Volume: 13
Number: 1
December 2016

To see an article directly, click its Title. To see abstract, click its [Abstract] link.

• A Systematic Literature Review on Segmentation and Recognition of Printed Bangla Characters
  Umme Hafsa Billah and Muhammad Asif Hossain Khan
  Abstract PDF
  A Systematic Literature Review on Segmentation and Recognition of Printed Bangla Characters
  Umme Hafsa Billah and Muhammad Asif Hossain Khan

  Abstract

  Optical character recognition (OCR) is a system for recognizing characters from a pictorial format such as from printed documents and representing them in editable digital format. A fully functional OCR is a compulsory tool for any language as it is the primary tool for developing a comprehensive corpus for that language. Such a corpus is the cornerstone for building every statistical natural language processing tool for a language. Though, industrial grade OCRs have been developed for almost all major languages such as English, French, Chinese, Japanese etc., it is yet to be developed for Bangla – a language spoken by almost 250 million people. The convoluted structure containing curvature shapes in Bangla script has made the development of a Bangla OCR particularly difficult. Over the past few decades many researchers have worked on different components of Bangla OCR, namely pre-processing of text images, segmentation of lines, words and characters, and recognition of segmented characters. Systematic analysis of numerous published literature reveals that there is hardly any synchronization among these works, which led to repeated handling of the same problem from scratch by different researchers and producing solutions with near similar accuracy. In this paper, we have tried to systematically organize the notable works on different components of Bangla OCR. We have listed the problems the earlier works tried to handle, the methodologies they adopted, the solutions they proposed, the accuracies they could achieve and they problems they identified which they could not handle. We hope that such a comprehensive literature review will help the future researchers have a clear picture on the progress of research in this field and help them to contribute to the unsolved problems or problems yet to have a satisfactory solution.
• A comprehensive survey of code offloading mechanisms for mobile cloud computing
  Mohammad Erfan , Bidoura Ahmad Hridita, Mohammad Shoyaib, and Md. Shariful Islam
  Abstract PDF
  A comprehensive survey of code offloading mechanisms for mobile cloud computing
  Mohammad Erfan , Bidoura Ahmad Hridita, Mohammad Shoyaib, and Md. Shariful Islam

  Abstract

  Resources, battery lifetime and storage ca- pabilities of mobile devices are affected by the compute intensive, resource intensive or energy drain applica- tions. These limitations of the mobile devices may be mitigated with the help of cloud computing by delegating the energy drain or computing intensive tasks to more resourceful servers and receiving the result from the server. This process (a.k.a code offloading) helps to increase performance and reduce energy consumption. We present a detailed survey of the state of the art code offloading mechanisms for mobile based applications to better understand the domain of code offloading. The paper provides a high level design architecture, proce- dure for code offloading and presents a comparison of existing mechanisms based on the level of execution, level of profiling, security requirements, code profiler and offloading adaptation context. This paper also highlights major research challenges related to code offloading and provide recommendations to take better decision on code offloading in order to build more powerful mobile applications.
• A Scenario Based API Recommendation System Using Syntax and Semantics of Client Source Code
  S. M. Shahnewaz, Husne Ara Rubaiyeat, Hasan Mahmud, Md. Kamrul Hasan
  Abstract PDF
  A Scenario Based API Recommendation System Using Syntax and Semantics of Client Source Code
  S. M. Shahnewaz, Husne Ara Rubaiyeat, Hasan Mahmud, Md. Kamrul Hasan

  Abstract

  In software development, developers fre- quently look for the examples and documentation pro- vided by vendor of application programming interface (API) libraries, forums, textbooks and unofficial web- sites. However, a vast number of example API usage scenarios, that are embedded in the billions of line of already developed code are largely unexploited. In this research work, we present an approach that analyses the syntax and semantics of already developed source code to extract API usage scenarios for the current structural context of the developer. Our approach can extract an API usage scenario merging heterogeneous API elements. We also developed a scenario ranking al- gorithm based on the interrelationships of API elements. In the course of scenario extraction and ranking we have developed some heuristics based on developers coding experience. We demonstrate a qualitative evaluation by reporting a user study involving users completing four development tasks. Results show that our developed tool takes around 40% less time compared to existing systems. In addition, we present the precision-recall based quantitative evaluation. The high precision and recall values indicate that our proposed system is able to retrieve most of the best-fit API usage scenarios from candidate source files. Therefore, experimental evaluation provide evidence that scenario based recom- mendation is appropriate to help developers.
• RIFE-MAC: Fair and Efficient Medium Access Control for IEEE 802.11 Based Wireless Mesh Networks
  Ashraful Alam Khan, Mahmudun Nabi, Abdullah-Al-Tariq, and Muhammad Mahbub Alam
  Abstract PDF
  RIFE-MAC: Fair and Efficient Medium Access Control for IEEE 802.11 Based Wireless Mesh Networks
  Ashraful Alam Khan, Mahmudun Nabi, Abdullah-Al-Tariq, and Muhammad Mahbub Alam

  Abstract

  Earth-moving improvements in communication technologies in recent years have given rise to exponential surge in high data-rate Wi-Fi applications alongside the IEEE 802.11 based Wireless Mesh Networks (WMNs) that carry the inherent promise of cost-efficient last mile broadband wireless connectivity by extending the range of IEEE 802.11 Wi-Fi beyond the one hop range. Yet IEEE 802.11 based WMN technology is not devoid of some treacherous shortcomings and challenges, mainly in the Medium Access Control (MAC) sub-layer. This is particularly because of MAC sub-layer overheads which abstract the effective exploitation of increased agility ensured by today’s physical layer (PHY). Besides, the multi-hop nature of WMNs triggers few more obstacles to medium access mechanism. The overhead of high collision probability and long idle listening during channel contentions, have restricted the expected improvement in data throughput efficiency. Moreover the IEEE 802.11 based WMNs exhibit severe unfairness and starvation of flows for gateway- destined traffics that are being originated from multi- hop stations. Such challenges in recent years have made IEEE 802.11 based WMNs an enticing research topic. In this paper we propose Receiver-Initiated Fair and Efficient MAC (RIFE-MAC), a novel distributed channel access mechanism for IEEE 802.11 based WMNs which not only efficiently maximizes the overall network throughput but also improves the fairness in channel access among flows in WMNs. To attain these desired goals, RIFE-MAC significantly reduces the collision probability, idle time in between consecutive transmissions and unfairness in channel access. We investigate the performance of RIFE-MAC through extensive simulation. Simulation results demonstrate the effectiveness and robustness of our approach compared to IEEE 802.11 MAC and few more relevant works.