OpticalCharacterRecognition (OCR) is an interesting and challenging field of research in pattern recognition, artificial intelligence and machine vision and is used in many real life applications like postal pin code sorting, bank cheque processing, job application form processing, vehicle number plate recognition, tax forms processing, digit recognition. A lot of research work has been done in this field considering the scope of the area. In the literature, various approaches are available for implementation of pre-processing, feature extraction and classification. G. S. Lehal and Nivedan Bhatt  have proposed a contour extraction technique. Reena Bajaj  have used three different types of feature namely, density features, moment features and descriptive features for classification of DevanagariNumerals. R. J. Ramteke  has presented a method based on invariant moments and the divisions of image for the recognition of numerals. U. Bhattacharya  have used a combination of Artificial Neural Network (ANN) and Hidden Markov Model (HMM) classifier. In this paper, a method of recognizing offlinehandwrittenDevanagari numeral using Daubechies-4 wavelet filter and multilayer perceptron neural network classifier is presented. The method is capable of providing recognition accuracy of about 60%-70%.
A number of methods relating to application of wavelets in characterrecognition have been reported . In one of the works reported, images of numerals are decomposed (single level) using Bi-orthogonal wavelets to four sub-images and each sub-image is normalized to the range [0, 1]. The feature vector is formed by these sub-images. Another approach is to get a signature of the character by converting the average image obtained after a three level wavelet transformation into binary image and use it for subsequent classification. Wavelet packets are also used where feature extraction is done by taking the wavelet packet transform of the character image using the best basis algorithm, for a desired number of multiresolution levels.
proposed . A Euclidian distance criterion and KNN classifier is used to classify the handwritten Kannada vowels. A total 1625 images are considered for experimentation and overall accuracy is found to be 85.53%. Rajashekararadhya and Vanaja proposed an offlinehandwritten numeral recognition technique for four south Indian languages like Kannada, Telugu, Tamil and Malayalam. In this work, they suggested a feature extraction technique, based on zone and image centroid. They used two different classifiers nearest neighbor and back propagation neural network to achieve 99% accuracy for Kannada and Telugu, 96% for Tamil and 95% for Malayalam . The work of Pal et al with gradient features and a classifier combination of SVM and MQDF have achieved a recognition rate of 95.13% for Devanagaricharacterrecognition . Another work by Pal et al has turned up an accuracy of 95.19% with curvature features and mirror image learning . Patil et al describes a complete system for the recognition of isolatedhandwrittenDevanagaricharacterusing Fourier Descriptor and Hidden-Markov Model (HMM). Before extracting the features, the images are normalized using image isometrics such as translation, rotation and scaling. After normalization, the Fourier features are extracted using Fourier Descriptor. An automatic system trained 400 images of image database and character model form with multivariate Gaussian state conditional distribution. A separate set of 100 characters was used to test the system. The recognition accuracy for individual character varies from 90% to 100%.Siddarth et al. have worked with features like zonal density, projection histogram, distance profile and background direction distribution (BDD), and classifiers SVM, KNN and probabilistic neural networks. SVM gave the highest recognition accuracy of 95.04% with zonal density and BDD features .
Principle Component Analysis PCA is a classical feature extraction and data representation technique widely used in pattern recognition. It is one of the most successful techniques in face recognition. But it has drawback of high computational especially for big size database. This paper conducts a study to optimize the time complexity of PCA (eigenfaces) that does not affects the recognition performance. The authors minimize the participated eigenvectors which consequently decreases the computational time. A comparison is done to compare the differences between the recognition time in the original algorithm and in the enhanced algorithm. The performance of the original and the enhanced proposed algorithm is tested on face94 face database. Experimental results show that the recognition time is reduced by 35% by applying our proposed enhanced algorithm. DET Curves are used to illustrate the experimental results.
of hyper planes in a high dimensional space for classification tasks. Here, a good separation is achieved by the hyper plane which has the largest distance to the nearest training data points of any class. In general, larger the margin, lower the generalization error of the classifier . In accordance with the above mentioned merits, SVMs are implemented in this paper. The advantage of using SVM is, it globally replaces all missing values and transforms nominal attributes into binary ones and also applies normalization to all values. Since ASR needs involving multiclass SVM, one Vs one type is chosen to test against all the other classes separately. The size of the training set needed for each SVM in one2Vs2one solution leads to a smaller computational effort with comparable accuracy rates. The significant factor in using SVM is choosing a good kernel function. In this work, the polynomial kernel of first order is used. The cost parameter value is set as 1. For this work, SVM has offered considerable results like HMM and MLP. SVMs have several advantages which are listed below:
Even though a large number of wavelets exist, one needs new wavelets for their specific applications. One of the basic wavelet categories is orthogonal wavelets. But it was hard to find orthogonal and symmetric wavelets. Symmetricity is required for perfect reconstruction. Hence, a need for orthogonal and symmetric arises. The solution was in the form of biorthogonal wavelets which preserves perfect reconstruction condition. Though a number of biorthogonal wavelets are proposed in the literature, in this paper four new biorthogonal wavelets are proposed which gives better compression performance. The new wavelets are compared with traditional wavelets by using the design metrics Peak Signal to Noise Ratio (PSNR) and Compression Ratio (CR). Set Partitioning in Hierarchical Trees (SPIHT) coding algorithm was utilized to incorporate compression of images.
Abstract — This paper presents a novel approach to recognize Grantha, an ancient script in South India and converting it to Malayalam, a prevalent language in South India using online characterrecognition mechanism. The motivation behind this work owes its credit to (i) developing a mechanism to recognize Grantha script in this modern world and (ii) affirming the strong connection among Grantha and Malayalam. A framework for the recognition of Grantha script using online characterrecognition is designed and implemented. The features extracted from the Grantha script comprises mainly of time-domain features based on writing direction and curvature. The recognized characters are mapped to corresponding Malayalam characters. The framework was tested on a bed of medium length manuscripts containing 9-12 sample lines and printed pages of a book titled Soundarya Lahari writtenin Grantha by Sri Adi Shankara to recognize the words and sentences. The manuscript recognition rates with the system are for Grantha as 92.11%, Old Malayalam 90.82% and for new Malayalam script 89.56%. The recognition rates of pages of the printed book are for Grantha as 96.16%, Old Malayalam script 95.22% and new Malayalam script as 92.32% respectively. These results show the efficiency of the developed system.
Using SignWriting any LGP speaker may be able to write any word into the system which will be able to store it as the set of cluster elements needed for its eventual animation. The only major obstacle for this process is the lack of a standard usage of SignWriting, which might increase the difficulty of identifying the clusters that compose a sign. Yet, this problem can be minimized through the imposition of a grid layout that makes obligatory the visual separation between each cluster. This will create a updater similar to SignWritting Studio TM4 interface but instead of creating an CSV or XML representation it will store the gloss as a set of parameters described in the previous section.
The proposed approach accomplishes the aim of recognizing Urdu characters in real time and its conversion into human speech. This system has an advantage over other systems as it serves as a basic architecture for handwritten Urdu characters captured in real time. In Table 3, TTS systems developed for different languages are compared and their characterrecognition accuracies are highlighted. In comparison, the overall OCR accuracy for the proposed methodology is 91%. In addition, the TTS developed here uses simple algorithms, has online recognition feature and a dedicated GUI for easy implementation. Most of the literature avail able doesn’t incorporate online characterrecognition. Moreover, although a dedicated GUI for TTS is included in many studies [1, 10], the GUI proposed here is very easy to use. On the basis on these advantages, this system can be extended to use for people with limited vision or are illiterate and unable to read important text messages and instructions.
In many speech recognition systems, endpoint detection and pattern recognition are used to detect the presence of speech in a background of noise. The beginning and end of a word should be detected by the system that processes the word. The problem of detecting the endpoints would seem to be easily distinguished by human, but it has been found complicated for machine to recognize. Instead in the last three decades, a number of endpoint detection methods have been developed to improve the speed and accuracy of a speech recognition system. This study uses the Malay language, which is a branch of the Austronesian (Malayo-Polynesian) language family, spoken as a native language by more than 33,000,000 persons distributed over the Malay Peninsula, Sumatra, Borneo, and the numerous smaller islands of the area, and widely used in Malaysia and Indonesia as a second language  .
Selection of Mother Wavelets-the choice of wavelets is crucial and determines the image compression performance. The choice of wavelet functions depends on the contents and resolution image. The best way for choosing wavelet function is based on the quality of objective picture and the quality of subjective picture. In this paper, we have considered MR images only and the major mother wavelets are including most popular wavelets such as Daubechies Wavelet, Coiflets Wavelet, Biorthogonal Wavelet, Reverse Biorthogonal Wavelet , Symlets Wavelet, Morlet Wavelet, and Discrete Mayer Wavelet. The different mother wavelets are studied on different classes of images based on the performance measurements, including novel proposed image quality index which are normally used for quality of images . These performances are computed for the cases of above six mother wavelets for compressing the images of different classes. Set Partition in Hierarchical Trees (SPIHT)  is the most popular image compression method. It provides such kind of features like as the highest image quality, progressive image transmission, fully embedded coded file, simple quantization algorithm, fast coding/decoding, completely adaptive, lossless compression, and exact bit rate coding and error protection. It makes use of three lists: (i) the List of Significant Pixels (LSP), (ii) List of Insignificant Pixels (LIP) and (iii) List of Insignificant Sets (LIS). These are coefficient location lists that contain their coordinates in this algorithm. After the initialization, the algorithm takes two stages for each level of threshold – the sorting pass (in which lists are organized) and the refinement pass (which does the actual progressive coding transmission). The result is showed in the form of a bit stream. It is capable of recovering the image perfectly by coding all bits of the transform.
The model is able to capture many of the features of the observations. For instance the storm tracks in the Northern Atlantic and Pacific oceans are well reproduced, as are the decks of stratus clouds off the coast of Peru. There are some discrepancies between the model and data, for instance the model underestimates the optical depth in the intertropical convergence zone (ITCZ). It also produces too large opti- cal depths over land in the winter, notably China, the Pa- cific coast of North America and the eastern USA in DJF and Chile in JJA, which may reflect an underestimate of the radii of the cloud particles in these locations at this time. Some of the other differences can be attributed to issues with the data, with errors introduced at high solar zenith angles, con- tamination from dust, and an underestimate of the amount of optically thick cloud (Marchand et al., 2010). For instance the discrepancy between the model and observations over the Sahara is likely a result of factors such as dust in the mea- surements rather than a deficiency in clouds in the model. However some of the discrepancies can be attributed to the simplified calculation of cloud optical depth and the limited resolution of the model. From the general good agreement we
Abstract— Surveillance of public spaces is often conducted with the help of cameras placed at elevated positions. Recently, drones with high resolution cameras have made it possible to perform overhead surveillance of critical spaces. However, images obtained in these conditions may not contain enough body features to allow conventional biometric recognition. This paper introduces a novel gait recognition system which uses the shadows cast by users, when available. It includes two main contributions: (i) a method for shadow segmentation, which analyzes the orientation of the silhouette contour to identify the feet position along time, in order to separate the body and shadow silhouettes connected at such positions; (ii) a method that normalizes the segmented shadow silhouettes, by applying a transformation derived from optimizing the low rank textures of a gait texture image, to compensate for changes in view and shadow orientation. The normalized shadow silhouettes can then undergo a gait recognition algorithm, which in this paper relies on the computation of a gait energy image, combined with linear discriminant analysis for user recognition. The proposed system outperforms the available state-of-the-art, being robust to changes in acquisition viewpoints.
This paper discussed the currency recognition system with a digitalized image processing system. This system can work for assisting visually impaired people to correctly determine the denomination of the currency notes. It can help to distinguish the original note from counterfeit currency. If the image exhibit information loses such as surface damage, noise level, sharpness issues and so on, the recognition may fail and the user has to do the processing again. The system had beenprogrammed by using MATLAB and it will include a user- friendly interface.
Abstract – Sign language plays a great role as communication media for people with hearing difficulties.In developed countries, systems are made for overcoming a problem in communication with deaf people. This encouraged us to develop a system for the Bosnian sign language since there is a need for such system. The work is done with the use of digital image processing methods providing a system that teaches a multilayer neural network using a back propagation algorithm. Images are processed by feature extraction methods, and by masking method the data set has been created. Training is done using cross validation method for better performance thus; an accuracy of 84% is achieved .
Fingerprint recognition is one of most popular and accuracy Biometric technologies. Nowadays, it is used in many real applications. However, recognizing fingerprints in poor quality images is still a very complex problem. In recent years, many algorithms, models… are given to improve the accuracy of recognition system. This paper discusses on the standardized fingerprint model which is used to synthesize the template of fingerprints. In this model, after pre-processing step, we find the transformation between templates, adjust parameters, synthesize fingerprint, and reduce noises. Then, we use the final fingerprint to match with others in FVC2004 fingerprint database (DB4) to show the capability of the model.
In their paper, Jones et al  have talked about how Language Modeling can be used to solve the problem of sentence recognition. They have used probabilistic grammar along with a Hidden Markov Identifier for adequately completing this task. In , an algorithm had been proposed to describe a framework for classifier combination in grammar-guided sentence recognition. Hybrid Techniques have also been used for the aforesaid problem. In , Hidden Markov Model (HMM) and Neural Network (NN) Model have been combined for the solution. Here, Word Recognition had been using a Tree- Structured dictionary while Sentence Recognition is done using a word-predecessor conditioned beam search algorithm to segment into words and word recognition. In , sentence recognition has been achieved which uses a template based pattern recognition and represents words as a series of diphone-like segments. In , word co- occurrence probability has been used for sentence recognition. The incurred results were also compared with the method using the Context Free Grammar. Binary Neural Networks have also been successfully used in the task of pattern recognition. Binary Hamming Neural Network has been applied to recognize sentences and have been found to sufficiently successful in this regard. The system proposed also takes advantage of greater speed of the Binary Networks to provide a very efficient solution to the problem of sentence recognition . David and Rajsekaran have talked about how Hopfield classifiers can be used as a tool in Pattern Classification . A combination of Hopfield Neural Network and Back Propagation Approach has also been used to propose method of vehicle license characterrecognition supported by a study of the relations among the study rate, error precision and nodes of the hidden layer .
In the recent years reversible logic design has promising applications in low power computing, optical computing, quantum computing. VLSI design mainly concentrates on low power logic circuit design. In the present scenario researchers have made implementations of reversible logic gates in optical domain for its low energy consumption and high speed. This study is all about designing a reversible Full adder using combination of all optical Toffoli and all optical TNOR and to compare it with the Full adder designed using all optical Toffoli gate in terms of optical cost. All optical TNOR gate can work as a replacement of existing NAND based All optical Toffoli Gate (TG). The gates are designed using Mach- Zehnder Interferometer (MZI) based optical switch. The proposed system is developed with the basic of reversibility to design all optical full Adder implemented with CMOS transistors. The design is efficient in terms of both architecture and in power consumption.
analisada. Por exemplo, seja a fun¸c˜ao suave por partes, como ilustrado na Fig. 7. Como pode ser visto na re- presenta¸c˜ao no plano posi¸c˜ao × escala (x×j), a maioria dos seus coeficientes wavelet s˜ao desprez´ıveis e podem ser desconsiderados. Esse ´e o princ´ıpio b´asico de muitas aplica¸c˜oes de wavelets.