Muhammad Haroon Yousaf

Implementation of VLSB Stegnography Using Modular Distance Technique

This work proposes a spanking new technique, Modular Distance Technique, to implement Variable Least Significant Bits Stegnography, in spatial domain, providing twofold security. It is an overriding and secure data embedding technique having low data hiding capacity of with least distortion. This is much immune to Steganalysis providing a large Key Size. This technique can be implemented with Euclidean, Chess Board and City Block distances with same data hiding capacity for each and contributing significantly to the key size. The key size of modular distance technique is almost 27 times of the size of the square of cover image. Low distortion made it difficult for intruder to detected hidden information and large key size make it difficult to extract the hidden information. This technique is contributing a data hiding capacity of 12.5–56.25 % with SNR ranging from 29.7 to 8 db. The hiding capacity and SNR varies with changing reference pixel, base of Mod and type of distance.

Multi-view human action recognition using 2D motion templates based on MHIs and their HOG description

In this study, a new multi-view human action recognition approach is proposed by exploiting low-dimensional motion information of actions. Before feature extraction, pre-processing steps are performed to remove noise from silhouettes, incurred due to imperfect, but realistic segmentation. Two-dimensional motion templates based on motion history image (MHI) are computed for each view/action video. Histograms of oriented gradients (HOGs) are used as an efficient description of the MHIs which are classified using nearest neighbor (NN) classifier. As compared with existing approaches, the proposed method has three advantages: (i) does not require a fixed number of cameras setup during training and testing stages hence missing camera-views can be tolerated, (ii) requires less memory and bandwidth requirements and hence (iii) is computationally efficient which makes it suitable for real-time action recognition. As far as the authors know, this is the first report of results on the MuHAVi-uncut dataset having a large number of action categories and a large set of camera-views with noisy silhouettes which can be used by future workers as a baseline to improve on. Experimentation results on multi-view with this dataset gives a high-accuracy rate of 95.4% using leave-one-sequence-out cross-validation technique and compares well to similar state-of-the-art approaches.

Optimized Audio Classification and Segmentation Algorithm by Using Ensemble Methods

Audio segmentation is a basis for multimedia content analysis which is the most important and widely used application nowadays. An optimized audio classification and segmentation algorithm is presented in this paper that segments a superimposed audio stream on the basis of its content into four main audio types: pure-speech, music, environment sound, and silence. An algorithm is proposed that preserves important audio content and reduces the misclassification rate without using large amount of training data, which handles noise and is suitable for use for real-time applications. Noise in an audio stream is segmented out as environment sound. A hybrid classification approach is used, bagged support vector machines (SVMs) with artificial neural networks (ANNs). Audio stream is classified, firstly, into speech and nonspeech segment by using bagged support vector machines; nonspeech segment is further classified into music and environment sound by using artificial neural networks and lastly, speech segment is classified into silence and pure-speech segments on the basis of rule-based classifier. Minimum data is used for training classifier; ensemble methods are used for minimizing misclassification rate and approximately 98% accurate segments are obtained. A fast and efficient algorithm is designed that can be used with real-time multimedia applications.

Real-time feet movement detection and tracking for controlling a Toy car

This paper proposes a new approach of controlling vehicles by detection of visual foot movement. It discusses the constraints imposed by the use of moving vehicle during real-time foot movement tracking. The visual feet tracking attributes ensure the fact that, unlike its counterpart in physical world, our driving does not involve effort of steering. The proposed technique comprises of two main parts; Software part and Hardware part. The Software part consists of the steps, namely, Initialization phase for initializing the software according to environmental conditions, Feature extraction phase to extract features, parameters calculation phase to calculate parameters in order to determine direction of motion. The hardware part consists of Toy cars RC alteration and Communication from personal computer using serial or parallel port. It works in real-time optimizing the problems caused by shadows. The output parameters are used for detection of any movement made by foot. So the input to the software part is the stream of images of foot and the output produced by the software part are the parameters specifying direction of motion. These parameters are input to the hardware part, which in this case is a toy car, and the hardware performs desired action.

A generic passive image forgery detection scheme using local binary pattern with rich models

Abstract Image forgery detection is one of the prominent areas from research and development perspective. This research work aims to propose a scheme for the detection of multiple types of image forgeries. In this paper, a generic passive image forgery scheme is proposed using spatial rich model (SRM) in combination with textural feature i.e. local binary pattern (LBP). Moreover, different sub-model selection strategies are implemented and analyzed to investigate the performance-to-model dimensionality trade-off. Ensemble multi-class classifier is used for classifying the features into different forgery classes. The proposed scheme is evaluated on the dataset generated from IEEE IFS-TC image forensics challenge containing 10 different kinds of forgeries. The results reveal that computing LBP on noise residuals in conjunction with co-occurrence matrices using BEST-q-CLASS feature selection strategy produces a model which performs efficiently for almost any set of modifications with accuracy of 98.4%.

Multi-view Human Action Recognition Using Histograms of Oriented Gradients (HOG) Description of Motion History Images (MHIs)

In this paper, a silhouette-based view-independent human action recognition scheme is proposed for multi-camera dataset. To overcome the high-dimensionality issue, incurred due to multi-camera data, the low-dimensional representation based on Motion History Image (MHI) was extracted. A single MHI is computed for each view/action video. For efficient description of MHIs Histograms of Oriented Gradients (HOG) are employed. Finally the classification of HOG based description of MHIs is based on Nearest Neighbor (NN) classifier. The proposed method does not employ feature fusion for multi-view data and therefore this method does not require a fixed number of cameras setup during training and testing stages. The proposed method is suitable for multi-view as well as single view dataset as no feature fusion is used. Experimentation results on multi-view MuHAVi-14 and MuHAVi-8 datasets give high accuracy rates of 92.65% and 99.26% respectively using Leave-One-Sequence-Out (LOSO) cross validation technique as compared to similar state-of-the-art approaches. The proposed method is computationally efficient and hence suitable for real-time action recognition systems.

A new algorithmic approach for fingers detection and identification

Gesture recognition is concerned with the goal of interpreting human gestures through mathematical algorithms. Gestures can originate from any bodily motion or state but commonly originate from the face or hand. Hand gesture detection in a real time environment, where the time and memory are important issues, is a critical operation. Hand gesture recognition largely depends on the accurate detection of the fingers. This paper presents a new algorithmic approach to detect and identify fingers of human hand. The proposed algorithm does not depend upon the prior knowledge of the scene. It detects the active fingers and Metacarpophalangeal (MCP) of the inactive fingers from an already detected hand. Dynamic thresholding technique and connected component labeling scheme are employed for background elimination and hand detection respectively. Algorithm proposed a new approach for finger identification in real time environment keeping the memory and time constraint as low as possible.

An Optimized and Fast Scheme for Real-Time Human Detection Using Raspberry Pi

Real-time human detection is a challenging task due to appearance variance, occlusion and rapidly changing content; therefore it requires efficient hardware and optimized software. This paper presents a real-time human detection scheme on a Raspberry Pi. An efficient algorithm for human detection is proposed by processing regions of interest (ROI) based upon foreground estimation. Different number of scales have been considered for computing Histogram of Oriented Gradients (HOG) features for the selected ROI. Support vector machine (SVM) is employed for classification of HOG feature vectors into detected and non-detected human regions. Detected human regions are further filtered by analyzing the area of overlapping regions. Considering the limited capabilities of Raspberry Pi, the proposed scheme is evaluated using six different testing schemes on Town Centre and CAVIAR datasets. Out of these six testing schemes, Single Window with two Scales (SW2S) processes 3 frames per second with acceptable less accuracy than the original HOG. The proposed algorithm is about 8 times faster than the original multi-scale HOG and recommended to be used for real-time human detection on a Raspberry Pi.

Computer vision based detection and localization of potholes in asphalt pavement images

Asphalt pavement distresses have significant importance in roads and highways. This paper addresses the detection and localization of one of the key pavement distresses, the potholes using computer vision. Different kinds of pothole and non-pothole images from asphalt pavement are considered for experimentation. Considering the appearance-shape based nature of the potholes, Histograms of oriented gradients (HOG) features are computed for the input images. Features are trained and classified using Naïve Bayes classifier resulting in labeling of the input as pothole or non-pothole image. To locate the pothole in the detected pothole images, normalized graph cut segmentation scheme is employed. Proposed scheme is tested on a dataset having broad range of pavement images. Experimentation results showed 90 % accuracy for the detection of pothole images and high recall for the localization of pothole in the detected images.

A novel vision based Approach for instructor's performance and behavior analysis

Performance analysis of instructors in the lecture room plays a significant role in maintaining the higher education quality and standards. This paper presents a novel approach for the evaluation of instructors performance and behavior in the lecture room. Proposed approach employs the lecture video using face recognition and pose estimation of instructor. Instructor time-in and time-out monitoring; walking, pointing, writing and addressing postures are focused for the analysis. Edge detection and texture based descriptor are suggested for the segmentation of lecture room scene; resulting in localization of white board and the presentation area. Gaussian mixture model and morphological operations are used for instructor detection during lecture. For time-in and time-out monitoring of instructor, face recognition is employed. For instructor pose estimation, morphological features of instructors upper limb are extracted in the space-time. Space-time features of instructors upper limb are classified into respective pose using Bayesian classification. Experimentation results shows 96% recognition rate for instructor selected postures. Proposed research work recommends an innovative enhancement to instructor performance analysis in the lecture room by generating a comprehensive activity analysis.