Showing papers on "Signature recognition published in 2021"

Offline Signature Recognition Using Image Processing Techniques and Back Propagation Neuron Network System

Abstract: As the technology improves, there are many new innovations which give the security frameworks with different methods that are used to identify a person. Signature recognition is one of methods used to identify person. In this paper, offline signature recognition using back propagation neuron network system and image processing techniques has been proposed. Preprocessing of signature can be done using image processing techniques which involves RGB2Gray conversion, filtering, adjusting, thresholding followed by canny edge detection and at the last image scaling applied to reduce the processing time. Processed image feature is extracted using back propagation neuron network system with defined number of neurons and hidden layers. Similarly, data set images undergo preprocessing operation and features are extracted. Based on number of layers that are hidden and neuron, better recognition rate is obtained. The proposed method shows that the experimental result has more success rate.

Offline Signature Recognition Using Deep Features

Abstract: Offline Signature Verification plays a significant role in Forensic offices. In our research, we investigate the performance of signature recognition utilizing features based on AlexNet (pretrained Convolution Neural Network model). All the investigations are performed on scanned signatures of fourteen users each from three datasets, namely CEDAR, UTSig, and BHsig260. Two classifiers, i.e., Support Vector Machine (SVM) and Decision Tree are utilized. Utilizing features based on Deep Convolution neural network and SVM as machine learning algorithms show better outcomes. The best output is achieved for Bengali signature recognition utilizing SVM with 100% accuracy. For Persian signature, we obtained an accuracy of more than 80% for each user. Twelve users out of fourteen users for Hindi signature are 100% recognized using SVM.

Multimodal biometrics of fingerprint and signature recognition using multi-level feature fusion and deep learning techniques

Abstract: Providing security in biometrics is the major challenging task in the current situation. A lot of research work is going on in this area. Security can be more tightened by using complex security systems, like by using more than one biometric trait for recognition. In this paper multimodal biometric models are developed to improve the recognition rate of a person. The combination of physiological and behavioral biometrics characteristics is used in this work. Fingerprint and signature biometrics characteristics are used to develop a multimodal recognition system. Histograms of oriented gradients (HOG) features are extracted from biometric traits and for these feature fusions are applied at two levels. Features of fingerprint and signatures are fused using concatenation, sum, max, min, and product rule at multilevel stages, these features are used to train deep learning neural network model. In the proposed work, multi-level feature fusion for multimodal biometrics with a deep learning classifier is used and results are analyzed by a varying number of hidden neurons and hidden layers. Experiments are carried out on SDUMLA-HMT, machine learning and data mining lab, Shandong University fingerprint datasets, and MCYT signature biometric recognition group datasets, and encouraging results were obtained.

Spatiotemporal spectral histogramming analysis in hand gesture signature recognition

Abstract: Dynamic signature recognition emerges to perfectly solve the hygiene concern due to its no-contact characteristic. Nevertheless, the recognition of dynamic texture is challenging compared with the static signature image due to their unknown spatial and temporal nature. In this work, we present a multi-view spatiotemporal approach based on spectral histogramming for hand gesture signature recognition. A Microsoft Kinect sensor is adopted to capture the motion of signing in a sequence of depth frames. The depth frame sequence is viewed from three directional sights to retrieve rich information, such as temporal changes at each spatial location, the signing motion flow of each vertical and horizontal spatial space in a temporal manner. Furthermore, the proposed approach performs feature description on different levels of locality. This function enables a multi-resolution analysis on this dynamic signature. The robustness of the proposed approach is reflected with the promising result by striking the state-of-the-art performance, as substantiated in the empirical results.

In-Air 3D Dynamic Signature Recognition using Haptic Devices

Abstract: Signature recognition is one of the most widespread and legally accepted methodology to authenticate a person’s identity. In this work, we show how a haptic device can be used to acquire in-air 3D signatures, and provide the time-dependent position and orientation characteristics needed to effectively perform user verification. Dynamic time warping and hidden Markov models are here employed to compare samples acquired during the enrolment and verification stages. The recognition performance achieved when testing the proposed system on samples captured from 52 subjects testify the effectiveness of the proposed approach. Furthermore, a longitudinal analysis carried out on data from a subset of 21 subjects, for which two recording sessions have been taken at an average distance of four months, demonstrates that effective recognition can be performed even at long time distances from the enrolment.

A Review on Pattern Recognition Using Machine Learning

Abstract: Machine learning (ML) techniques have gained remarkable attention in past two decades including many fields like computer vision, information retrieval, and pattern recognition. This paper presents a literature review on pattern recognition of various applications like signal processing, agriculture sector, healthcare sector, signature recognition, and different model analysis using ML techniques. The focus of our survey is at the ML techniques, classification techniques and deep learning model, and improves the accuracy rate for the automatic decision making algorithms.

Real-Time Signature Recognition Using Neural Network

Abstract: The human signature is an important biometric feature that is used to identify human identity. It is essential in preventing falsification of documents in numerous financial, legal, and other commercial settings. The computerized system enters many aspects of our life, security is one of them, continues developing in computer vision and artificial network leads researcher to develop computerized signature recognition. This paper proposed a real-time algorithm for signature recognition. It is based on client and server operation.in which, client agent captures a signature and sends it to the server through the network. The server receives data and performs processing on it. Processing algorithm is based on weightless neural network. It is chosen for its simplicity and few numbers of sample required for training. The algorithm is tested and evaluated and show the ability to process 4.7 images per second.

Deep Learning Based Handwritten Signature Recognition

Abstract: A handwritten signature commonly practiced route for confirming the authenticity of legal documents. The verification of the signature is critical as it varies every time and may change with age, behavior, and environment. This paper presents a Deep learning model based on the CNN architecture to verify the signature. For experimental purpose, the feature extraction portion of the GoogleNet model has been used to transfer value calculation and the classification layer was retrained using back propagation with the concept of transfer learning. The classification layer of the Deep learning model was retrained with 25 classes of signature image dataset with each class consisting of 85 signatures. After training, the model was evaluated with a testing dataset of 15 signatures from each class. The mean testing precision of the neural network architecture with signature dataset was found to be 95.2 %.

Feature Selection for Handwritten Signature Recognition Using Neighborhood Component Analysis

Abstract: Advance trends adopted in classification of images using deep learning networks have paved opportunities for handwritten signature identification and verification. Convolution neural network has demonstrated to be profoundly proficient in image classification, yet this methodology has a few weaknesses also. Notably, a huge number of images are required for training to achieve high level of a precision for classification. Transfer learning with pretrained deep convolution neural network model can be utilized to conquer these issues. To bring about further enhancements in terms of recognition rate, storage and computational time, feature selection procedure plays a key role by distinguishing the most significant features. In this work, we used a pretrained DCNN model, i.e., GoogLeNet for feature extraction using transfer learning and further used neighborhood component analysis to reduce the irrelevant features. Support vector machine is used for classification. Experiments have been conducted using three signature datasets, i.e., CEDAR, SigComp2011 and UTSig.

Robustness for Authentication of the Human Using Face, Ear, and Gait Multimodal Biometric System

Abstract: Biometrics is the science that deals with personal human physiological and behavioral characteristics such as fingerprints, handprints, iris, voice, face recognition, signature recognition, ear recognition, and gait recognition. Recognition using a single trait has several problems and multimodal biometrics system is one of the solutions. In this work, the novel and imperative biometric feature gait is fused with face and ear biometric features for authentication and to overcome problems of the unimodal biometric recognition system. The authors have also applied various normalization methods to sort out the best solution for such a challenge. The feature fusion of the proposed multimodal biometric system has been tested using Min-Max and Z-score techniques. The computed results demonstrate that Z-Score outperforms the Min-Max technique. It is deduced that the Z-score is a promising method that generates a high recognition rate of 95% and a false acceptance rate of 10%.

Backpropagation Algorithm on Implementation of Signature Recognition

Abstract: Many things are required by all parties, especially in the process of recognition of one's identity, ranging from health care, maintenance of bank accounts, aviation services, immigration and others.Many ways of proving one's identity and the most popular one is using a signature.The signature is used as an identification system which serves to recognize a person's identity.Recognition process is still done manually by matching the signature by the person concerned.Therefore, the very need for a system that is able to analyze and identify the characteristics of the signature, so it can be used as an alternative to simplify the process of introducing people’s signature.Artificial neural networks can be used as one of the solutions in identification of signatures.Artificial neural network is a branch of science of artificial intelligence that is capable of processing information with the performance of certain characteristics.Artificial neural networks have some method such as perceptron, Hopfield discrete, Adaline, Backpropagation, and Kohonen.In this paper, the artificial neural network with back propagation method is applied in the process of signature and patternrecognition which provided a solution that is able to analyze and recognize people's signature.Implementation of the application of neural networks in pattern recognition signature can further be applied to any computer that handles problems in the process of matching one's data.

Handprinted Character and Online Signature Recognition Using Residual Convolutional Network: A Comparative Study

Abstract: Residual convolutional networks proved superiority in the image recognition field in addition to many other pattern recognitions related problems. A special benefit of residual convolutional networks has been proven in the field of handwriting recognition, especially in handprinted characters and online signature recognition. Two published research papers addressed the two problems with highly accurate results that compete with the state-of-the-art techniques. Both problems have two well-known challenges in object classification: intra-class variability and inter-class similarity. In the character recognition problem, different users produce different shapes for the same character in addition to the user himself, who produces different shapes for the same character resulting in high intra-class variability. Inter-class similarity exists due to the similarity between different characters. In handwritten signature recognition, intra- class variability is introduced by the same person where no two signatures of the same person may coincide. On the other hand, persons may have similar signatures resulting in inter-class similarity. Solving the two pattern recognition problems using deep learning techniques requires the selection of the best technique that can model the handwriting process, but at the same time, it requires a large training data. In the character recognition case, large training samples are available from many datasets resources. A recent data set representing this challenge is the EMINST dataset. An optimized residual architecture has been introduced to give an excellent solution for this problem in one of the comparison papers. In the case of signature recognition, a different solution based on a residual network has been introduced in a second paper because practically training samples that can be collected from new users are very rare. In this paper, the two techniques are compared, and conclusions that may generalize well in other problem domains are stated.

Scaled-Magnitude Multi-Channel Correlation Filters for Multimodal Biometric Recognition

Abstract: We propose a novel variant of the multi-channel correlation filters (MCCF), namely the scaled-magnitude MCCF (SM-MCCF). The SM-MCCF is characterized by a scaling factor on the magnitude response, which has phase-only spectrum and conventional magnitude and phase spectra as the corner cases. We show that the SM-MCCF design technique, when applied to a multimodal biometric authentication system based on face and handwritten signature recognition, outperforms the conventional MCCF and SVM classifiers under low SNR conditions. Furthermore, the utility of the SM-MCCF is also explored for multimodal fusion with image features for face and handwritten signatures with i-vectors for speech data. Our experimental results indicate that SM-MCCF provides a reasonable improvement in performance, in terms of the EER and recognition rate, as opposed to the MCCF in both moderately and severely degraded scenarios. Moreover, we also demonstrate that the feature level fusion is advantageous than score fusion as the level of abstraction in feature representation is lesser when compared to score level representations.

Preliminary Study on the Behavioral Traits Obtained from Signatures and Writing Using Deep Learning Algorithms

Abstract: This work describes how a new multimodal data base is constructed to try to solve the link between signature/handwriting and personality. With the help of two devices, one of them responsible to report the mechanical writing process (using a tablet) and the other one to acquire brain activity (with an EEG headset) we will be able to carry out different sessions through a set of experiments. Because the data base is not completed yet, and it is well known that deep learning requires larges amount of data, the main results about signature and personality factors were not good enough. The different deep convolutional neural networks (DCNN) tested does not obtain a reasonable minimum threshold. However the same incomplete data base gives promising results when solving a completely different problem such as signature recognition (where a performance of 80% was reached) using the same DCNN architecture.

Intensified Biological Signature Recognition in the Wild: A Case Study

Abstract: The biological signature recognition seems to be a solved problem, but when these systems are tested in the wild their accuracy collapsed abruptly. A case study of “Face Recognition Through Camera (FRTC)” was designed to make a state of art for biological signature recognition in the wild system with minimum data set required to provide the highest accuracy. The FRTC system was designed by integrating the best qualities of modern recognition systems present in recent times (OpenCV and Face API) to intensify the biological signature recognition in the wild. The case study FRTC would be tested with the classroom attendance without prior information to the students. It was expected that FRTC would be more accurate than OpenCV and Face API. The results have shown the FRTC is more accurate, 7.02% and -1% than OpenCV and Face API respectively.

Multi-lingual Hybrid Handwritten Signature Recognition Based on Deep Residual Attention Network

Abstract: The writing styles of Uyghur, Kazak, Kirgiz and other ethnic minorities in Xinjiang are very similar, so it is extremely difficult to extract the effective features of handwritten signatures of different languages by hand. To solve this problem, a multi-lingual hybrid handwritten signature recognition method based on deep residuals attention network was proposed. Firstly, an offline handwritten signature database in Chinese, Uyghur, Kazak and Kirgiz was established, with a total of 8,000 signed images. Then, the signature image is pre-processed by grayscale, median filtering, binarization, blank edge removal, thinning and size normalization. Finally, transfer learning method is used to input the signature image into the deep residual network, and the high-dimensional features are extracted automatically by the fusion channel attention for classification. The experimental results show that the highest recognition accuracy of this method is 99.44% for multi-lingual hybrid handwritten signature database, which has a high application value.

Dynamic handwritten signature identification using spiking neural network

Abstract: The article proposes a method for dynamic signature identification based on a spiking neural network. Three dynamic signature parameters l(t), xy(t), p(t) are used, which are invariant to the signature slope angle, and after their normalization, also to the signature spatial and temporal scales. These dynamic parameters are fed to the spiking neural network for recognition simultaneously in the form of time series without preliminary transformation into a vector of static features, which, on the one hand, simplifies the method due to the absence of complex computational transformation procedures, and on the other hand, prevents the loss of useful information, and therefore increases the accuracy and reliability of signature identification and recognition (especially when recognizing forged signatures that are highly correlated with the genuine). The spiking neural network used has a simple training procedure, and not all neurons of the network are trained, but only the output ones. If it is necessary to add new signatures, it is not necessary to retrain the entire network as a whole, but it is enough to add several output neurons and learn only their connections. In the results of experimental studies of the software implementation of the proposed system, it’s EER = 3.9% was found when identifying skilled forgeries and EER = 0.17% when identifying random forgeries.