Volume: 14
Number: 1
December 2018

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

• Performance Comparison of Feature Descriptors in Offline Signature Verification
  ABM Ashikur Rahman, Md Redwan Karim, Rafsanjany Kushol and Md. Hasanul Kabir
  Abstract PDF
  Performance Comparison of Feature Descriptors in Offline Signature Verification
  ABM Ashikur Rahman, Md Redwan Karim, Rafsanjany Kushol and Md. Hasanul Kabir

  Abstract

  Handwritten Signature is a widely used biometric in daily life as a mean of identity verification of an individual. For offline signature verification both accuracy and speed are important parameters. Accuracy may vary as the samples from signature datasets show a high intra-class variability. As these properties depend on the feature descriptor taken to represent the signature image, this selection is very important. In this study we provide a comparative performance evaluation of wellknown histogram based descriptors like SIFT and SURF and a wide variety of binary descriptors like BRIEF, ORB, BRISK and FREAK in the application of handwritten signature verification. We compare the performance of these feature descriptors against speed and accuracy. After the experimental analysis we have observed that binary features like ORB is faster with moderate accuracy but SIFT-like descriptors give better accuracy. Among them the combination of FAST feature detection and BRIEF descriptor is the fastest one but with lowest accuracy.
• Towards A Smartphone Based Lifelogging System for Reminiscence
  Md. Abed Rahman, A. M. Esfar E Alam, Md. Hasan Mahmud and Md. Kamrul Hasan
  Abstract PDF
  Towards A Smartphone Based Lifelogging System for Reminiscence
  Md. Abed Rahman, A. M. Esfar E Alam, Md. Hasan Mahmud and Md. Kamrul Hasan

  Abstract

  The proliferation of mobile devices have enabled us with the capability to store and make sense of passively gathered records of everyday human activities. This is called lifelogging. Every mobile device currently comes with a range of sensing abilities which includes but not limited to a camera, accelerometer, GPS and digital compass. All of these can be used to gather data unobtrusively. This data can then be made sense of by leveraging cloud computing. In this work, we build towards making a complete smartphone based lifelogging system, one that unobtrusively saves data as well as shows how it can be used to benefit the user. Another focus of lifelogging is to help users remember past events i.e. reminiscence. However, being able to find a good way to trigger memories is a challenge in itself. We explore the possibility of using music and background noise as a memory recalling tool and see the implications that it can have on reminiscence in a smartphone based lifelogger.
• Computational investigation of pulsatile blood flow through sinusoidal stenoses of varying lengths
  Md. Abdul Karim Miah, Ifat Rabbil Qudrat Ovi, Nasimul Eshan Chowdhury, Shorab Hossain, and Md. Nurul Absar Chowdhury
  Abstract PDF
  Computational investigation of pulsatile blood flow through sinusoidal stenoses of varying lengths
  Md. Abdul Karim Miah, Ifat Rabbil Qudrat Ovi, Nasimul Eshan Chowdhury, Shorab Hossain, and Md. Nurul Absar Chowdhury

  Abstract

  Numerical simulations have been performed for two dimensional axisymmetric, laminar, pulsatile flow through modelled, sinusoidal stenosis with a mean Reynolds number of 578 with a maximum of 938 and minimum of 328. Blood has been considered as the Newtonian fluid. Investigations have been carried out for different lengths of the sine curve forming the stenosis shape and compared for the same severity of 56%. Input sinusoidal pulse of 2.9 Hz, corresponding to 345 milliseconds of time period, has been applied. Womersley number of 7.75 has been considered in the flow. Radial velocity distribution, vorticity and wall shear stress (WSS) distribution have been taken as the key parameters for analyzing and comparing stenosis of varying lengths. It has been concluded for the considered sinusoidal shaped stenosis that with the decrease in the lengths of the stenosis the vorticity, wall shear stress etc. gets concentrated towards the center of the stenosis and the peak values get higher indicating high risk in stenosis with smaller lengths.
• Printed Bangla Character Image Segmentation: A Font Invariant Approach
  Muhammad Asif Hossain Khan, Anindya Sundar Paul and Muhammad Jawad Iqbal
  Abstract PDF
  Printed Bangla Character Image Segmentation: A Font Invariant Approach
  Muhammad Asif Hossain Khan, Anindya Sundar Paul and Muhammad Jawad Iqbal

  Abstract

  Optical character recognition is the technology which enables conversion of photographed text into searchable and editable documents. The complex nature of the feature set of Bangla characters has made it quite difficult to design a fairly accurate OCR. There have been multiple OCR solutions for Bangla based on template matching and deep learning, but none of them have achieved industrial grade accuracy. To date Google’s OCR engine Tesseract is one of the best performing OCRs for Bangla. OCR solutions have two major parts: segmentation and recognition, with segmentation being the more challenging part. In this research work we focused on improving the segmentation module of Tesseract, identifying issues unique to Bangla and resolving few of them. The proposed method successfully segments five vowel modifiers from their consonant bases where Tesseract fails. Experiment results conducted using five different fonts reveal that our proposed segmentation algorithm USHA shows notable improvement over Tesseract in segmenting Bangla scripts.
• Fingertip Detection and Finger Identification Approach for Hand Gesture Recognition using Microsoft Kinect
  Ahmed Al Marouf, Md. Kamrul Hasan and Hasan Mahmud
  Abstract PDF
  Fingertip Detection and Finger Identification Approach for Hand Gesture Recognition using Microsoft Kinect
  Ahmed Al Marouf, Md. Kamrul Hasan and Hasan Mahmud

  Abstract

  Hand gesture recognition is one of the recent research attractions in human computer interaction (HCI) and augmented reality (AR). Recognizing different hand gestures depend mostly on the posture of the hand which varies based on the orientation of fingers. To classify them, the orientation of fingertips and the names of individual finger are vital features to be considered. The most challenging task in hand gesture recognition is to determine the palm point and use the point to find all the fingertips. In this paper, a novel approach is proposed to determine the center of the palm and to detect the fingertips. The procedure of fingertip detection includes adaptive Hill Climbing algorithm applied on distance graphs. This paper also propose a novel approach for finger identification based on the relative distances among the fingertips and valley points. The experimental results shows up to 94% accuracy based on its inputs.