http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
- 中文 을 입력하시려면 zhongwen을 입력하시고 space를누르시면됩니다.
- 北京 을 입력하시려면 beijing을 입력하시고 space를 누르시면 됩니다.
RISS 인기검색어
- 검색키워드
- 저자 : Muhammad Haroon Yousaf (검색결과 3 건)
- 무료
- 기관 내 무료
- 유료
-
HMM-based Scheme for Smart Instructor Activity Recognition in a Lecture Room Environment
Asim Raza,Muhammad Haroon Yousaf,Hassan Ahmed Sial,Gulistan Raja 한국산학기술학회 2015 SmartCR Vol.5 No.6
Instructor activity recognition can certainly play its part as an important parameter in evaluating and improving the performance of an instructor. This paper presents a single-layered sequential approach for instructor activity recognition in the lecture room environment. A hidden Markov model (HMM) scheme is selected as a sequential approach for activity recognition. The proposed system incorporates the five major activities of the instructor in the lecture room, i.e. walking, writing, pointing towards the board, standing, and pointing towards presentations. Background/foreground modelling is carried out using a Gaussian mixture model (GMM) for instructor detection in the lecture room. Mesh features are selected to represent the instructor. After vector quantization, features are passed to the HMM for activity recognition. Time is tracked, and the occurrences of each activity are counted to elaborate on the activities the instructor performed during the lecture. The proposed scheme proved to be efficient owing to its high accuracy rate of over 90 percent in recognizing five different activities of an instructor as tested in a MATLAB simulation environment.
-
Video augmentation technique for human action recognition using genetic algorithm
Nudrat Nida,Muhammad Haroon Yousaf,Aun Irtaza,Sergio A. Velastin 한국전자통신연구원 2022 ETRI Journal Vol.44 No.2
Classification models for human action recognition require robust features and large training sets for good generalization. However, data augmentation methods are employed for imbalanced training sets to achieve higher accuracy. These samples generated using data augmentation only reflect existing samples within the training set, their feature representations are less diverse and hence, contribute to less precise classification. This paper presents new data augmentation and action representation approaches to grow training sets. The proposed approach is based on two fundamental concepts: virtual video generation for augmentation and representation of the action videos through robust features. Virtual videos are generated from the motion history templates of action videos, which are convolved using a convolutional neural network, to generate deep features. Furthermore, by observing an objective function of the genetic algorithm, the spatiotemporal features of different samples are combined, to generate the representations of the virtual videos and then classified through an extreme learning machine classifier on MuHAVi-Uncut, iXMAS, and IAVID-1 datasets.
-
Hadia Bashir,Fawad Hussain,Muhammad Haroon Yousaf 한국산학기술학회 2015 SmartCR Vol.5 No.3
A human body consists of a complex 3D structure. Conversion of 3D structures into 2D leads to a loss of information and may result in incorrect disease diagnosis. This issue has grasped the attention of researchers involved in 3D modeling. MRI scans consist of a large number of 2D slices, which makes 3D reconstruction a complex and time-consuming task. We propose an efficient algorithm that uses limited MRI slices to reconstruct a 3D image on the basis of matching criteria, which aids in the selection of most appropriate slices, which therefore significantly reduces computational complexity and increases accuracy. The methodology involves the acquisition of a brain MRI, pre-processing, OTSU’s segmentation for the identification of suspicious areas, and rule-based classification to extract a tumor area. For appropriate slice selection, Rapid Mode image matching is utilized, 3D modeling is performed using a cubic reconstruction scheme, and finally the tumor volume is calculated. Performance of proposed work is tested on the XNAT datasets of 21 patients. We achieved 96.6% accuracy and concluded that it can be efficiently used in all clinical applications.
ScienceON연관 검색어 추천
이 검색어로 많이 본 자료
활용도 높은 자료
