Papers by Mariofanna Milanova
International Journal of Advanced Computer Science and Applications, 2018
A new method to detect human health-related actions (HHRA) from a video sequence using an Android... more A new method to detect human health-related actions (HHRA) from a video sequence using an Android camera. The Android platform works not only to capture video images through its camera, but also to detect emergency actions. An application for HHRA is to help monitor unattended children, individuals with special needs or the elderly. The application has been investigating based on TensorFlow Object Detection Application Program Interface (API) technique with Android studio. This paper fundamentally focuses on the comparison, in terms of improving speed and detection accuracy. In this work, two promising new approaches for HHRA detection has been proposed: SSD Mobilenet and Faster RCNN Resnet models. The proposed approaches are evaluated on the NTU RGB+D dataset, which it knows as the present greatest publicly accessible 3D action recognition dataset. The dataset has been split into training and testing dataset. The total confidence scores detection quality (total mAP) for all the actions classes are 95.8% based on the SSD-Mobilenet model and 93.8% based on Faster-R-CNN-Resnet model. The detection process is achieved using two methods to evaluate the detection performance using Android camera (Galaxy S6) and using TensorFlow Object Detection Notebook in terms of accuracy and detection speed. Experimental results have demonstrated valuable improvements in terms of detection accuracy and efficiency for human healthrelated actions identification. The experiments have executed on Ubuntu 16.04LTS GTX1070 @ 2.80GHZ x8 system.
With the fast evolution of medical imaging study, a great interest in skin cancer detection has b... more With the fast evolution of medical imaging study, a great interest in skin cancer detection has been investigated with numerous computer algorithms. Generally, skin lesions are examined with a limited quantity of ground truth labeling. The most important part of the medical image's detection is calculating the localization function which is normally evaluated on the Intersection over Union threshold (IoU). It helps to locate the lesion accurately to collect dominant features of the skin lesion. In this work, an object localization for skin lesion detection has been proposed using SSD- Mobilenet model on ISIC 2018 as a training and testing datasets. To evaluate the detection performance, the detection process has been achieved using two different methods; a real-time mobile application of Android camera (Galaxy S6), and Jupyter Notebook of TensorFlow Object Detection Application Program Interface (API). The total confidence scores detection quality (total mAP) is 96.04% with a to...
International Journal of Advanced Computer Science and Applications, 2017
There is a great benefit of Alzheimer disease (AD) classification for health care application. AD... more There is a great benefit of Alzheimer disease (AD) classification for health care application. AD is the most common form of dementia. This paper presents a new methodology of invariant interest point descriptor for Alzheimer disease classification. The descriptor depends on the normalized Hu Moment Invariants (NHMI). The proposed approach deals with raw Magnetic Resonance Imaging (MRI) of Alzheimer disease. Seven Hu moments are computed for extracting images' features. These moments are then normalized giving new more powerful features that highly improve the classification system performance. The moments are invariant which is the robustness point of Hu moments algorithm to extract features. The classification process is implemented using two different classifiers, K-Nearest Neighbors algorithm (KNN) and Linear Support Vector Machines (SVM). A comparison among their performances is investigated. The results are evaluated on Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The best classification accuracy is 91.4% for KNN classifier and 100% for SVM classifier.
International Journal of Advanced Computer Science and Applications, 2017
This paper presents a new model of scale, rotation, and translations invariant interest point des... more This paper presents a new model of scale, rotation, and translations invariant interest point descriptor for human actions recognition. The descriptor, HMIV (Hu Moment Invariants on Videos) is used for solving surveillance camera recording problems under different conditions of side, position, direction and illumination. The proposed approach deals with raw input human action video sequences. Seven Hu moments are computed for extracting human action features and for storing them in a 1D vector which is constringed as one mean value for all the frames' moments. The moments are invariant to scale, translation, or rotation, which is the robustness point of Hu moments algorithm. The experiments are evaluated using two different datasets; KTH and UCF101. The classification process is executed by calculating the Euclidean distance between the training and testing datasets. Human action with minimum distance will be selected as the winner matching action. The maximum classification accuracy in this work is 93.4% for KTH dataset and 92.11% for UCF101.
2006 IEEE International Conference on Granular Computing, 2006
Signal decomposition techniques prove to be useful in the analysis of neural activity, as they al... more Signal decomposition techniques prove to be useful in the analysis of neural activity, as they allow for identification of supposedly distinct neuronal structures (i.e., sources of activity). Applied to measurements of brain activity in a controlled setting as well as under exposure to an external stimulus, they allow for analysis of the impact of the stimulus on those structures. The link between the stimulus and a given source can be confirmed by a classifier that is able to "predict" if a given signal was registered under one or the other condition, solely based on the components. Very often, however, statistical criteria used in traditional decomposition techniques turn out to be insufficient to build an accurate classifier. Therefore, we propose to utilize a novel hybrid technique based on multi-objective evolutionary algorithms (MOEA) and rough sets (RS) that will perform decomposition in the light of the classification problem itself.
This paper presents a new algorithm for human action recognition in videos. This algorithm is bas... more This paper presents a new algorithm for human action recognition in videos. This algorithm is based on a combination of two different feature types extracted from Aligned Motion Images (AMIs). The AMI is a method for capturing the motion of all frames in a human action video in one image. The first feature is a contourbased type and is employed to grasp boundary details of the AMI. It relies on the 1st and 2nd discrete time differential of the chorddistance signature feature, so it is called Derivatives of ChordDistance Signature (DCDS). The second feature is a silhouette-based type that is used to capture regional appearance details. It catches most of the visual components for the AMI using a Histogram of Oriented Gradients (HOG) feature. Combining both features creates a complementary feature vector that makes it possible to obtain an optimal correct recognition rate of 100%. For the classification, the algorithm is utilized two different classifiers: K-Nearest-Neighbor (KNN) and...
This paper explains research based on improving real time face recognition system using new Radix... more This paper explains research based on improving real time face recognition system using new Radix-(2 × 2) Hierarchical Singular Value Decomposition (HSVD) for 3rd order tensor. The scientific interest, aimed at the processing of image sequences represented as tensors, was significantly increased in the last years. Current home security solutions can be cost-prohibitive, prone to false alarms, and fail to alert the user of a break-in while they are away from the home. Because of advancements in facial detection and recognition techniques made in the past decade, we propose a home security system that takes advantage of this technology. To create such a system at a low cost requires algorithms that are powerful enough to detect users in various environmental conditions and fast enough to process real time video on weaker hardware. Experiments comparing the efficiency of two different decomposition techniques applied for face recognition in real time.
This report is published in the interest of scientific and technical information exchange, and it... more This report is published in the interest of scientific and technical information exchange, and its publication does not constitute the Government's approval or disapproval of its ideas or findings. i Standard Form 298 (Rev. 8-98) Prescribed by ANSI Std. 239.18 REPORT DOCUMENTATION PAGE Form Approved OMB No. 0704-0188 Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing this collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
Proceedings of the 9th International Conference on Neural Information Processing, 2002. ICONIP '02., 2002
This paper presents a novel approach to classification of decomp osed cortical evoked potentials ... more This paper presents a novel approach to classification of decomp osed cortical evoked potentials (EPs). The decomposition is based on learning of a sparse set of basis functions using an Artificial Neural Network (ANN). The basis functions are generated according to a probabilistic model of the data. In contrast to the traditional signal decomposition techniques (i.e. Principle Component Analysis or Independent Component Analysis), this allows for an overcomplete representation of the data (i.e. number of basis functions that is greater than the dimensionality of the input signals). Obviously, this can be of a great advantage. However, there arises an issue of selecting the most significant components from the whole collection. This is especially important in classification problems based upon the decomposed representation of the data, where only those components that provide a substantial discernibility between EPs of different groups are relevant. To deal with this problem, we propose an approach based on the Rough Set theory's (RS) feature selection mechanisms. We design a sparse coding-and RS-based hybrid system capable of signal decomposition and, based on a reduced component set, signal classification.
2015 International Symposium on Innovations in Intelligent SysTems and Applications (INISTA), 2015
The paper presents a new technique for archiving and protecting content of visual medical informa... more The paper presents a new technique for archiving and protecting content of visual medical information. A special format is developed based on a new Inverse Pyramid decomposition. The images are archived with the highest quality but their restoration is performed in accordance with the application. The image content is protected by inserting multiple fragile watermarks that can be extracted by authorized users only. The fragile watermark is inserted as additional decomposition layer and does not influence the image quality. This approach permits the creation of archiving systems with hierarchical access control.
Proceedings of the 17th International Conference on Pattern Recognition, 2004. ICPR 2004., 2004
International Journal of Computer Science, Engineering and Applications, 2013
Morphological active contours for image segmentation have become popular due to their low computa... more Morphological active contours for image segmentation have become popular due to their low computational complexity coupled with their accurate approximation of the partial differential equations involved in the energy minimization of the segmentation process. In this paper, a morphological active contour which mimics the energy minimization of the popular Chan-Vese Active Contour without Edges is coupled with a morphological edge-driven segmentation term to accurately segment natural images. By using morphological approximations of the energy minimization steps, the algorithm has a low computational complexity. Additionally, the coupling of the edge-based and region-based segmentation techniques allows the proposed method to be robust and accurate. We will demonstrate the accuracy and robustness of the algorithm using images from the Weizmann Segmentation Evaluation Database and report on the segmentation results using the Sorensen-Dice similarity coefficient.
Studies in Computational Intelligence, 2005
This paper presents a new image compression method based on the Inverse Difference Pyramid (IDP) ... more This paper presents a new image compression method based on the Inverse Difference Pyramid (IDP) decomposition, and one specific application of this method aimed at layered image transfer via a standard communication networks. A basic feature of the IDP method ...
2011 IEEE International Conference on Information Reuse & Integration, 2011
The Cellular Neural Networks (CNN) model is now a paradigm of ceIlular analog programmable multid... more The Cellular Neural Networks (CNN) model is now a paradigm of ceIlular analog programmable multidimensional processor array with distributed local logic and memory. CNNs consist of many paraIlel analogue processors computing in real time. One desirable feature is that these processors arranged in a two dimensional grid, only have local connections, which lend themselves easily to VLSI implementations. The connections between these processors are determined by a cIoning template, which describes the strength of nearest-neighbour interconnections. The cloning templates are spaceinvariant, meaning that a11 the processors have the same relative connections. In this paper first we describe the architecture of CNN. Next, a new application of CNN using them for the 3D scene analysis is studied. .
In this paper we present a new approach for face recognition using Singular Values Decomposition ... more In this paper we present a new approach for face recognition using Singular Values Decomposition (SVD) to extract relevant face features and seven states Hidden Markov Model (HMM) as classifier. The SVD-HMM system has been evaluated on two databases-the Olivetti Research Laboratory (ORL) face database and YALE database. In order to gain more speed and higher recognition rate effective modifications of the original images are proposed.
This paper investigates the applicability of the Neural Network approach for image quality assess... more This paper investigates the applicability of the Neural Network approach for image quality assessment. The aim is to predict the subjective quality score, namely the difference mean opinion score (DMOS) obtained from human observers, by incorporating a neural network algorithmic approach utilizing extracted statistical features from test and original images. To ease this approach, a MATLAB user interface is developed and presented here. To validate the proposed approach, an image database is selected consisting of various distortion types as test bed in which a DMOS value is provided for each distorted image. Experimental results show that the obtained output of Neural Network correlates well with DMOS values and Neural Network can mimic human observers.
Uploads
Papers by Mariofanna Milanova