Fig 5 - uploaded by Manju Jose
Content may be subject to copyright.
Source publication
Handwriting is one of the mostly accepted and natural means of communication among the people across the globe. Handwriting recognition has always been a challenge due to the fact that the same writer can write the same character in diverse ways. The advancement and technical merits of deep convolutional neural networks has made it very popular for...
Context in source publication
Context 1
... of finding individual learning rates of all the parameters. The model accuracy was 84% when 2 convolutional layers were deployed along with 6 epochs. So the model was enhanced by trying various combinations of convolutional layers, fully connected layers and epochs until the desired accuracy was achieved. The accuracy of the model is given in Fig. 5. The training accuracy was 84.15% during the first epoch and further increased to above 99% for the remaining all epochs. The highest validation accuracy of 96.48% was obtained for epoch 4 and then the accuracy decreased further. The lowest validation accuracy obtained was 95.45% which was attained for epoch 2. The test accuracy of the ...
Similar publications
In the process of communication, modulation signal recognition and classification are an important part of non-cooperative communication. Automatic modulation recognition technology of communication signals based on feature extraction and pattern recognition is a key research object in the radio field. The use of neural network can achieve automati...
Citations
Each individual possesses a unique signature that is primarily employed to verify personal identity and authenticate legally binding documents or facilitate significant transactions, a method commonly utilized for verifying their identity. The utilization of this technology is restricted to the authentication of biometric recognition in a range of financial, legal, banking, insurance, and various other business documents. Techniques for recognizing signatures are employed to determine the specific user associated with a particular signature. In recent years, a significant number of researchers have focused on the implementation of novel approaches in this area, with a notable increase in the prevalence of deep learning techniques. To enhance the understanding of the evolution of offline handwritten signature recognition among researchers, this manuscript adopts a structured methodology to categorize this research, drawing primarily from studies found in set major databases. This study assesses methodologies for offline handwritten signature recognition by implementing predetermined inclusion and exclusion criteria. It explores various aspects, such as feature extraction and challenges in classification. In recent years, there have been noticeable advances and new developments. The paper accentuates the dominance of deep learning research directions in this specific domain. Differing from existing surveys, this paper does not confine itself to a particular research phase but meticulously outlines each stage, aspiring to guide future researchers in their investigations.
The acquisition of a fluid and legible handwriting in elementary school has a positive impact on multiple skills (e.g., reading, memory, and learning of novel information). In recent years, the growing percentages of children that encounter mild to severe difficulties in the acquisition of grapho-motor parameters (GMPs) has highlighted the importance of timely and reliable assessments. Unfortunately, currently available tests relying on pen and paper and human-based coding (HBC) require extensive coding time, and provide little or no information on motor processes enacted during handwriting. To overcome these limitations, this work presents a novel screen-based platform for Grapho-motor Handwriting Evaluation & Exercise (GHEE). It was designed to support both fully automatic machine-based coding (MBC) of quantitative GMPs and human-machine interaction coding (MBC+HBC) of GMPs accounting for qualitative aspects of a child’s personal handwriting style (i.e., qualitative GMPs). Our main goal was to test: the GHEE coding approach in a relevant environment to assess its reliability compared to HBC; the efficacy of human-machine interaction in supporting coding of qualitative GMPs; and the possibility to provide data on kinematic aspects of handwriting. The preliminary results on 10 elementary school children showed reliability of fully automatic MBC of quantitative GMPs with respect to traditional HBC, a higher resolution of mixed human-machine interaction systems in assessing qualitative GMPs, and suitability of this technology in providing new information on handwriting kinematics.
