Automatic Human Motion Detection and Tracking in Complex and Dynamic Environment
Phyo Phyo Aung 1, Cho Cho Myint 2
Department of Information Technology, West Yangon Technological University Yangon, Myanmar
Detecting human beings accurately in a video surveillance system is crucial for diverse application areas for the secure environments. The main objective of the paper is to detect and track motion of a person in a certain area where any persons are prohibited to enter. The first step of the detection process is to detect an object which is in motion. Motion object detection could be performed by background subtraction using Gaussian Mixture Model (GMM), morphological operations and median filtering. Based on geometrical features such as boundingbox and regional area, of the foreground objects in each frame image, they are determined whether human or not. If there cannot be made any decision for human motion, the foreground objects are cropped and sent to support vector machine (SVM) to classify human motion. The accuracy of the proposed method is measured with many experiments on different environments.
Keywords - human motion, Gaussian Mixture Model, Support Vector Machine, complex and dynamic environment, motion detection and tracking
A Secret Text Message Hiding Technique Using Indexes Based Text in Text Steganography
Thu Zar Win1, Aye Wai Oo2, Cho Cho Myint3
Department of Information Technology Engineering, West Yangon Technological University, Myanmar
Steganography is one of the art of cryptography and data hiding for secret hidden writing. This paper is one of the text in text steganography to hide secret text message in white space of cover text using indexing approach. There are two main sections, embedding section and extraction section. In the encoding section, an array is firstly constructed equally with the length of cover text and all cover texts except white spaces are assigned with 1’s in this array. As the second, the characters in the secret text message which are the same with characters in the cover text are searched and retrieved their indexes of cover text. All cover indexes are assigned by appending 9. Other characters of the message are also changed into ASCII indexes. These indexes are converted into octal numbers. These octal numbers are put in the position of white spaces of the array. Next, a random number array which is the same length with the index array is constructed. After that the two arrays are combined and merged in the cover text. In the message extraction process, the random array is subtracted from the combined array to produce the index array. From white space positions of the index array, all characters of the secret message are recovered by the reverse of embedding process. The performances of the proposed technique will be presented and discussed with various experiments.
Keywords – text in text steganography, cover index, ASCII index, stego text, white spaces
Improvement in AODV Routing Protocol for VANET (Vehicular Ad Hoc Network)
May Zar Win#1, Khin Khat Khat Kyaw*2, Cho Cho Myint#3
#Information Technology Department, West Yangon Technological University Yangon, Myanmar
Vehicular Ad Hoc Network (VANET) is a part of Mobile Ad Hoc Network (MANET).VANET is the emerging technology which enables the vehicles to communicate using wireless technologies.VANET protocols are important for road-safety, traffic efficiency and management applications. In the mobile Ad Hoc networking paradigm, there is no fixed infrastructure and packets are delivered to their destinations through wireless multi hop connectivity. Among the current routing protocols, reactive routing protocols are favoured in MANET because they help in reducing overheads by continuously sending the data for better communication. Many researchers have analyzed the popular reactive routing protocols such as AODV, DSDV, DSR, PUMA and TORA. Among these protocols, Ad-hoc On-demand Distance Vector (AODV) is the most popular and robustness routing protocol. Nowadays when AODV is used in VANET, there are a lot of weak points to overcome since VANET uses inter vehicle communication rather than vehicle to road side unit communication. Reducing routing overheads and bandwidth consumption are important issues to make AODV usable for VANET. For these issues, we have changed route discovery phase and route maintenance phase by optimization of route time-outs, hello mechanism and expanding ring search based on TTL. By using NS2 simulation, we show that proposed protocol can reduce network end-to-end delay and routing overhead. Moreover, it also increases packet delivery ratio and throughput.
Keywords - VANET, MANET, Ad-hoc, AODV, NS2.
Performance Analysis of DSDV Routing Protocol in MANET
Daw Nge , Dr.Khin Khat Khat Kyaw
Information Technology Department, West Yangon Technological University Myanmar
Since the mobile applications are widely used in human society, the efficiency of Mobile ad hoc network (MANET) needs better solutions. The topology of network dynamically changes in real world applications. Therefore, the routing protocols used in MANET must be fast and save. Consequently, the performance analysis of MANET routing protocols becomes an interesting case in networking research area. In this paper, the performance of Dynamic Destination-Sequenced Distance-Vector Routing (DSDV) routing protocol has been analysed in term of throughput, end-to-end delay, packet delivery ratio (PDR) and packet loss.
Keywords –MANET, DSDV, NS2, routing, PDR, throughput
Optimization of Energy Efficient MAC Layer Protocol for Mobile Ad Hoc Network
Su Su Hlaing#1, May Zin Oo*2
1Department of Information Technology, West Yangon Technological University, Myanmar
Medium Access Control (MAC) protocols is an important part in many channel access for the using of same frequency spectrum in all nodes of MANET. As these nodes normally operate depending on battery power, energy efficiency is a significant research issue. In this paper, a new approach, modification of MAC protocol (MAC-mod), is proposed with the purpose of optimize the average energy consumption by reducing the number of ACK packets. MAC-mod keeps nodes that are not involved in transmission in sleep mode to optimize energy consumption. The performance of original MAC protocol (MAC-ori) and MAC-mod is compared by measuring the average throughput and average energy consumption. Experiments are carried out by varying node density with Manhattan Grid (MG) mobility model. The simulations were carried out using Network Simulator (NS-2).
Keywords—ad hoc network, energy consumption, IEEE 802.11, mobility model, throughput
Evidence Effecting on Analysis Time in Database Forensics
Ohn Mar Myint #1, Mie Mie Su Thwin *2
Information Technology Department , West Yangon Technological University
The business related data of most organization and company are maintained in database system. Within database system, the misuse of data causes some consequences for company. Intruders are unauthorized both user and insider restricted users. To detect high privileged user’s misuse of database within organization is difficult. Database comprises the number of location where the evidence of fraudulent activity carried on database and their metadata. Database forensics analysis can help to determine intruder in system, the fraudulent activity carried on database system. Internal structures are used as a vital base line to collect evidence during a forensic investigation. This paper will contribute to a better understanding of database forensics and crucial evidence collection process of database forensics analysis, especially MySQL database because it is the most widely used for commercial and government unit in Myanmar. MySQL is ease of use for the speedy and cost-effective delivery of both new and improved web based applications. The overall benefit of this paper is that no additional logs are needed and we used replication process. Furthermore, our approach is invariant to retroactive malicious modifications by any attacker. This assures the authenticity of the evidence and strengthens the chain of custody.
Keywords - MySQL, database, forensics, logs, artifacts
Continuous Speech Recognition System Based on Deep Convolutional Neural Network for Myanmar
Yin Win Chit#1, Soe Soe Khaing*2, Yi Yi Myint#1
#1Faculty of Information and Communication Technology
University of Technology (Yatanarpon Cyber City), Pyin Oo Lwin, Myanmar,
Automatic Speech Recognition (ASR) system, that translates the speech signal into text words, is still a challenge in the continuous speech signal. Continuous speech recognition systems develop with four separated steps: segmentation of the speech signal, feature extraction, classification and recognizing the words. These steps can be modeled with the various methods. Among them, the combination model of the dynamic threshold based segmentation, Mel-Frequency Cepstral Coefficient (MFCC) feature extraction method and Deep Convolutional Neural Network (DCNN) is proposed in this paper. Especially, DCNN-AlexNet has been applied in image processing because it can perform as a highly accurate, effective and powerful classifier. In the training and classification step of this system, the advantages of DCNN in image processing are applied using the MFCC feature images. The main purpose of this system is to transform the MFCC features of the speech signal to MFCC features images with various frame size for three layers of input images of DCNN. The three layers 32*32*3 images are used for the input images of DCNN-AlexNet to support the recognition step of the system. The experiments shows that the DCNN speech recognition system achieves the average Word Error Rate (WER) of 11.5 % on the proposed MFCC images training dataset and WER of 13.75% on the MFCC features value matrix training dataset.
Keywords – Automatic Speech Recognition, Mel-Frequency Cepstral Coefficient, Deep Convolutional Neural Network, Word Error Rate
Hybrid Strength of Cryptography and Steganography for User Authentication
San San Newl1, KhaingKhaing Soe2, MyaMya Htay3, Nay Aung Aung4, Tin Tin Thein5
Information Technology Support and Maintenance Department
University of Computer Studies (Pakokku)
PakokkuTownship , Myanmar
Cryptography and steganography are the two popular methods available to provide security. They provide security in their own ways, but to add multiple layers of security it is always a good practice to use combination of these techniques. In this paper, user password is embedded within the cover media by using Least Significant Bit (LSB) steganography, an extra secure method in which to protect data. Then, the image is divided into two shares using visual cryptography where one of the shares is stored in the database and the other kept by the user. These shares are collected and stacked to get the original image. Then user password is de-embedded from the cover image and compared with user’s login password. Therefore, two levels of security have been provided using the proposed hybrid technique. In addition, the proposed system consists of three experiments. Firstly, we analyze the strength of passwords based on their lengths (8, 9, 10, 11, 12 characters) and composition using brute force attack. Then, the quality of images with peak signal-to-noise ratio (PSNR) and mean square error (MSE) between original images and stego-images is measured. Finally, the time complexities of registration and verification phase are examined.
Keywords: authentication, visual cryptography, LSB steganography, hybrid technique, password
Weather Forecasting Model for Growing Crops in Pakokku Township Using Artificial Neural Network (ANN)
Khaing Khaing Soe1, Mya Mya Htay2, San San Newl3, Nay Aung Aung4, Tin Tin Thein5
Information Technology Support and Maintenance Department
University of Computer Studies (Pakokku)
Pakokku Township , Myanmar
Weather forecasting system is one of the major important applications in agricultural field. Most of the crops depend upon weather condition. But weather forecasting system has been many challenging problem around the world. If a farmer knows the next year rainfall then the ideal crops to grow can be selected else, the crops might be damaged. In this paper, to overcome from that problem the system predicts rainfall based on weather dataset (rainfall, dry bulb, wet bulb, humidity, soil, dew point and temperature) from 2005 to 2014 from Meteorological and Hydrological Department in Pakokku. Multilayer Neural Network and Data Mining Techniques are used for predicting rainfall. This information supports for the farmers in Pakokku Township to get good profit about the growing crops.
Keywords- Artificial Neural Network, Data Mining Techniques, weather forecasting, profit, rainfall
Application of Graph Theory to Effective Travelling in Ayeyarwady Region
Nila Aung Khaing #1, Lin Lin Naing #2
# Faculty of Computing, University of Computer Studies, Hinthada Hinthada Township, Ayeyarwady Division, Myanmar
This paper presents the effective ways for travelling in Myanmar using Dijkstra’s Algorithm. It is an iterative algorithm and the basic idea is searching a graph by finding path, starting at a point, and exploring adjacent nodes from there until the destination node is reached. Generally, the goal is to obtain the shortest path to the destination. In this paper the well-known Dijkstra’s Single-Source Shortest Path (SSSP) algorithm is used to find optimal paths from one place to another, such as cities or other interesting places in Myanmar. This algorithm was reviewed with values which come from weights on edges according to actual situations on the road, such as costs, distances, and travel times of some famous cities in Ayeyarwady region. The key issue to be addressed in this work is to find the shortest paths from one source point (city) to others by comparing the weighted values (i.e., costs, distances and travel times) between any two different paths with their edge lengths (roads) that assigned by actual values for saving cost and time for effective travelling.
Keywords - Path Finding, Weighted Graph, Routing, Shortest Paths, Dijkstra’s Algorithm.
Performance Evaluation of SVM and NB in the Opinion Mining from Social Media
Thiri Yadanar#1, Htun Htun*2, Cho Cho Myint#3
#1 Information Technology Department, West Yangon Technological University Yangon, Myanmar
The proliferation of social network sites allows users to communicate with each other by using several tools like chats, discussion forums, comments etc. At any type, any category and any word user professionally feel free to express their emotion as comments. These all comments have some features and attributes with it. These comments or status are really useful which are actually viewed as their ‘OPINIONS’. Opinions are really important when we are analysing any of comments, views and emotions with images. Then, we introduce and present our original method of opinion classification and we test the presented algorithm on real world datasets, reporting on the results. The aim of our research is to select three features POSITIVE, NEGATIVE and NEUTRAL from these comments respectively. It performs evaluations experiments for each classifier results which can be worked for feature mining of user opinions on social media. Then, we presented the opinion classification and we test the real-world datasets, reporting on the results. The research is compared with the performance of Support Vector Machine (SVM) and Naïve Bayes (NB), will also be presented.
Keywords - Opinion Mining, Social media, Support Vector Machine, Naïve Bayes (NB), Text Analysis
Robust Abandoned Object Detection using Gaussian Mixture Model and Kalman Filtering
Kyi Kyi Win
Department of Information Technology West Yangon Technological University, Yangon, Myanmar
Today, the usage of intelligent based visual surveillance systems such as AI based CCTV are more and more increase to monitor the security of public area for keeping citizens safety. In the visual surveillance system, more than one AI based object detections are usually included. Among them, abandoned object detection is one of the most important and interesting processes in which suspected left object is detected and activate alarm system. The main objective of the paper is to detect security-suspected objects such as bags, boxes, luggages and etc. in public area. In this paper, background subtraction from input video sequence is performed by using Gaussian Mixture Model. Result foreground moving and stationary objects are checked and track using Kalman filter. Target stationary objects are verified by using multiclass support vector machine (SVM) and then declared as abandoned Objects. The accuracy of proposed system will be proved with various target objects in different locations.
Keywords – Background Subtraction, Abandoned object detection, Kalman filter, Gaussian Mixture Model, Support Vector Machine (SVM)
Comparison of Energy Efficiency in MAC Protocols for Mobile Ad Hoc Network
Su Su Hlaing
Department of Information Technology, West Yangon Technological University, Myamar
With the rapid development in wireless communication technologies, mobile ad hoc networks (MANETs) have emerged as an important part of the envisioned future ubiquitous communication. Energy-efficient design in MANETs is more important and challenging than with other wireless networks. The traffic loads in MANETs with the absence of an infrastructure are heavier than in other wireless networks with fixed access points or base stations, and thus MANETs have more energy consumption. Energy-efficient design needs to consider the trade-offs between different network performance criteria. In this paper, medium access control (MAC) protocol and MAC modification (MAC-mod) protocol are used. MAC-mod protocol reduces the number of ACK packets to optimize the energy consumption and puts the some nodes that are not included in transmission, go to sleep mode. Experiments are carried out by varying the node density with Reference Point Group (RPGM) and Random Waypoint (RWP) mobility models. The simulations are done by Network Simulator (NS-2).
Keywords – MANETs, IEEE 802.11, DCF, MAC, mobility models, energy-efficient