Handwriting Recognition: Fundamentals and Applications

By Fouad Sabry

What Is Handwriting Recognition

Handwriting recognition (HWR) is the ability of a computer to accept and interpret comprehensible handwritten input from sources such as paper documents, pictures, touch-screens, and other devices. Handwritten text recognition (HTR) is another name for handwriting recognition. Handwriting recognition (HWR) is also known as handwritten text recognition (HTR). By using optical scanning or intelligent word recognition, the image of the written text can be sensed "off line" from a piece of paper. This can be done in a number of different ways. A different option is for the movements of the pen tip to be sensed "on line," for instance by a pen-based computer screen surface. This is a task that is typically simpler because there are more hints available. Formatting, accurate character segmentation, and the identification of words that are most likely to be written are all taken care of by a handwriting recognition system.

How You Will Benefit

(I) Insights, and validations about the following topics:

Chapter 1: Handwriting Recognition

Chapter 2: Artificial Neural Network

Chapter 3: Optical Character Recognition

Chapter 4: Recurrent Neural Network

Chapter 5: Long Short-term Memory

Chapter 6: Deep Learning

Chapter 7: Signature Recognition

Chapter 8: Handwritten Biometric Recognition

Chapter 9: MNIST Database

Chapter 10: History of Artificial Neural Networks

(II) Answering the public top questions about handwriting recognition.

(III) Real world examples for the usage of handwriting recognition in many fields.

(IV) 17 appendices to explain, briefly, 266 emerging technologies in each industry to have 360-degree full understanding of handwriting recognition' technologies.

Who This Book Is For

Professionals, undergraduate and graduate students, enthusiasts, hobbyists, and those who want to go beyond basic knowledge or information for any kind of handwriting recognition.

• Language: English
• Publisher: One Billion Knowledgeable
• Release date: Jul 6, 2023

Book preview

Chapter 1: Handwriting recognition

Whether on paper, in a picture, or on a touch screen, handwriting recognition (HWR) or handwritten text recognition (HTR) refers to a computer's capacity to accept and analyze legible handwritten input. Optical scanning (optical character recognition) and intelligent word recognition allow for the off line sensing of printed text from paper. On the other hand, a pen-based computer screen surface may detect the pen's motions on line, making this a simpler job thanks to the increased availability of hints. Formatting, accurate character segmentation, and locating the most likely word choices are all taken care of by a handwriting recognition system.

Automatically converting text in a picture into letter codes useable inside computer and text-processing tools is what offline handwriting recognition is all about. The information gleaned from this form is considered a fixed illustration of handwriting. Due to the wide variety of individual writing styles, offline handwriting recognition presents a significant challenge. Moreover, as it stands, OCR engines are optimized for reading machine-printed text, whereas ICR is meant to read text that was manually printed (all capital letters).

Scanning a form or document is a common practice for offline character recognition. Therefore, it will be necessary to separate the scanned picture into its component characters. There are tools available that can carry out this procedure. However, there are a number of typical flaws at this stage. The most frequent occurrence is when two linked characters are returned as a single sub-image. This creates a significant issue during the recognizing phase. The likelihood of a chain of characters is reduced by several methods, however.

A recognition engine is then utilized to determine which digital character corresponds to each extracted character. There are now a number of various methods of recognition accessible.

To some extent, feature extraction functions like neural network recognizers. However, developers still have to decide which characteristics they consider most important by hand. This method allows the recognizer to have more say over the identifying attributes. However, due to the fact that the attributes are not automatically acquired by the system, this method needs far more time to implement than a neural network.

The newer methods of character recognition are far more concerned with identifying the whole line of text that has been segmented than the older methods were. In particular, they center on machine learning methods that can learn visual characteristics, doing away with the need for feature engineering. To generate character probabilities, state-of-the-art algorithms employ convolutional networks to extract visual information across several overlapping windows of a text line picture.

Using a particular digitizer or PDA equipped with a sensor that detects pen-tip motions and pen-up/pen-down switching, online handwriting recognition may automatically convert text as it is being typed. Digital ink refers to the information that represents handwriting in a digital format. The collected data is transformed into a format that computers and text-editing programs can understand.

Typical components of a web-based handwriting recognition interface:

anything to write with, like a pen or a stylus.

a touchscreen that is either a part of or next to the output screen.

application program that reads the writing surface and converts the stylus' actions into editable text.

Online handwriting recognition may be simplified into a few standard procedures:

preprocessing, feature extraction and

classification

The goal of preprocessing is to remove noise from the incoming data that might compromise the accuracy of the recognition. The next stage involves the collection of features. Once the preprocessing methods have sent back a vector field in two or more dimensions, the next step is to extract the higher-dimensional data. The goal of this stage is to flag key features that will be used by the recognition algorithm. Writing characteristics like as pressure, speed, and direction variations may be recorded here. Classification is the last major stage. Here, the retrieved characteristics are mapped to classes using various models, revealing the letters or phrases they represent.

In the early 1980s, commercial devices were released that used handwriting recognition as an alternative to the keyboard. The Pencept Penpad and the Inforite POS terminal are two examples of handwritten terminals. Several commercial solutions, such as those from Pencept and others, have been produced to replace a PC's keyboard and mouse with a single pointing/handwriting system since the PC consumer market exploded. In September 1989, GRiD Systems produced the GRiDPad, the first commercially available tablet-style portable computer. It ran on a modified version of Microsoft's DOS.

Tablet computers powered by GO Corp.'s PenPoint OS first appeared in the early 1990s, with releases from hardware manufacturers including NCR, IBM, and EO. PenPoint made extensive use of handwriting recognition and gestures and made those features available to other programs. The first ThinkPad tablet computer was IBM's creation and included the company's handwriting recognition technology. Later, IBM's Pen for OS/2 and Microsoft Windows also used this recognition method. There was no commercial success with any of them.

Because of technological advancements, handwriting recognition software is now able to fit into devices with a smaller form factor than tablet computers, making it a common input method for portable digital assistants. With the Apple Newton, the public was introduced to the benefits of a minimalistic user interface in a personal digital assistant. Unfortunately, the device's software's inaccuracy in attempting to learn a user's writing patterns prevented it from being commercially successful. However, by the time the Newton OS 2.0 was released, with its much enhanced