“The whole is greater than the sum of its parts,” said Aristotle. Understanding a phenomenon requires multiple perspectives. The more sources of information we consider, the more precise our interpretation will be and the greater our ability to detect hidden patterns and relationships. This is precisely what happens in electronic handwriting analysis applied to fields such as education, healthcare, and the detection of forged signatures, among others.
Currently, we write with a pen and paper or on a digital tablet, for example. However, writing isn’t just the trace left by the pen; it’s a process that involves the movement of the arm, forearm, and wrist. Tablets can’t capture this information, and even if motion-capture sensors on the arm can, they are invasive and disrupt the natural feel of the process.
Now, is there a way to estimate these movements and better understand the writing process without interfering with the writer?
The key is in the anthropomorphic form of the robots
A robotic arm can reproduce handwriting with great precision. Doing so makes it easy to record kinematics (velocities and accelerations) and dynamics (forces involved in joint movement). By assimilating these robotic movements to those of a human arm, a valuable source of information is obtained.
This is because robotic arms, with their degrees of freedom and range of movements, resemble the structure and function of the human arm. In some models, it is even possible to identify equivalences with specific body parts, such as the trunk, forearm, upper arm, elbow, wrist, and even the fingers in those that incorporate grippers.
Therefore, when a robot imitates the writing process, all its parts move in a coordinated manner. These movements allow the robot to estimate how a human arm would move when making the same stroke, and then incorporate this source of information into the data provided by the digital tablet.
Exploiting robotics in the study of writing
The most interesting aspect of these new robotic features in the study of writing is their added value in the application of technological systems in fields such as security, health, and education.
For example, it’s possible to improve fraud detection in handwritten signatures. Due to threats such as professional forgers or mass bots , automatic signature verification systems have become more vulnerable. Robotic stroke analysis has already proven effective in detecting impostors .
Furthermore, this type of analysis has proven useful in improving the classification of drawings or writings made by patients with neurodegenerative problems.
It has also been used in the early detection of writing disorders such as dysgraphia . Its promising results are currently under review for publication in a scientific journal.
In the field of education, ensuring that a child’s handwriting—especially in the early stages of early childhood—is appropriate for their age is crucial for anticipating potential problems in the development of fine motor skills and preventing future dysfunctions. In this process, robotic features can make a difference.
What if I don’t have a robot?
Isaac Newton and advanced trigonometric relationships offer an alternative solution. Based on a given robotic model, the Lagrangian formulation (a formula for calculating an object’s trajectory, taking into account kinetic energy and inverted potential energy ) allows these robotic characteristics to be calculated from a theoretical perspective, reliably estimating, with sufficient precision, the values of the kinematics and dynamics of the robotic arm for handwriting analysis.
However, a robotic model faithful to the human anatomy of an arm requires a sufficiently complex physical-mathematical model, whose real-time calculation represents a challenge due to its high consumption of computational resources.
Fortunately, recent studies have developed artificial intelligence (AI) capable of estimating the kinematics and dynamics of a robotic arm’s movement with six degrees of freedom (i.e., the robotic arm can move independently at six different points within its structure). This AI only requires data collected by a tablet to function. After processing it by a trained model, it is possible to estimate the robotic characteristics associated with writing in near real time.
Science is advancing, and finding new ways to analyze a phenomenon is essential to better understanding it and making more informed decisions.
Author Bios: Moisés Díaz Cabrera is Professor of Applied Physics, Cristian Rodríguez Rodríguez is a PhD student, José Juan Quintana Hernández is Associate Professor of Robotics and Automation and Miguel Ángel Ferrer Ballester is Professor of Signal Theory and Communications all at the University of Las Palmas de Gran Canaria
Tags: robotics
