Classifying handwriting samples according to their type (i.e. natural, disguised, traced, simulated or unintentionally unnatural) is an important task in handwriting analysis. It may facilitate the ...collection of writing standards and also help experts to assess the differences between questioned material and comparative samples or to choose the best writing features and the most relevant examination protocol for the case. Current research aimed to create a method for classifying the type of a handwriting sample using discriminant analysis. Five basic types (i.e. natural, disguised, traced, simulated and unintentionally unnatural) and some subtypes were included in this study. Participants (N = 139) wrote their full signatures, fictional signatures or a short sentence. Motor and dimensional features were assessed. The methods proved to be more than twice as accurate in classifying samples according to their type than a random choice probability (e.g. 44% as opposed to 17% for the 6-types classifier). This proof-of-a-concept study demonstrates that handwriting samples may be classified according to their type with satisfying accuracy based on their writing features and statistical tools of discriminant analysis. However, further studies are necessary to improve the accuracy of the method.
Recognizing lines of unconstrained handwritten text is a challenging task. The difficulty of segmenting cursive or overlapping characters, combined with the need to exploit surrounding context, has ...led to low recognition rates for even the best current recognizers. Most recent progress in the field has been made either through improved preprocessing or through advances in language modeling. Relatively little work has been done on the basic recognition algorithms. Indeed, most systems rely on the same hidden Markov models that have been used for decades in speech and handwriting recognition, despite their well-known shortcomings. This paper proposes an alternative approach based on a novel type of recurrent neural network, specifically designed for sequence labeling tasks where the data is hard to segment and contains long-range bidirectional interdependencies. In experiments on two large unconstrained handwriting databases, our approach achieves word recognition accuracies of 79.7 percent on online data and 74.1 percent on offline data, significantly outperforming a state-of-the-art HMM-based system. In addition, we demonstrate the network's robustness to lexicon size, measure the individual influence of its hidden layers, and analyze its use of context. Last, we provide an in-depth discussion of the differences between the network and HMMs, suggesting reasons for the network's superior performance.
Reconstructing the trajectory from the static image of handwritten ink traces is useful in many practical applications envisaging handwriting analysis and recognition from offline data, as it allows ...the use of methods, algorithms, and tools that deal with online data, achieving better results than those achieved on offline data. In this work, the trajectory recovery is addressed by combining a general graph traversal algorithm with knowledge about the processes involved in human learning of motor skills to perform voluntary and complex movements. The effectiveness of the proposed approach has been quantitatively and extensively evaluated on large and publicly available datasets, containing English and French multi-stroke words and isolated characters. The experimental results show that our approach outperforms the existing ones in terms of Root Mean Square Error and Dynamic Time Warping distance between the recovered trajectories and the actual ones. Furthermore, an “off-the-shelf” online recognition system provided with the trajectory recovered from offline samples showed an overall reduction of 6.8% with respect to the recognition rate achieved by the system when provided with online data; the reduction, however, drops to 2.4% once preprocessing errors are not taken into account.
Reservoir computing systems utilize dynamic reservoirs having short-term memory to project features from the temporal inputs into a high-dimensional feature space. A readout function layer can then ...effectively analyze the projected features for tasks, such as classification and time-series analysis. The system can efficiently compute complex and temporal data with low-training cost, since only the readout function needs to be trained. Here we experimentally implement a reservoir computing system using a dynamic memristor array. We show that the internal ionic dynamic processes of memristors allow the memristor-based reservoir to directly process information in the temporal domain, and demonstrate that even a small hardware system with only 88 memristors can already be used for tasks, such as handwritten digit recognition. The system is also used to experimentally solve a second-order nonlinear task, and can successfully predict the expected output without knowing the form of the original dynamic transfer function.
•Handwriting evaluation for young children learning.•Handwriting analysis and segmentation.•Handwriting learning with digital pen-based tablets.
As part of an innovative e-education project, a ...digital workbook is being developed to help teach handwriting at school for children aged three to seven. The main objective of this project is to offer an advanced digital writing experience at school by using pen-based tablets. In this paper, an automatic qualitative analysis process of cursive handwriting words is presented. This approach is original because the goal is not to recognise the word that was handwritten by children (it is an explicit instruction) but to design a precise evaluation of the quality of his handwriting production to give them a real-time feedback. The presented method is based on a specific explicit elastic letter spotting segmentation able to deal with the imprecision of the handwriting of young children. This approach is suited to automatically and precisely highlight the difficulties encountered by children (adding or missing letters, incorrect shapes...). The validation of the proposed approach has been done on a dataset collected in French preschools and primary schools from 231 children. Beyond quantitative results, this paper reports the very positive impact of using this digital workbook that allows children to work independently with online and real-time feedbacks.
This paper demonstrates a framework for offline handwriting recognition using character spotting and autonomous tagging which works for any alphabetic script. Character spotting builds on the idea of ...object detection to find character elements in unsegmented word images. An autonomous tagging approach is introduced which automates the production of a character image training set by estimating character locations in a word based on typical character size. Although scripts can vary vividly from each other, our proposed approach provides a simple and powerful workflow for unconstrained offline recognition that should work for any alphabetic script with few adjustments. Here we demonstrate this approach with handwritten Bangla, obtaining a character recognition accuracy (CRA) of 94.8% and 91.12% with precision and autonomous tagging, respectively. Furthermore, we explained how character spotting and autonomous tagging can be implemented for other alphabetic scripts. We demonstrated that with handwritten Hangul/Korean obtaining a Jamo recognition accuracy (JRA) of 93.16% using a tiny fraction of the PE92 training set. The combination of character spotting and autonomous tagging takes away one of the biggest frustrations—data annotation by hand, and thus, we believe this has the potential to revolutionize the growth of offline recognition development.
Using memristor crossbar arrays to accelerate computations is a promising approach to efficiently implement algorithms in deep neural networks. Early demonstrations, however, are limited to ...simulations or small‐scale problems primarily due to materials and device challenges that limit the size of the memristor crossbar arrays that can be reliably programmed to stable and analog values, which is the focus of the current work. High‐precision analog tuning and control of memristor cells across a 128 × 64 array is demonstrated, and the resulting vector matrix multiplication (VMM) computing precision is evaluated. Single‐layer neural network inference is performed in these arrays, and the performance compared to a digital approach is assessed. Memristor computing system used here reaches a VMM accuracy equivalent of 6 bits, and an 89.9% recognition accuracy is achieved for the 10k MNIST handwritten digit test set. Forecasts show that with integrated (on chip) and scaled memristors, a computational efficiency greater than 100 trillion operations per second per Watt is possible.
Large memristor arrays composed of hafnium oxide are demonstrated with suitability for computing matrix operations at higher power efficiency than digital systems. The nonmemory application of memristors is performed in an analog computing platform. Computational operations with 6 bit equivalent precision are shown and utilized to directly compute neural network inference within a memristor crossbar.
Although several types of architecture combining memory cells and transistors have been used to demonstrate artificial synaptic arrays, they usually present limited scalability and high power ...consumption. Transistor-free analog switching devices may overcome these limitations, yet the typical switching process they rely on-formation of filaments in an amorphous medium-is not easily controlled and hence hampers the spatial and temporal reproducibility of the performance. Here, we demonstrate analog resistive switching devices that possess desired characteristics for neuromorphic computing networks with minimal performance variations using a single-crystalline SiGe layer epitaxially grown on Si as a switching medium. Such epitaxial random access memories utilize threading dislocations in SiGe to confine metal filaments in a defined, one-dimensional channel. This confinement results in drastically enhanced switching uniformity and long retention/high endurance with a high analog on/off ratio. Simulations using the MNIST handwritten recognition data set prove that epitaxial random access memories can operate with an online learning accuracy of 95.1%.
LSTM: A Search Space Odyssey Greff, Klaus; Srivastava, Rupesh K.; Koutnik, Jan ...
IEEE transaction on neural networks and learning systems,
10/2017, Letnik:
28, Številka:
10
Journal Article
Odprti dostop
Several variants of the long short-term memory (LSTM) architecture for recurrent neural networks have been proposed since its inception in 1995. In recent years, these networks have become the ...state-of-the-art models for a variety of machine learning problems. This has led to a renewed interest in understanding the role and utility of various computational components of typical LSTM variants. In this paper, we present the first large-scale analysis of eight LSTM variants on three representative tasks: speech recognition, handwriting recognition, and polyphonic music modeling. The hyperparameters of all LSTM variants for each task were optimized separately using random search, and their importance was assessed using the powerful functional ANalysis Of VAriance framework. In total, we summarize the results of 5400 experimental runs (≈15 years of CPU time), which makes our study the largest of its kind on LSTM networks. Our results show that none of the variants can improve upon the standard LSTM architecture significantly, and demonstrate the forget gate and the output activation function to be its most critical components. We further observe that the studied hyperparameters are virtually independent and derive guidelines for their efficient adjustment.
This article comprehensively surveys Arabic Online Handwriting Recognition (AOHR). We address the challenges posed by online handwriting recognition, including ligatures, dots and diacritic problems, ...online/offline touching of text, and geometric variations. Then we present a general model of an AOHR system that incorporates the different phases of an AOHR system. We summarize the main AOHR databases and identify their uses and limitations. Preprocessing techniques that are used in AOHR, viz. normalization, smoothing, de-hooking, baseline identification, and delayed stroke processing, are presented with illustrative examples. We discuss different techniques for Arabic online handwriting segmentation at the character and morpheme levels and identify their limitations. Feature extraction techniques that are used in AOHR are discussed and their challenges identified. We address the classification techniques of non-cursive (characters and digits) and cursive Arabic online handwriting and analyze their applications. We discuss different classification techniques, viz. structural approaches, Support Vector Machine (SVM), Fuzzy SVM, Neural Networks, Hidden Markov Model, Genetic algorithms, decision trees, and rule-based systems, and analyze their performance. Post-processing techniques are also discussed. Several tables that summarize the surveyed publications are provided for ease of reference and comparison. We summarize the current limitations and difficulties of AOHR and future directions of research.