Signature recognition

About: Signature recognition is a research topic. Over the lifetime, 2138 publications have been published within this topic receiving 37605 citations.

Gesture recognition using active body parts and active difference signatures

Abstract: The introduction of low cost depth cameras along with advances in computer vision have spawned an exciting new era in Human Computer Interaction. Real time gesture recognition systems have become commonplace and attention has now turned towards making these systems invariant to within-user and user-to-user variation. Active difference signatures have been used to describe temporal motion as well as static difference from a canonical resting position. Geometric features, such as joint angles, and joint topological distances can be used along with active difference signatures as salient feature descriptors. To achieve robustness to natural gesture variation, this paper introduces active body part recognition along with these features into the Hidden Markov Model framework. The proposed method is bench-marked against other methods, achieving state of the art results on the MSR3D and ChaLearn datasets.

Intelligent data recognition of DNA sequences using statistical models

Abstract: The intelligent data acquisition in biological sequences is a hard and challenge problem since most biological sequences contain unknowledgeable, diverse and huge data. However, the intelligent data acquisition reduces a demand on the use of high computation methods because the data are more compact and more precise. We propose a novel approach for discovering sequence signatures, which are sufficiently distinctive information in identifying the sequences. The signatures are derived from the best combination of the n-grams and the statistical scoring models. From our experiments in applying them to identify the Influenza virus, we found that the identifiers constructed by too short n-gram signatures and inappropriate scoring models get low efficiency since the inappropriate combinations of n-gram signatures and scoring models bring about unbalanced class and pattern score distribution. However, the other identifiers provide accuracy over 80% and up to 100%, when they apply an appropriate combination. In addition to accomplishing in the signature recognition, our proposed approach also requires low computation time for the biological sequence identification.

Dynamic signature recognition method based on touch screen mobile device

Abstract: The invention relates to a dynamic signature recognition method based on a touch screen mobile device, and belongs to the technical field of bioidentification. On the basis of easily-obtained signature polar angle characteristics, user signatures are matched to achieve dynamic signature recognition on a mobile terminal. The dynamic signature recognition method comprises the following steps: firstly, preprocessing an initial user signature data obtained by sampling to enable two adjacent points to have equal time interval, and on the basis of a centroid of a signature in an X-Y plane and a golden centroid, normalizing the size of a character shape; secondly, selecting a pole to establish a polar coordinate system, calculating characteristic sequences of an polar angle and a polar radius of the signature, and extracting stable polar value points from the polar angle characteristics as separating points; finally, on the basis of sample signatures to be measured and a separating point sequence of a template, searching and matching to obtain a global optimal matching scheme and the best similarity so as to judge the identity of a signer. According to the technical scheme, signature recognition on the terminal has the advantages of low cost, high accuracy, high calculating speed and the like.

Signature Detection and Identification Algorithm with CNN, Numpy and OpenCV

Abstract: The purpose of the article is to present a simple signature detection algorithm and its subsequent signature identification using a deep learning model for processing images based on a convolutional neural network. To solve the task of the image recognition, a binary classification has been performed to predict text or signature and signature classifications to determine the author of this signature. The proposed algorithm is interesting in the preliminary processing of scanned documents with signatures in order to extract the area with the signature and transfer it to the trained model. The research results are presented for documents of the same type, in which the signature is located in the same place. To select a specific element in the document we are using the tensor-slicing operations on Numpy arrays. To extract areas with text and signature, OpenCV tools are used. The results on the ready-made neural network model studies on a small dataset are presented in this article. Good results have been achieved in recognizing the famous writers’ signatures. The proposed algorithm demonstrates the possibility of using the classical convolution network model for solving specific practical problems. The studies can be recommended to students in the study of neural networks to understand the basics of deep learning and apply a ready-made model as a template for solving practical problems in the field of computer vision.