Handwriting Synthesis Research Articles

DeepWriteSYN: On-Line Handwriting Synthesis via Deep Short-Term Representations

This study proposes DeepWriteSYN, a novel on-line handwriting synthesis approach via deep short-term representations. It comprises two modules: i) an optional and interchangeable temporal segmentation, which divides the handwriting into short-time segments consisting of individual or multiple concatenated strokes; and ii) the on-line synthesis of those short-time handwriting segments, which is based on a sequence-to-sequence Variational Autoencoder (VAE). The main advantages of the proposed approach are that the synthesis is carried out in short-time segments (that can run from a character fraction to full characters) and that the VAE can be trained on a configurable handwriting dataset. These two properties give a lot of flexibility to our synthesiser, e.g., as shown in our experiments, DeepWriteSYN can generate realistic handwriting variations of a given handwritten structure corresponding to the natural variation within a given population or a given subject. These two cases are developed experimentally for individual digits and handwriting signatures, respectively, achieving in both cases remarkable results. Also, we provide experimental results for the task of on-line signature verification showing the high potential of DeepWriteSYN to improve significantly one-shot learning scenarios. To the best of our knowledge, this is the first synthesis approach capable of generating realistic on-line handwriting in the short term (including handwritten signatures) via deep learning. This can be very useful as a module toward long-term realistic handwriting generation either completely synthetic or as natural variation of given handwriting samples.

Assessing similarity in handwritten texts

Today, people rely almost full time on digital texts. It is not surprising that handwriting earned a special status, and solutions to mimic real handwriting became attractive. A particular field called handwriting synthesis generates renderings of text which resemble natural writing but are synthesized from actual handwriting samples. The main idea behind samples’ current solutions is to collect enough samples to capture a given subject’s writing style, and therefore be able to reproduce it in new texts, with natural variability. Nevertheless, the question remains of how much input variability is enough to represent specific handwriting. In this paper, we address sample acquisition for handwriting synthesis. We conducted a study comparing written text similarity between two sets of samples, one using augmented pangrams (with a total of 473 characters) and the other using general texts (with 1586 characters). Our results show that the samples collected with pangrams are statistically equivalent in variation with samples collected using general texts, with many benefits, particularly the shorter time needed to collect the samples. We also made our data collection publicly available, providing a valuable original resource for future research.

Handwriting perceptual classification and synthesis using discriminate HMMs and progressive iterative approximation

This paper handles the problem of online handwriting synthesis. Indeed, this work presents a probabilistic model using the embedded hidden Markov models (HMMs) for the classification and modeling of perceptual sequences. At first, we start with a vector of perceptual points as input seeking a class of basic shape probability as output. In fact, these perceptual points are necessary for the drawing and the recovering of each basic shape where each one is designed with an HMM built and trained with its components. Each path through these possibilities of control points represents an observation that serves as input for the following step. Secondly, the already detected sequences of observations which represent a segment formed an initial HMM and the concatenation of multiple ones leads to a global HMM. To classify a global HMM, we should codify it by searching the best path of initial HMM. The best path is obtained by computing the maximum of likelihood of the different basic shapes. In order to synthesize the handwritten trace, and to recover the best control points sequences, we investigated the progressive iterative approximation. The performance of the proposed model was assessed using samples of scripts extracted from IRONOFF and MAYASTROON databases. In fact, these samples served for the generation of the set of control points used for the HMMs training models. In experiments, good quantitative agreement and approximation were found for the generated trajectories and a more reduced representation of the scripts models was designed.

Handwriting synthesis: classifications and techniques

Handwriting synthesis is the automatic generation of data that resemble natural handwriting. Although handwriting synthesis has recently gained increasing interest, the area still lacks a stand-alone review. This paper provides classifications for the different aspects of handwriting synthesis. It presents the applications, techniques, and evaluation methods for handwriting synthesis based on the several aspects that we identify. Then, it discusses various synthesis techniques. To the best of our knowledge, this paper is the only stand-alone survey on this topic, and we believe it can serve as a useful reference for the researchers in the field of handwriting synthesis.

An Arabic handwriting synthesis system

In this paper, we present an Arabic handwriting synthesis system. Two concatenation models to synthesize Arabic words from segmented characters are adopted: Extended-Glyphs connection and Synthetic-Extensions connection. We use our system to synthesize handwriting from a collected dataset and inject it into an expanded dataset. We experiment by training a state-of-the-art Arabic handwriting recognition system on the collected dataset, as well as on the expanded dataset, and test it on the IFN/ENIT Arabic benchmark dataset. We show significant improvement in recognition performance due to the data that was synthesized by our system.

Probabilistic Synthesis of Personal-Style Handwriting

The goal of personal-style handwriting synthesis is to produce texts in the same style as an individual writer by analyzing the writer's samples of handwriting. The difficulty of handwriting synthesis is that the output should have the characteristics of the person's handwriting as well as looking natural, based on a limited number of available examples. We develop a synthesis algorithm which produces handwriting that exhibits naturalness based on the probabilistic character model.

Style-preserving English handwriting synthesis

This paper presents a novel and effective approach to synthesize English handwriting in the user's writing style. We select the most important features that depict the handwriting style, including character glyph, size, slant, and pressure, special connection style, letter spacing, and cursiveness. The features can be efficiently computed with the aid of our specially designed sample collecting user interface. Given ASCII text, the user handwriting is synthesized hierarchically. First, character glyphs are sampled and shape variation is added. Second, words are generated by aligning the character glyphs on the baseline with proper horizontal inter-character space and vertical offset from the baseline. The heads and tails of the letters may be trimmed to avoid severe overlap and facilitate possible connections between neighboring letters. Adjacent letters may be connected to each other by polynomial interpolation. Finally, after the pressure is assigned, the handwriting is rendered word by word and then line by line. The experimental results prove the capability of our system to adapt to the user's writing style.

Latent Doodle Space

Abstract We propose the concept of a latent doodle space, a low‐dimensional space derived from a set of input doodles, or simple line drawings. The latent space provides a foundation for generating new drawings that are similar, but not identical to, the input examples. The two key components of this technique are 1) a heuristic algorithm for finding stroke correspondences between the drawings, and 2) the use of latent variable methods to automatically extract a low‐dimensional latent doodle space from the inputs. We present two practical applications that demonstrate the utility of this idea: first, a randomized stamp tool that creates a different image on every usage; and second, “personalized probabilistic fonts,” a handwriting synthesis technique that mimics the idiosyncrasies of one's own handwriting. Keywords: sketch, by‐example, style learning, scattered data interpolation, principal component analysis, radial basis functions, Gaussian processes, digital in‐betweening, handwriting synthesis

Combining shape and physical modelsfor online cursive handwriting synthesis

This paper proposes a novel learning-based approach to synthesizing cursive handwriting of a user's personal handwriting style by combining shape and physical models. In the training process, some sample paragraphs written by a user are collected and these cursive handwriting samples are segmented into individual characters by using a two-level writer-independent segmentation algorithm. Samples for each letter are then aligned and trained using shape models. In the synthesis process, a delta log-normal model based conditional sampling algorithm is proposed to produce smooth and natural cursive handwriting of the user's style from models.

Network-Based Approach to Korean Handwriting Analysis

It is well known that the stochastic approach using the HMM and dynamic programming-based search is particularly suited to the analysis of time series signals including on-line handwriting. The starting point of this research is a network of HMMs which models the whole set of characters. Then it is followed by the assertion that the HMM for the on-line script can be applied to not only on-line character recognition but also to the handwriting synthesis and even pen-trajectory recovery in off-line character images. The solutions to these problems are based on the single network of HMMs and the single principle of DP-based state-observation alignment. Given an observation sequence, the search for the best path in the network corresponds to the recognition. Given a character model, the search for the best observation sequence corresponds to the handwriting generation. The proposed framework has been shown to work nicely through a set of tests on Korean characters.