Jasjit S. Sodhi ¹, Maitreyee Dutta ², Naveen Aggarwal ³

Affiliations
1 Department of Biotechnology, Himachal Pradesh University, Shimla - 171005, Himachal Pradesh, IN
2 Department of Computer Science, N.I.T.T.T.R, Chandigarh - 160019, Punjab, IN
3 Department of Computer Science, U.I.E.T, Punjab University, Chandigarh - 160014, Punjab, IN

Abstract

Objectives: This paper presents the use of machine learning technique in order to eliminate the components affecting the human decision making process. To assist decision making in career counseling, an Artificial Intelligence model was implemented using Artificial Neural Network (ANN) in MATLAB for predicting vocational stream of pursuit based on the behavioral characteristic of the beneficiary. Methods/Statistical Analysis: The Differential Aptitude Test (DAT) battery; and Scientific Knowledge and Aptitude Test (SKAT) were used to assess an individual’s specific abilities in different areas. The training data set for the ANN model was procured in the form of normalized scores based on occupational/vocational profiles as given by the authors of DAT battery and SKAT. The trained ANN was tested with normalized stanine scores of 100 tenth class studentsraised through cluster random sampling technique for predicting a vocational stream of pursuit. In order to evaluate its accuracy and efficiency, three techniques of classification were employed. The unclassified data was classified by using Discriminant function analysis, ANN and two classifications were obtained from the trained counselors. Findings: The classification result obtained from the above mentioned techniques were compared and was found that the ANN system and Discriminant function analysis agreed approximately by 91% over all the test cases. The results of the statistical method support the classification made by the ANN. The two counsellors were in agreement with ANN’s classification output by approximately 81%. However, the counsellors disagreed with each other’s prediction approximately by 27% over all the test cases. The experimental results support the hypothesis that the proposed machine learning technique performed better than the prediction made by counsellors. Application/Improvements: In the Indian scenario, the developed machine learning system may be used as a standalone system in places where there is the paucity of counsellorsor can assist in the career decision making provided by human beings.

Keywords

Artificial Neural Network, Career Counseling, Decision Support System, Differential Aptitude Test Battery, Discriminant Function Analysis, Scientific Aptitude and Knowledge Test.

Full Text

   

Staffy Kingra ¹, Naveen Aggarwal ¹, Raahat Devender Singh ¹

Affiliations
1 UIET, Panjab University, Chandigarh - 160014, Punjab, IN

Abstract

Objectives: This paper summarizes all the proposed techniques involved in digital video inter-frame forgery detection for MPEG-1, 2, 4 and H.264/AVC encoded videos. Methods/Statistical Analysis: Double compression detection techniques are classified here on the basis of footprints analyzed during detection. The detection methods designed for videos that use fixed GOP structure for first and any of the subsequent compression are different from the videos that use different GOP structure. Video inter-frame forgery detection techniques are then analyzed on the basis of type of forgery they detect and the type of codec used for video encoding. Findings: Digital videos often provide forensic evidence in legal, medical and surveillance applications but are more prone to inter-frame forgeries, which are not only easy to perform but are equally difficult to detect as well. The analysis of the literature ascertained that majority of the proposed techniques are dependent on the number of frames tampered and video codec used to encode the videos. Among these proposed techniques, double MPEG compression was best detected using the technique which utilizes Benford’s law and was proposed by Chen and Shi on MPEG-1 and MPEG-2 encoded videos. On the other hand, Wang et al. gave sound results for all kinds of inter-frame forgeries on MPEG-2 encoded videos by utilizing the measure of optical flow consistency. Since very few authors focussed on forgery detection in MPEG-4 encoded videos and thereby such techniques have not been discussed in many survey papers. Moreover, digital cameras especially surveillance cameras which generate massive amount of videos these days have built-in MPEG-4 codec because it offers a better compression rate. Application/Improvements: Video forensics domain, therefore, is in dire need of a technique that will detect any kind of inter-frame forgery in MPEG-4 encoded videos.

Keywords

Digital Forensics, Inter-frame Forgery, Video Forgery Detection, Video Forgery Detection Survey, Video Tampering Detection.

Full Text