Mihir Narayan Mohanty

A non-rigid motion estimation algorithm for yawn detection in human drivers

This work focuses on the estimation of possible fatigue or drowsiness by detecting the occurrence of yawns with human drivers. An image processing technique has been proposed to analyse the deformation occurring on drivers face and accurately identify the yawn from other types of mouth opening such as talking and singing. The algorithm quantifies the degree of deformation on lips when a driver yawns. The image processing methodology is based on study of non-rigid motion patterns on 2D images. The analysis is done on a temporal sequence of images acquired by a camera. A shape-based correspondence of templates on contours of a particular region is established on the basis of curvature information. The shape similarity between the contours is analysed, after decomposing with wavelets at different levels. Finally, the yawn is correlated with fatigue-induced behaviour of drivers on a simulator.

Development of a New Algorithm Based on SVD for Image Watermarking

The research on watermarking has been increasing day-by-day since past decade. It has been largely driven by its important applications in digital copyrights management and protection. To provide more watermarks and to minimize the distortion of the watermarked image, a novel technique is presented in this paper. In this paper, the singular value decomposition (SVD)-based image watermarking scheme is proposed. The output result of SVD is more secure and robust. SVD is often used to develop robust watermarking algorithms. However, the SVD-based algorithms exhibit false-positive problem and pose security concern. In this work, we try to overcome this problem. In the proposed schemes, the host image is first decomposed into sub-bands by applying discrete wavelet transform (DWT). The watermark image is embedded in all the sub-bands by modifying the singular values of each sub-band. Next to it, we propose to encrypt and embed the singular values of the watermark image instead of original singular values. RSA algorithm has been used for the encryption process. Peak signal-to-noise ratio (PSNR) is used to measure the imperceptibility of the proposed schemes. The simulation result shows its efficacy.

Optimized Clustering Method for CT Brain Image Segmentation

Though image segmentation is a fundamental task in image analysis; it plays a vital role in the area of image processing. Its value increases in case of medical diagnostics through medical images like X-ray, PET, CT and MRI. In this paper, an attempt is taken to analyse a CT brain image. It has been segmented for a particular patch in the brain CT image that may be one of the tumours in the brain. The purpose of segmentation is to partition an image into meaningful regions with respect to a particular application. Image segmentation is a method of separating the image from the background, read the contents and isolating it. In this paper both the concept of clustering and thresholding technique with edge based segmentation methods like sobel, prewitt edge detectors is applied. Then the result is optimized using GA for efficient minimization of the objective function and for improved classification of clusters. Further the segmented result is passed through a Gaussian filter to obtain a smoothed image.

Analysis of outliers in system identification using WLMS algorithm

Outliers play an important role in adaptive systems. The rank-based Wilcoxon approach to linear regression problems in statistics are usually insensitive to outliers. This paper aim towards the Wilcoxon approach in Least Mean Square Algorithm. Also it has been applied for System Identification problem with Gaussian noise. The traditional LMS algorithm is generally well suited for identification of linear static systems where the probability of addition of outliers to data input is minimal. The investigation regarding the performance the performance analysis, error curve and deviation in presence of outliers are presented. Simulation results show that the Wilcoxon norm based LMS have better robustness against outliers.

Analysis of stressed human speech

Stress is a condition in which the normal speech of a person is changed. It has been observed that stressed syllables are longer than unstressed ones. The investigation of the speakers stress is based on specific changes in short-time spectrum of phonemes. Comparative results between the normal speech and stressful speech are achieved. In case of speech under stress, the obtained spectra differ towards the higher frequency due to enhanced pitch modulation observed in the envelope of the normal speech spectrum. In this work, a new database of speech under stress and normal condition consisting of data collected during the examination of final year students. Frequency domain analysis is most useful for speech processing and is done in this paper. It is observed that stress primarily affects vowel duration, whereas syllable final consonants have little stress variation and the durations typically differ 20%-40% between stressed and unstressed syllables.

An Optimized Cluster Based Routing Technique in VANET for Next Generation Network

Since last few years, research in the field of vehicular networking has gained much attention and popularity among the industries and academia. Intelligent approach for such technology is the challenge. In this paper, we have taken an attempt to optimize the routing algorithm for vehicular adhoc networking (VANET). Ant Colony Optimization (ACO) is an optimization technique and is applied based on clustering technique. To improve the safety factor and efficiency and to develop an intelligent transport system, it is highly conceptual with the wireless technology. It is a special type of MANET, because of the variation of routing protocols. Even if the protocols of MANET are feasible, they are not able to provide the optimum throughput required for a fast changing vehicular ad hoc network. Positions of the vehicles create the zone and the optimization is zone based. Ant Colony algorithm is combined with zone based clustering algorithm to improve the result. This approach combines the advantages of both the techniques, the ant colony algorithm as well as the zone based routing algorithm. Routing overhead has been compared between AODV, MARDYMO and TACR protocols and depicted in the graphical plots.

On the Use of MFCC Feature Vector Clustering for Efficient Text Dependent Speaker Recognition

The paper describes an experimental study and the development of a computer agent for Speaker recognition. It presents an efficient method to verify authorised speakers and identify them using MFCC Feature vector clustering. For clustering of the MFCC features, Vector Quantisation using Linde-Buzo-Gray (LBG) algorithm has been presented. This approach proves to be an efficient ASR technique.

Morphological based segmentation of brain image for tumor detection

Interpretation of bio-medical image contents is one of the most challenging field in computer vision for medical diagnosis. In context to that it has received much awareness of researchers to meet the challenges. The purpose of Image segmentation is to partition an image into meaningful regions with respect to a particular application. Edge is a basic as well as an important feature of an image. For further processing, detecting edges is one of the most important aspects in image segmentation. It is a process of identifying and locating sharp discontinuities in an image. In this paper, the brain image is considered for analysis and detection. Initially the region of interest is found, that helps to detect the particular content of the image and set the boundary of it. Basic morphological operations is used for edge detection. For this purpose the thresholding using histogram is done. The result obtained using Gaussian filter shows better performance than other methods. Comparison measure shows for MSE, PSNR & SSIM.

Emotion recognition using MLP and GMM for Oriya language

Emotion recognition of human beings is one of the major challenges in the modern complicated world of political and criminal scenario. In this paper an attempt is taken to recognise two classes of speech emotions as high arousal like angry, surprise and low arousal like sad and bore. Linear prediction coefficients (LPC), Mel-frequency cepstral coefficient (MFCC) and perceptual linear prediction (PLP) features are used for emotions recognition using multilayer perceptron (MLP) and Gaussian mixture model (GMM) classifier. Two different databases of four emotions, one of five children and other one of a professional actor has been used in this work. Emotion recognition performance of LPC, PLP and MFCC features has been compared with two classifiers, MLP and GMM. MFCC features with MLP classifier and PLP features with GMM classifier has performed best in their respective categories.

Design of FIS-Based Model for Emotional Speech Recognition

Human beings have emotions associated with their acts and speeches. The emotional expressions vary with moods and situations. Speech is an important medium through which people express their feelings. Prosodic, spectral, and other parameters of speech vary with the emotions. The ability to represent the emotional speech varies with the type of features chosen. In an attempt to recognize such an emotional content of speech, one of the spectral features (linear prediction coefficients), have been first tested by the fuzzy interference system. Next to it hybridization of LPC features with different prosodic features were compared with LPC features for recognition accuracy. Results show that the hybridization of features can classify emotions better with the FIS system.