Search

Drones are capable of exploring large areas due to their swift mobility and utilization efficiency, so their use in various fields such as surveillance is increasing. When a drone performs its mission, the estimation of drone's location relies on GPS signals. Therefore, it becomes difficult to accurately estimate the drone's location in environments where GPS signals are unstable or in denied areas, which directly affects the safety and mission performance of the drone. Drone missions by autonomous flight are often carried out along a given route. This can be done by extracting and comparing the feature points of the image data from the camera mounted on the drone and the image data on the already constructed map using the SIFT algorithm. It allows the drone’s location to be estimated under GPS-denied environments. In this paper, we aim to construct aerial map data and verify the technology for estimating and correcting the drone’s location through image data received from the drone. [ABSTRACT FROM AUTHOR]

With the expansion of online financial services due to the use of mobile phones, stores, stores and branches are closing. Banks are also reducing offline business networks and reorganizing the financial transaction environment through mobile banking. However, financial transactions are inconvenient for the elderly, who do not use online services. In this paper, multimodal AI technology is utilized for senior-friendly services in mobile banking apps. Research UI/UX aspects of mobile banking apps and present prototypes. Two authentication modes are used for non-face-to-face identity verification using multimodal AI technology: visual (face verification) and auditory (speaker verification). For face identification, we use the FaceMash ML solution based on the Mediapipe framework, and for speaker identification, we use d-vector, a speech feature extraction method, in a deep learning DNN model. If multimodal AI technology is reflected, it will be possible to realize a convenient and safe senior-friendly use environment and contribute to the development of senior citizens' financial capabilities. [ABSTRACT FROM AUTHOR]

In this paper, we explore customer churn in a telecommunication company, measuring the probability of users discontinuing their service. Our investigation employs various machine learning models, including Decision Tree, Random Forest, XGBoost, LightGBM, SVM, Logistic Regression, and Deep Neural Network. To accomplish this, we leverage data collected from a California-based telecommunication company, initially containing 38 feature data. In order to reduce complexity, we apply Lasso regression to select the five most crucial features for determining customer churn: household situation, service satisfaction, loyalty, payment capability, and contract type. Through performance evaluation, we demonstrate that accurate predictions of customer churn can be achieved even with five features, emphasizing the significance of feature selection. [ABSTRACT FROM AUTHOR]

With the rapid development of deep neural network theory and application research, the effect of face detection has been improved. However, due to the complexity of deep neural network calculation and the high complexity of the detection environment, how to detect face quickly and accurately becomes the main problem. This paper is based on the relatively simple model of the MTCNN model, using FDDB (Face Detection Dataset and Benchmark Homepage), LFW (Field Label Face) and FaceScrub public datasets as training samples. At the same time of sorting out and introducing MTCNN(Multi-Task Cascaded Convolutional Neural Network) model, it explores how to improve training speed and Increase performance at the same time. In this paper, the dynamic image pyramid technology is used to replace the traditional image pyramid technology to segment samples, and OHEM (the online hard example mine) function in MTCNN model is deleted in training, so as to improve the training speed. [ABSTRACT FROM AUTHOR]

Glaucoma is a typical neurodegenerative disease that causes permanent vision deficits. Early detection is important to prevent vision loss as the patient is unaware of the damage and severity because pain is not induced during the process and progresses gradually. To predict glaucoma, fundus image-based deep learning models are being actively developed. In the traditional approach, original fundus photographs including red, green, and blue (RGB) channels were used as a learning material of deep learning models. In this paper, we propose a new pre-processing method to produce the learning images by segmenting vascular, optic nerve papillae, and optic nerve indentation images from the original fundus images, and we built a deep learning model; previous RGB channels are replaced into our three segmented images. We test the proposed method using two datasets and confirm that proposed method shows better performance than previous approaches that use original fundus photographs. [ABSTRACT FROM AUTHOR]

In this paper, we propose a new approach for tracking each human in a surveillance camera network with various resolution cameras. When tracking human on multiple non-overlapping cameras, the traditional appearance features are easily affected by various camera viewing conditions. To overcome this limitation, the proposed system utilizes facial information along with appearance information. In general, human images captured by the surveillance camera are often low resolution, so it is necessary to be able to extract useful features even from low-resolution faces to facilitate tracking. In the proposed tracking scheme, texture-based face descriptor is exploited to extract features from detected face after face frontalization. In addition, when the size of the face captured by the surveillance camera is very small, a super-resolution technique that enlarges the face is also exploited. The experimental results on the public benchmark Dana36 dataset show promising performance of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Recently, due to the popularization of 3D depth cameras, new researches and opportunities have been made in research conducted on RGB images, but estimation of human hand pose is still classified as one of the difficult topics. In this paper, we propose a robust estimation method of human hand pose from various input 3D depth images using a learning algorithm. The proposed approach first generates a skeleton-based hand model and then aligns the generated hand model with three-dimensional point cloud data. Then, using a random forest-based learning algorithm, the hand pose is strongly estimated from the aligned hand model. Experimental results in this paper show that the proposed hierarchical approach makes robust and fast estimation of human hand posture from input depth images captured in various indoor and outdoor environments. [ABSTRACT FROM AUTHOR]

The butterfly species recognition technology based on machine learning using images has the effect of reducing a lot of time and cost of those involved in the related field to understand the diversity, number, and habitat distribution of butterfly species. In order to improve the accuracy and time efficiency of butterfly species classification, various features used as the inputs of machine learning models have been studied. Among them, branch length similarity(BLS) entropy or color intensity entropy methods using the concept of entropy showed higher accuracy and shorter learning time than other features such as Fourier transform or wavelet. This paper proposes a feature extraction algorithm using RGB color intensity entropy for butterfly color images. In addition, we develop butterfly recognition systems that combines the proposed feature extraction method with representative ensemble models and evaluate their performance. [ABSTRACT FROM AUTHOR]

In this paper, a lightweight network with fewer parameters compared to the existing object detection method is proposed. In the case of the currently used detection model, the network complexity has been greatly increased to improve accuracy. Therefore, the proposed network uses EfficientNet as a feature extraction network, and the subsequent layers are formed in a pyramid structure to utilize low-level detailed features and high-level semantic features. An attention process was applied between pyramid structures to suppress unnecessary noise for prediction. All computational processes of the network are replaced by depth-wise and point-wise convolutions to minimize the amount of computation. The proposed network was trained and evaluated using the PASCAL VOC dataset. The features fused through the experiment showed robust properties for various objects through a refinement process. Compared with the CNN-based detection model, detection accuracy is improved with a small amount of computation. It is considered necessary to adjust the anchor ratio according to the size of the object as a future study. [ABSTRACT FROM AUTHOR]

Binary classification has been broadly investigated in machine learning. In addition, binary classification can be easily extended to multi class problems. To successfully utilize machine learning methods for classification tasks, preprocessing and feature extraction steps are essential. These are important steps to improve their classification performances. In this paper, we propose a new learning method based on weighted least squares. In the weighted least squares, designing weights has a significant role. Due to this necessity, we also propose a new technique to obtain weights that can achieve feature transformation. Based on this weighting technique, we also propose a method to combine the learning and feature extraction processes together to perform both processes simultaneously in one step. The proposed method shows the promising performance on five UCI machine learning data sets. [ABSTRACT FROM AUTHOR]

Multimodal refers to the combination or integration of data from multiple sources or modalities. Multimodal data integration involves merging diverse sources, such as imaging, clinical, and genetic data, to obtain a more comprehensive view of patient health. This process, facilitated by deep learning, has emerged as a transformative technique in the biomedical field, enabling a comprehensive understanding of patient health and leading to improved decision-making. This study provides an overview of the general pipeline for multimodal data integration techniques. We start by examining different types of biomedical data and exploring feature extraction and preprocessing techniques for these modalities. Moreover, we discuss data-level, feature-level, and decision-level fusion methods in detail, specifying scenarios where each fusion approach is most effective. The review also addresses the use cases and existing applications of multimodal data fusion, highlighting their impact on disease diagnosis, prognosis, treatment planning, and personalized medicine. [ABSTRACT FROM AUTHOR]

In this paper, we propose a real-time system for counting people. The proposed system uses an impulse radio ultra-wideband(IRUWB) radar to estimate the number of people in a given location. The proposed system uses learning-based classification methods to count people more accurately. In other words, a feature vector database is constructed by exploiting the pattern of reflected signals, which depends on the number of people. Subsequently, a classifier is trained using this database. When a newly received signal data is acquired, the system automatically counts people using the pre-trained classifier. We validated the effectiveness of the proposed algorithm by presenting the results of real-time estimation of the number of people changing from 0 to 10 in an indoor environment. [ABSTRACT FROM AUTHOR]

Commercialized speech recognition systems that have an accuracy recognition rates are used a learning model from a type of speaker dependent isolated data. However, it has a problem that shows a decrease in the speech recognition performance according to the quantity of data in noise environments. In this paper, we proposed the vector quantization based speech recognition performance improvement using maximum log likelihood in Gaussian distribution. The proposed method is the best learning model configuration method for increasing the accuracy of speech recognition for similar speech using the vector quantization and Maximum Log Likelihood with speech characteristic extraction method. It is used a method of extracting a speech feature based on the hidden markov model. It can improve the accuracy of inaccurate speech model for speech models been produced at the existing system with the use of the proposed system may constitute a robust model for speech recognition. The proposed method shows the improved recognition accuracy in a speech recognition system. [ABSTRACT FROM AUTHOR]

Recently, with the sharp increase of digital visual media such as photographs, animations, and digital videos, it has been necessary to generate mosaic blocks in a static or dynamic image intentionally or unintentionally. In this paper, we suggest a new method for detecting mosaic blocks contained in a color image using boundary features. The suggested method first extracts Canny edges in the image and finds candidate mosaic blocks with the boundary features of mosaic blocks. The method then determines real mosaic blocks after filtering out non-mosaic blocks using geometric features like size and elongatedness features. Experimental results show that the proposed method can detect mosaic blocks robustly rather than other methods in various types of input images. [ABSTRACT FROM AUTHOR]

Parkinson’s disease is the second most common degenerative brain disease after Alzheimer’s in old age. Symptoms of Parkinson’s disease are factors that reduce the quality of life in daily life, such as shaking hands, slowing behavior and cognitive function. Parkinson’s disease that can slow the progression of the disease through early diagnosis. To diagnoze Parkinson’s disease early, an algorithm was implemented to extract features using wav2vec and to diagnose the presence or absence of Parkinson’s disease with deep learning(ANN). As a results of the experiment, the accuracy was 97.47%. It was better than the results of diagnosing Parkinson’s disease using the existing neural network. The audio voice file could simply reduce the experiment process and obtain improved results. [ABSTRACT FROM AUTHOR]