Mahesh Jangid

Gradient Local Auto-Correlation for handwritten Devanagari character recognition

This manuscript is focus on the utilization of object detection algorithm GLAC (Gradient Local Auto-Correlation) for the handwritten character recognition (HCR) problem. HOG and SIFT are already used in this (HCR) field except GLAC which produced good results than HOG and SIFT for object detection problem like human in images, pedestrian detection and image patch matching. This paper utilized GLAC algorithm to recognize the handwritten Devanagari characters. GLAC applied on two handwritten Devanagari databases, ISIDCHAR and V2DMDCHAR. The images of databases are also normalized with and without preserving aspect ratio. Using GLAC method and SVM classifier, the best results obtained on ISIDCHAR and V2DMDCHAR are 93.21%, 95.21 % respectively that justified the utilization of GLAC algorithm for character recognition problem.

Mobile Big Data: Malware and Its Analysis

The quick extension of mobile big data by telecoms vendors, has presented a flexible worldwide platform that creates a user interface for many app data bases or app stores. Big data is a tremendously well-known idea, however what are we truly talking about? From a security point of view, there are two particular issues: securing the app stores or app databases with its source of information in big data context and utilizing big data procedures to break down, and even anticipate, security flaws. The main issue arise that many hackers or attackers are targeting mobile big data in the form of signaling big data, mobile traffic big data, location-based big data, and heterogeneous data in app store. In this paper, we are taking Android-based mobile operating system for experimental setup. This paper contains an extraction technique to extract the malware in different big data context and also analysis of these malware. We have worked with many Mallarme family (as approx 40 K malware) in mobile big data and result of the whole analysis is approx 90% to identify the current malware in mobile big data.

Filter vs. Wrapper approach for optimum gene selection of high dimensional gene expression dataset: An analysis with cancer datasets

In Bioinformatics, gene dataset experiments are generating thousands of gene expression measurements, which generally used to collect information from tissue and cell samples regarding gene expression differences. Optimum gene selection from such gene expression datasets and their classification plays an important role for disease prediction & diagnosis. Further the task ahead to understand that, what is the best way of gene selection to get maximum classification accuracy from such high dimensional gene expression dataset, whether the filter is the best way to rely upon or wrapper approach can be the best suitable, beyond that which classifier works well with filter and with wrapper? To answer the question, in this paper, the performance of the filter vs. wrapper gene selection technique is being evaluated by supervised classifiers over three well known public domain datasets viz. Ovarian Cancer, Lymphomas & Leukemia. For optimal gene selection, ReliefF method is used as a filter based gene selection and Random gene subset selection algorithm is used as a wrapper based gene selection. For classification, different linear as well as an ensemble classifiers have been tested for their performances. This paper also tries to bring the fact of timing details so that through analysis, it can get derived upon that which approach is more appropriate for better time management as well as with high accuracy of the selected dataset.

Handwritten Devanagari Character Recognition Using Layer-Wise Training of Deep Convolutional Neural Networks and Adaptive Gradient Methods

Handwritten character recognition is currently getting the attention of researchers because of possible applications in assisting technology for blind and visually impaired users, human–robot interaction, automatic data entry for business documents, etc. In this work, we propose a technique to recognize handwritten Devanagari characters using deep convolutional neural networks (DCNN) which are one of the recent techniques adopted from the deep learning community. We experimented the ISIDCHAR database provided by (Information Sharing Index) ISI, Kolkata and V2DMDCHAR database with six different architectures of DCNN to evaluate the performance and also investigate the use of six recently developed adaptive gradient methods. A layer-wise technique of DCNN has been employed that helped to achieve the highest recognition accuracy and also get a faster convergence rate. The results of layer-wise-trained DCNN are favorable in comparison with those achieved by a shallow technique of handcrafted features and standard DCNN.

Deep ConvNet with Different Stochastic Optimizations for Handwritten Devanagari Character

In this paper, we present a deep learning model to recognize the handwritten Devanagari characters, which is the most popular language in India. This model aims to use the deep convolutional neural networks (DCNN) to eliminate the feature extraction process and the extraction process with the automated feature learning by the deep convolutional neural networks. It also aims to use the different optimizers with deep learning where the deep convolution neural network was trained with different optimizers to observe their role in the enhancement of recognition rate. It is discerned that the proposed model gives a 96.00% recognition accuracy with fifty epochs. The proposed model was trained on the standard handwritten Devanagari characters dataset.

Counting and Classification of Vehicle Through Virtual Region for Private Parking Solution

This paper presents a new and efficient way to track the movement of vehicles and counting them for the purpose of better and economical vehicle parking system. The numbers of the vehicles are growing very rapidly which produced the problem of the parking. There are diverse techniques that have been introduced by many researchers those are used according to the need and scale. This paper is addressed the problem of vehicle parking for small scale by using the surveillance camera. A background subtraction is used to detect moving vehicle. The moving vehicles are tracked using the Gaussian mixture model and a foreground mask is created. Dilation is applied to remove inconsistent and noise particles. The variable intensity distribution is plotted on a histogram and changes are identified to count vehicles. Each parking spot can accommodate only certain numbers of vehicle. If the count has reached the maximum limit then the display unit guides the driver to next available parking spot.

Real-Time Bottle Detection Using Histogram of Oriented Gradients

Object detection (Mohan et al. in PAMI, 2001 [1]; Lowe in IJCV 60(2):91–110, 2004 [2]) is one of the pivotal computer vision problems, which is still welcoming new and improved solutions. This area of object detection has seen many attempts made toward the detection of different objects. In this paper, we described a method of bottle detection based on histogram of oriented gradients which proves to be superior to the rest, in terms of both detection rate and error rate when used with a linear SVM classifier. This work is to serve the purpose of water bottle detection and classification in a video feed captured by a robot in the office to serve the needs of person. This will help in automating the servant work and reduce human involvement as well as dependency.

Accuracy Enhancement of Devanagari Character Recognition by Gray level Normalization

The recognition of Indian handwritten languages are attracting the attention of researchers due to wide variation and scope. This manuscript is focus on the Accuracy enhancement of handwritten Devanagari character recognition using background elimination and gray level normalization techniques. Devanagari is a famous and widely used script in India. In last few decades, the gradient based approaches have become a best choice to extract the features from handwritten characters. GLAC (Gradient Local Auto-Correlation) feature extraction technique is used for the experiment. All the experiments have done on standard handwritten Devanagari database (36172) and obtained (95.94%) higher recognition rate.

Touching character segmentation of Devanagari script

Segmentation of characters is one of the major step in OCR system. Devanagari script is a two dimensional form of symbol. It is very inconvenient to treat each form of character as a separate symbol because such combinations are very large in number. Segmentation is a process which separates words into characters. In this paper, we deal with segmentation of offline handwritten document especially touching character segmentation also we have discussed the process involved in touching character segmentation and the problem in those processes with their solution. The proposed method is tested on the database set of 250 handwritten touching characters. Experimental results indicate that the method used correctly segments the touching characters.

Similar handwritten devanagari character recognition by critical region estimation

The recognition of handwritten similar shape characters are still a challenging problem in HOCR system. Almost entire scripts are suffering with this kind of problem. This manuscript is considered the similar shape character problem in Devanagari script. An algorithm is proposed first to estimate the similar character pairs in Devanagari Script and 7 pairs are identified by investigating the confusion matrix. Similar shape characters have very minor difference in shape thats why at the time of recognition (classification) phase, classifier is being confused with another similar shape characters. This problem can be solved by estimate that minor difference called critical region in the similar shape characters and used critical region to extract the more features before classification phase. The critical region is estimated by Fisher discrimination function. A new kind of masking techniques used to extract the features from ISIDCHAR (standard Devanagari database) and 96.58 % recognition rate is obtained finally after getting the 81.94% improvement in similar character recognition.