Title of ePoster... M. Author1, D. Author2, M. Author3 November 15 - 16, 2018, Grand Hotel Starý Smokovec, the High Tatras, Slovakia Title of ePoster... M. Author1, D. Author2, M. Author3 Abstract — In this paper, the research on the identification of a person in the video broadcast through the IP stream is presented. The software solution is created based on the analysis of face detection methods and current related researches. The crucial part is the identification of individual (university staff) in the IP stream. Training is carried out prior to the identification process. The software solution in form of the web application is created to enable live streaming. It further promotes the functionalities of discussion and voting on questions. System administrator interface is created to allow control over person identification and management of questions. Another essential part focuses on the processing and storing data from discussion or voting. The research verifies the reliability of identified persons through the series of experimental tests. I. INTRODUCTION The presented research is focused on the development of a system for enabling the live broadcast from assembly of Poprad city council or may be applied to any other analogous event related to education. Proposed system incorporates automatic PTZ camera system control being based on the active speaker. Broadcasting per se is to be accessible through web interface or dedicated mobile applications. These applications will allow citizens to engage in the current process through discussion and voting. The main objective of the research is identification of objects, specifically human beings. As the object of interest, the human face is selected. The identification is to be deployed on the video streamed through the IP stream. The methods and procedures for face recognition are analyzed and also deployed in the created solution. Another part of this research is proposal of a web solution for broadcasting of live stream from the council gathering. II. OPEN CV LIBRARY The library called OpenCV [1] is built on a common infrastructure designed for the development of computerbased applications. Library is released under a BSD license and hence it is free for both, academic and commercial user. While utilizing over 2500 algorithms, it offers a wide range of tools for recognition of faces, Identification of objects, capturing human motion in a video, recording object movement, finding similar images from an image database, eye tracking etc. Further, it implements the eigenface algorithm for face recognition. A. Face recognition using Eigenfaces algorithm In general, the algorithm selects the relevant information from the image and encodes it as efficiently as possible. In case of face recognition, a set of images of the same individual is evaluated for differences. The algorithm calculates eigenvectors from a covariance matrix from a set of face images [2]. Each image from the set containing the face creates an eigenvector. Vectors define the differences between the frames. A set of vectors is referred to as eigenfaces. Each face can be represented by a linear combination of these eigenfaces, where their amount can be reduced to those with the highest values and hence higher system efficiency can be achieved. The goal of the algorithm is to develop a system that does not compare the images to each other but rather their characteristic weights. The algorithm may be implemented through the following steps: 1. Formation of a face-image database, calculation of the eigenfaces and determination of the face space. 2. Calculation of its set of weights, once a new image is found. 3. Determination if the face is present in the image. 4. Finally, checking if the image corresponds to a known face stored in database. VI. CONCLUSION The goal of presented research was development of a solution for identification of the person visible in the broadcast through the IP stream. The paper analyses two libraries capable of face recognition that both use different algorithm for the recognition process. Another part is considered with the identification of persons in the IP stream. This includes development of experimental software solution consisted of three separate parts. The result of this research includes also testing of the recognition reliability using the chosen solution. From the measured data, charts are formed and consider several obstructions that may appear in the overall recognition. ACKNOWLEDGMENT This publication is the result of the Project implementation: University Science Park for Innovation Applications Supported by Knowledge Technology, Phase II., ITMS: 333333333333, supported by the Research & Inovation Operational Programme funded by the ERDF. We support research activities in Slovakia/This project is being co-financed by the European Union. ICETA 2018 Conference Office elfa, s.r.o. Park Komenského 7, 040 01 Košice Slovakia tel.: + 421-55-625 3839 fax: + 421-55-726 5195 e-mail: iceta@elfa.sk www.iceta.sk Back to Schedule