Assignment Sample on Development of a Neural Network
1.0 Introduction
Identification is a subject including several domains, including facial recognition, thumb pad recognition, picture recognition, identification of character, acknowledgement of numbers, etc. The “HDR system” is an intelligent system capable of classifying hand-held characters as human. Segmentation of character is an essential aspect of many computer views and pictures with difficulties such as image processing and computer vision, platform recognition, etc. Due to varied handwriting styles, the categorization of handwritten characters is a harder one. Handwritten recognition of the “Arabic character (HACR)” in the recent decade has received great attention. With fast development of deep learning algorithms, scientists have achieved great progress in this sector. Language is a sort of Semitic speech spoken by millions around the world in nations of the Middle Eastern as the mother tongue. In general, the letters of the “Arabic alphabet” consist of 28 characters which are displayed. Deep Learning models consist of input and numerous nonlinear concealed lays and output neurons, which means that there are very huge numbers of interconnections and characteristics. To prevent overlap, the convolutional neural network needs a very broad range of instances. The “Convolutional Neural Network” is one class type with substantially smaller and easier to train variables “(CNN)”. CNN is an advanced, multi-layer genetic algorithm that extracts given input characteristics and qualities. “Convolution Neural Network” has a rear method training for neural net
2.0 Background
A mix of co-operative hybrid classifiers combine to improve the overall system performance in the field of automated face detection. The combination approaches include two basic categories: information fusing and choice melting. The fusion of the feature is derived from a combination of several characteristics into a vector followed by a decoder. Some methods stressed the usefulness of this field of basic assimilation. Ensemble learning procedures include the decisions of various classifiers; every system is built in various characteristics. More recent approaches have shown that the combined high ranking strategy enhances performance Feature location is a key idea for designing robust recognition algorithms. Applied sliding window based extraction strategies for localizing different automatic Arabian recognition software, references. The application of several sliding windows was also shown to yield higher classification accuracy. To accomplish location on regular “HoG” features, a unique type of localizations was employed, dubbed a pyramidal distribution of texture classification. Some techniques have calculated a mixture of statistics and physical properties, while others have calculated skeletal traits. The dispersion of curvature characteristics in various ways was achieved and shown to represent Arabic handwriting efficiently pyramidal characteristics were also retrieved and great performance in handwriting letters was attained. Input strings have to be encoded at various levels with the “pyramid graphs of characters (PHOC)”. Also the orientated “gradient pyramid histogram (PHoG)” exceeded normal “gradient histogram (HoG)” The newer systems were also calculated on the basis of bibliographical adaptations. For Hebrew handwritten text, many sorts of characteristics were collected. Certain techniques calculated a mixture of quantitative and physical properties, large point sources.
3.0 Justification of the Research
The sector is now witnessing a sharp growth in the quantity of data to be evaluated. It therefore becomes highly important to design data classification techniques that are scalable and can be readily deployed both inside the “neural network system” and different “cloud services” for comprehensive data analyses (Carbune et al 2020). In order to achieve this necessary functionality on the platform and to provide user experience as smoothly as possible, an optimized approach is therefore highly required in computer science technology for differentiating the large clutter datasets that can be implemented in the aforementioned sections. The identification of the handwriting is an important approach. So by the development of the neural network handwriting recognition will be done, which will be a major breakthrough in the research field.
3.1 Research aim
The aim of the research is to identify the various methods of development of a neural network to identify and recognize the handwriting with the help of machine learning. The research will also aim to identify the advantages of “Convolution Neural Network” in recognition of handwriting. The research will also aim to understand the scope of application of the “neural network system” in machine learning. Thus its application in the recognition of the handwriting will also be done. The concept regarding the implementation of “Deep Learning” will also be analyzed.
3.2 Research Objectives
There are some research objectives which are related to this research. The primary objectives are
- Understand the process of neural network using machine learning
- A research study that represents the effort in the process of neural network systems.
- Define the scope of application of the neural network system in machine learning.
- To illustrate the literature review to profit from a detailed understandable topic.
- To gain a comparative analysis of the analysis of the suggested neural network system.
- To develop a plan for the project and attach a Gantt chart in the time horizon
- To understand the various networks for the recognition of handwriting
3.3 Research Questions
- How does a neural network system work?
- How is machine learning connected to a neural network?
- What is the scope of the study?
- What is the future scope of the study?
- How can handwriting analysis be done by a neural network?
4.0 Literature Review
According to Baldominos 2017, “Convolutional neural networks (CNNs)” have received much attention in recent decades and have had a strong influence, partly because of their remarkable conduct in especially challenging supervised classification tasks. These systems have expected to be remarkably strong in the handling of information and time series, pictures, audio and video analyses of a unique kind. The development of effective signal preparing systems and functions technique extracts to provide suitable stored procedures suited to the classification job is an important component of the solutions in these difficulties. The impact of this neural network in the identification and the reading of handwriting is huge. So they are adopted in various firms to achieve success efficiently. A “Convolution Neural Network” can be defined as a deep-learning system capable of taking an input picture, attaching significance to distinct attributes in the picture and distinguishing between them. In this research by the author the main aim was to read and identify “Arabic handwritings” using the network. The author observes that the reading of the arabic characters is a challenging task, but with the use of the Convolution Network which uses an algorithm, the task will be much easier and efficient.
According to Loey 2017, the “Handwritten character Recognition System” is an intelligent system that is able to detect handwritten characters like a human sight. Character classification plays an essential role in many images with challenges along with the classification of “optical character”, “plate identification”, etc. Due to the many forms of individuals’ handwriting, categorization of hand written languages is a difficult endeavor (Toledo et al 2017). So the necessity of the implementation of the “Deep neural networks” becomes necessary. The author observes that the use of this algorithm helps to identify and recognize various “machine learning procedures” and functionality. The detailed study was done on the research and the techniques of identifying the handwriting was identified. A high number of instances are necessary to prevent “over fitting” in the “deep neural network”. The “Convolution Neural Network” is one major kind of deep neural network with lower features and quicker to develop.
The above figure shows the proposed structure for the identification of handwriting and according to the author the method is very useful and productive. The application of machine learning is done and thus the suitable model is produced. So the proposed method by the author can be identified as one of the most efficient models for the recognition of handwriting.
According to Khaleel 2020, the implementation of “Deep learning” has made another major contribution in the field of recognition of handwriting. The offline mode of handwriting recognition is a major breakthrough. The author observes that, despite all the advancement in technology, one can never reject that ancient civilizations contributed much and offered precious, digitally preserved handwritten texts. “Automatic manuscript recognition” focuses on the reservation of historic and modern manuscript collections. “Forgery detection” and “signature analysis” are important in many other sectors. Many energy and resources are spent on the automated recognition and digitization of a textual content. The online and offline modes of recognition are the two major methods that are classified by the experts (Baldominos et al 2018). “Automatic handwriting identification” needs the construction of a composite recognition system, the measurement of varying impacts and the joint use from several classifiers for improving the performance. There are mainly two types of approaches in the offline recognition of handwriting. The traditional approach is followed in majority. “Deep learning algorithms” provide greater identification performance in image recognition and constructing surveillance systems than other commercial feature sensors. “Deep neural networks” is one of the most effective steps of information accumulation.
The above figures show the implementation of the “deep learning” in the recognition word and handwriting. The training model of the database is constructed and after that the matching is done with one of the already existing “clustered layers”. The algorithm will be used to implement the process of matching. The algorithm will help to perform the whole process efficiently. The figure below will show the algorithm.
Each “matching cluster” comprises classes with much the same range of geographical and mathematical characteristics specified. Common categories may be a mix of multiple cluster nodes. Some shorter sets can also become a component of other combo packs which can construct the correspondent as a “binary search tree”, thereby lowering the amount of clusters in the database to “O (log (v)”, where v is the number of clusters in the database.
4.1 Literature gap
Through the literature review various journals by different authors are studied and analyzed. The analysis has helped to understand the application of “Convolution Neural Network” in the recognition of handwriting. The algorithms that can be used in this process are also studied. The literature review has also helped to identify the method of “offline recognition” which is considered to be one of the major breakthroughs in this field. Various methodologies were studied and a strong concept about the various methods of identification handwriting was developed. But there are some areas that need to be covered. The literature review did not cover many parameters like other neural network development techniques. The review was concentrated only on one method and thus the comparison was not done. In future this field must be kept in mind while performing the literature review.
4.2 Future Scope
From the literature review and the gap that is described above the future scope can be easily identified. The comparison with various other techniques is an option. The development of the algorithm with more data can also improve the efficiency of the research. The applications of the neural network in the recognition of handwriting should be discussed in more detail including the detail of clustering. For this objective, the algorithms must be examined in a wide variety of data, such as the unorganized sector or movements of the search query, that assist us to comprehend how these strategies are implemented in real time. The investigation is also important to give a comprehensive study of the various approaches to determine the feasibility and usefulness of the neural networks. All these should be planned for future research. Thus the research will be more efficient and productive. The fields will be studied in more detail and much more useful information can be gathered.
5.0 Methodology
5.1 Research Approach
The approach of the research comprises all the key assumptions, specialized data collection, treatment and dissemination processes. Academic research methodology. Therefore, the structure of the research subject is very necessary to know. This particular research follows the “Qualitative research approach”. The data collection will be from the secondary sources. The method of data collection will be discussed in the next action. Another technique that aims to offer software with valuable knowledge is the ‘Deductive Research method.’ The deductive technique to acquire appropriate information through the “secondary data collecting method” will be a very important approach for this work. The approach to analyzing the relation of most investigators with research activities is deductible. The scientist investigates what everyone else has done, analyzes current theories on any phenomenon that he or she is examining and then tests these theories. This technique also helps to fully recognize the differences in studies conducted while identifying gaps. This will also allow researchers to work successfully when they perceive these loopholes or lack of information. So this is the main research approach that will be followed for the research that will be conducted. It is inferred that the approach has many advantages that will help to complete the research in an efficient manner.
5.2 Data Collection Method
Data collecting techniques are important because the perspective and the technique of the researcher define the usage and implications of the results. In this targeted research, “secondary data collection” is used to gather various relevant information. The use of “qualitative” data analysis is done in the research. So it can be inferred that the data collection approach that is used in the research is very effective and efficient. The data collection method that is selected has a huge impact on the success of the research. Information was gathered from all available research articles, periodicals, internet outlets and from other current sources in “secondary data gathering” and investigators. Secondary data collection is the strategy by methods of which assets may obtain data on the proposed study fast. Data from several sources are also acquired to help management skills and save rates for personal study work. As science advances, the amount of social media channels and Web platforms led to large numbers of “secondary data” with a wide range of options. The “secondary strategy of data collection” will also provide researchers with extra information on contemporary class strategies which are necessary to determine different criteria and provide research study with present financial and wisdom.
5.3 Ethical Consideration
Among the most significant components of studying are moral considerations. Research participants must answer this here in this paragraph without prejudice. The security of the research respondents should be a huge problem. The entire approval of the patients really should be sought before the experiment. All knowledge will be collected using a “secondary research approach” for that focused study article. This protects or eliminates any existing source of data duplication in all sorts of ethical restrictions. For the sole purpose of collecting information, research will be conducted. The day that will be collected will be used only for the purpose of research and no misconduct will be performed.
6.0 Limitations of the Research
The objective of the research is dedicated to developing the systematic review within the context of the recent program, as a component of the “secondary research method” given in the current project. Consequently, the fundamental limits of the explanation of alternative methods are dependence on the literature (Wigington et al 2018). This will be one of the main limitations of the research. The dependency on the available resources and not using anything will be a major concern and in the suture research this should be removed. Another limitation is that the discussion is based on only one model and not on various models. Thus no comparison was performed. This is another sector that should be given concentration in future. Except these there are no other limitations and the research will be performed in an efficient manner and results will be obtain
8.0 Conclusion
The proposal has helped to understand the various techniques for the recognition of handwriting. The literature review section has helped to understand the views of various authors on “Convolution Neural Network” and its application in handwriting recognition. The proposal also discusses the approach of the research and the data collection method. All the methods are chosen after long discussion and it is inferred that the selected methods will be best suited for the research. The research will be performed by keeping in mind all the ethical considerations. The various limitations that are discussed in the proposal will be overcome in the future to make the research more productive. Thus all the aims and objectives of the research are met and the success of the research is inevitable.
Reference List
Journals
Bhunia, A.K., Das, A., Bhunia, A.K., Kishore, P.S.R. and Roy, P.P., 2019. Handwriting recognition in low-resource scripts using adversarial learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 4767-4776).
Bluche, T. and Messina, R., 2017, November. Gated convolutional recurrent neural networks for multilingual handwriting recognition. In 2017 14th IAPR international conference on document analysis and recognition (ICDAR) (Vol. 1, pp. 646-651). IEEE.
Carbune, V., Gonnet, P., Deselaers, T., Rowley, H.A., Daryin, A., Calvo, M., Wang, L.L., Keysers, D., Feuz, S. and Gervais, P., 2020. Fast multi-language LSTM-based online handwriting recognition. International Journal on Document Analysis and Recognition (IJDAR), pp.1-14.
Chammas, E., Mokbel, C. and Likforman-Sulem, L., 2018, April. Handwriting recognition of historical documents with few labeled data. In 2018 13th IAPR International Workshop on Document Analysis Systems (DAS) (pp. 43-48). IEEE.
Cilia, N.D., De Stefano, C., Fontanella, F. and di Freca, A.S., 2019. A ranking-based feature selection approach for handwritten character recognition. Pattern Recognition Letters, 121, pp.77-86.
Dargan, S., Kumar, M., Ayyagari, M.R. and Kumar, G., 2019. A survey of deep learning and its applications: A new paradigm to machine learning. Archives of Computational Methods in Engineering, pp.1-22.
Dutta, K., Krishnan, P., Mathew, M. and Jawahar, C.V., 2018, August. Improving cnn-rnn hybrid networks for handwriting recognition. In 2018 16th international conference on frontiers in handwriting recognition (ICFHR) (pp. 80-85). IEEE.
El-Sawy, A., Loey, M. and El-Bakry, H., 2017. Arabic handwritten characters recognition using convolutional neural network. WSEAS Transactions on Computer Research, 5, pp.11-19.
Fanany, M.I., 2017, May. Handwriting recognition on form document using convolutional neural network and support vector machines (CNN-SVM). In 2017 5th international conference on information and communication technology (ICoIC7) (pp. 1-6). IEEE.
Ghanim, T.M., Khalil, M.I. and Abbas, H.M., 2020. Comparative study on deep convolution neural networks DCNN-based offline Arabic handwriting recognition. IEEE Access, 8, pp.95465-95482.
Kusetogullari, H., Yavariabdi, A., Cheddad, A., Grahn, H. and Hall, J., 2019. Ardis: a swedish historical handwritten digit dataset. Neural Computing and Applications, pp.1-14.
Rabhi, B., Elbaati, A., Hamdi, Y. and Alimi, A.M., 2019, September. Handwriting recognition based on temporal order restored by the end-to-end system. In 2019 International Conference on Document Analysis and Recognition (ICDAR) (pp. 1231-1236). IEEE.
Toledo, J.I., Dey, S., Fornés, A. and Lladós, J., 2017, November. Handwriting recognition by attribute embedding and recurrent neural networks. In 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR) (Vol. 1, pp. 1038-1043). IEEE.
Wigington, C., Tensmeyer, C., Davis, B., Barrett, W., Price, B. and Cohen, S., 2018. Start, follow, read: End-to-end full-page handwriting recognition. In Proceedings of the European Conference on Computer Vision (ECCV) (pp. 367-383).
Know more about UniqueSubmission’s other writing services:
