Dr. Al-Ma'adeed is the first to research writer recognition in Arabic
Dr. Somaya Al-Ma’adeed received her BSc in computer science from Qatar University in 1994. She received her MSc in mathematics and computer science from Alexandria University, Egypt, in 1999 and went on to get her PhD in computer science from Nottingham University in 2004. Since 1994, she has worked on research based at Qatar University, specializing in character recognition, writer identification, speech recognition, tendering systems and document management. Dr. Al-Ma’adeed has published around 20 papers in these areas and is currently an Assistant Professor in the Computer Science and Engineering Department at Qatar University.
Q: Your research has been drawing attention—can you explain what you are doing?
A: My work is about writer identification. Just as you can know people from their fingerprints, irises or faces, you can know people from their handwriting. Through research and experiment, we found that people can be distinguished only by their handwriting. In fact, the recognition rate is more than 90 percent, and this is useful.
So, for example, there was a criminal in the United States who kidnapped a child and killed him. They found him from his handwriting.
Q: What kinds of techniques are used to distinguish one type of handwriting from another?
A: Mostly we rely on features. For instance, we can tell a lot from the angles of handwriting. We measure angles within a pixel of space using computer software. We can perform different exercises like finding the sum all of the angles.
Angle is just one type of feature we analyze; there are more than 90 features to distinguish the handwriting. There is writer identification involving the signature, or identification involving regular writing. We look at both the small-scale and large-scale features. We discovered that some features are more distinguishable than other features, such as the degree of the angles. These are more consistent within the way a person writes, and with software, we can recognize the pattern of writing by the features.
Q: What made you interested in this?
A: I was interested in character recognition, how to know what people write in Arabic. If someone writes something in Arabic, how do we translate it into something that the computer recognizes? When I realized that this field was [saturated], I wanted to do something new.
I started this work on handwriting recognition by running a student senior project supported by an internal university grant, and I expanded it when I saw that it was working.
I’m the first person to conduct this specific kind of research in Arabic. I’ve been working in this field of pattern and character recognition, and I’m the lead principal investigator, collaborating with Ahmed Bouridane at Northumbria University in the UK.
Q: Any milestones you’d like to share?
A: In 2011, there were two competitions, one for writer identification and one for signature identification. These involved Arabic, English (Latin-based script) and Chinese. We organized the Arabic one, so we couldn’t participate, but we won the Latin-based writer identification, for offline.
Q: What is offline?
A: The difference between offline and online is that for online subjects write in a test bed that’s connected to the computer, and the computer senses the pressure and direction of the writing. This is easier than offline because there are more features to analyze and compare. For offline recognition, it’s more difficult, and we won that for the Latin language.
Another competition involved signature identification, and, again, we won the Latin-based module. We are a team of one postdoc, one research assistant and two co-PIs, yet we are a strong collaboration, because each of us has different tasks, different specialties.
In this competition, we scored 100 percent on an English document and 90 percent on a Greek document. We submitted two features for each.
Q: How did you decide on two features?
A: Dr. Bouridane’s features [of specialization] and mine combine in different ways. Features, like the direction of letters and the angle of lines, mix with other features. We can’t know which feature is better. The choice also depends on the information in the database; in these competitions, they don’t give you the whole dataset. We don’t know which part we will analyze in advance; we work with what is provided. The features are embedded in mathematical formulas, and we have a program that combines features to find which work together better than others.
Q: Is it true that you are looking into patents around this work?
A: We are looking into patents, one covering signature identification, and one on writer identification. We haven’t published a lot because it is difficult to get a patent if your methods are widely published.
To avoid publishing our work and results, we have published how we did in competitions. This allowed us to have a chance at the patent.
Q: What do you hope your research does for society?
A: I hope that it will be useful in forensic applications and in the courts to verify signatures on important documents. I’m sure it can be used in forensics.
I’m also working with the Minister of Interior now here in Qatar. We had a workshop where we invited him and asked him to show us what he does, and he showed us how they identify writers manually and we showed him how we can do that by computer.
In Qatar it’s not legal or widely acceptable to collect online signatures; they must be in hardcopy. But in Europe, it’s becoming more widespread so we hope that a new system will be accepted in here.
In the meantime, my students are working and learning from this. It is a senior project for them, to work with the software and analyze the features.
Q: Can you say a little about the support you have received from QNRF?
A: I am thankful to QNRF. They respond so well to our requests to adjust things when we needed to. They are very responsible. I thank Sheikha Mozah for establishing the NPRP fund, because five years ago Qatar University didn’t have this kind of research activity. We were only teaching. We didn’t have the resources to do research, so I feel lucky now.
Dr. Al-Ma'adeed, we thank you very much for this interview and wish you the best of luck with your future research.
© copyright QNRF 2015