatment in clinics. With this in mind, the research presented here has emerged from the question: “Can data be easily collected and analyzed without causing discomfort to patients?” Therefore, the aim of this study is to provide a novel monitoring system for quantifying sleep quality. The data acquisition system is equipped with multimodal sensors, including a three-axis accelerometer and a pressure sensor. To identify sleep quality based on measured data, a novel algorithm, which uses numerous physiological parameters, was proposed. Such parameters include non-REM sleep time, the number of apneic episodes, and sleep durations for dominant poses. To assess the effectiveness of the proposed system, three participants were enrolled in this experimental study for a duration of 20 days. From the experimental results, it can be seen that the proposed monitoring system is effective for quantifying sleep quality.

New Potential Functions with Random Force Algorithms Using Potential Field Method

Autonomous mobile robot path planning is a common topic for robotics and computational geometry. Many important results have been found, but a lot of issues are still veiled. This paper first describes new problem of symmetrically aligned robot-obstacle-goal (SAROG) when using potential field methods for mobile robot path planning. In addition, we consider constant robot speed for practical use. The SAROG and the constant speed involve two potential risks: robot-obstacle collision and local minima trap. For dealing with the two potential risks, we analyze the conditions of the collision and the local minima trap, and propose new potential functions and random force based algorithms. For the algorithm verification, we use WiRobot X80 with three ultrasonic range sensor modules.