Pith. sign in

REVIEW 6 major objections 6 minor 40 references

Neural Architecture based on Fuzzy Perceptual Representation For Online Multilingual Handwriting Recognition

T0 review · 6 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read Decomposing online handwriting into nine fuzzy perceptual stroke codes lets LSTM recognizers exceed 98 percent accuracy on Arabic and Latin digits and words.

desk verdict Plausible pipeline, but the accuracy claims don't survive contact with the paper's own tables, and the fuzzy ground-truth protocol needs serious scrutiny before anyone bets on it. read the letter →

arxiv 1908.00634 v1 pith:OUVTO2X6 submitted 2019-08-01 cs.CV

classification cs.CV
keywords onlinehandwritingrecognitionBeta-ellipticmodelperceptualcodesfuzzylogicLSTMconvolutionalgroundtruthtrainingArabicandLatinscripts
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes two online handwriting recognition systems for Arabic and Latin scripts and claims both exceed 98 percent accuracy on several databases. The first, OnHSR-LSTM, segments handwriting into strokes with a Beta-elliptic model, labels each stroke with one of nine fuzzy perceptual codes, and trains a long short-term memory (LSTM) network with fuzzy ground truth to recognize the resulting code sequences. The second, OnHR-convLSTM, learns directly from raw (x,y,z) pen signals using a convolutional LSTM. The strongest reported result is 98.5 percent recognition of Arabic digits on the MAYASTROUN database. If the claim holds, a small fixed set of perceptual primitives plus sequence learning can replace much handwriting-specific feature engineering and transfer across scripts.

What carries the argument

The load-bearing object is the nine-code perceptual alphabet—Valley, left-oblique shaft, shaft, right-oblique shaft, four half-occlusions, and Occlusion—which the paper asserts can form any script. The Beta-elliptic model turns the pen trace into elliptic strokes bounded by three beta-points, and each stroke receives fuzzy membership in one of four elementary perceptual code regions via a hand-set overlap constant. An LSTM maps the code sequence to a label, CTC removes timing alignment, and fuzzy ground truth replaces one-hot targets with membership values computed from Euclidean distances between stroke lengths.

What would settle it

A direct reproduction of the MAYASTROUN digit experiment where the perceptual-code front end is replaced by raw (x,y,z) coordinates, keeping the same LSTM and fuzzy training, settles whether the reported 98.5 percent is caused by the perceptual representation; if raw-input accuracy is not clearly lower, the central claim is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that online handwriting is a concatenation of perceptual codes: with the Beta-elliptic model, any trace can be segmented into elliptic strokes, each stroke receives fuzzy membership over four elementary perceptual code regions, and sequences of these codes identify characters, digits, and words. On that representation, an LSTM with CTC labeling and fuzzy ground truth training reaches 98.5 percent on MAYASTROUN Arabic digits, 97.5 percent on the UNIPEN Latin set, and up to 98.4 percent on a mobile-acquired Arabic database. The companion system, OnHR-convLSTM, reads raw (x,y,z) signals through convolutional LSTM layers and reaches 98 percent on IRONOFF words. The authors therefore claim the perceptual front end is not required for high accuracy, but that it is the component that survives injected noise, dropping only to 96 percent where the raw-signal convLSTM falls to 50.45 percent.

Load-bearing premise

The load-bearing premise is that every Arabic or Latin online trace can be reliably cut into strokes that each belong to one of nine perceptual shapes, a coverage claim the paper states without proof or segmentation error analysis.

Editorial extensions

If this is right

  • On Arabic digits from MAYASTROUN, OnHSR-LSTM with fuzzy ground truth reaches 98.5 percent, exceeding the circular-trajectory baseline and the smart-device recognizer reported for the same task.
  • Fuzzy ground truth training improves accuracy over framewise training in every reported pairing: 91.5 to 97.5 percent on LMCA/ADAB letters, 90.5 to 97.5 percent on UNIPEN set 1(c), and 90 to 96 percent on ADAB.
  • OnHR-convLSTM reaches 98 percent on IRONOFF words and 97 percent on ADAB words, making the raw-signal convolutional LSTM the stronger word-level pipeline in these experiments.
  • Under injected noise, OnHSR-LSTM keeps 96 percent accuracy while OnHR-convLSTM on raw (x,y,z) drops to 50.45 percent, supporting the claim that Beta-elliptic preprocessing removes tremor and sampling noise before classification.
  • The same nine-code perceptual alphabet is used for both Arabic and Latin scripts, so the system is claimed to be multilingual without changing the feature representation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural test the authors do not run is to apply the same nine-code pipeline to another cursive script such as Farsi or Urdu; if the alphabet is truly universal, accuracy should remain in the high 90s with no architectural change.
  • Because fuzzy membership assigns each stroke to several codes with degrees, the method could flag low-confidence strokes for human review or active learning; the paper reports accuracies but does not analyze per-stroke ambiguity.
  • The 50.45 percent collapse of the raw convLSTM under noise suggests the perceptual front end functions as an implicit denoiser; comparing it against standard tremor-removal filters on Parkinsonian handwriting would quantify how much of the gain is segmentation versus preprocessing.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

6 major / 6 minor

Summary. The paper proposes two online handwriting recognition systems: OnHSR-LSTM, which segments a script into Beta-elliptic strokes, converts each stroke into fuzzy Elementary and Basic Perceptual Codes, and classifies the resulting code sequence with an LSTM and CTC; and OnHR-convLSTM, which applies a convolutional LSTM directly to raw (x,y,z) traces. Experiments are reported on MAYASTROUN, LMCA/ADAB, UNIPEN, IRONOFF, and an in-house mobile database, with rates up to 98.5%. The abstract claims that the recognition rate exceeds 98%, but this is contradicted by several rows in the paper's own tables. The best result for system 1 is obtained under a 'Fuzzy Ground Truth' training protocol whose target memberships are computed from the same stroke-length features that form the network input, which makes the reported gain over framewise training difficult to interpret.

Significance. If the reported results were valid, the paper would offer a distinctive integration of perceptual-code theory with deep sequence classifiers for multilingual online handwriting, with potential applications to Arabic and Latin scripts and to tremor/robustness analysis. The paper's strengths include the use of an established Beta-elliptic segmentation model, the use of CTC for sequence alignment, and the application of data augmentation and dropout. However, the central quantitative claim is not currently supported: the abstract's >98% statement does not match the tabulated rates, the evaluation protocol is severely underspecified, and the best system-1 result relies on a training-target construction that is a deterministic function of the input representation. The contribution as presented is therefore not established, despite the plausibility of the general approach.

major comments (6)
  1. [Abstract and Tables 4-9] The abstract's claim that 'the recognition rate exceeds 98%' is not supported by the paper's own numbers. Table 4 reports framewise rates of 91.5%, 94.5%, and 90.0% and Fuzzy Ground Truth rates of 97.5%, 98.5%, and 96.0%; Table 5 reports 90.5% and 97.5%; Table 9 reports 93% and 98%. Only a few settings reach or exceed 98%, so the unqualified headline claim is contradicted by the experimental section.
  2. [Section 4.3.2, Algorithm 2] The Fuzzy Ground Truth training target is constructed as the Euclidean distance between a reference stroke-length series {lr} and the input-derived segmented stroke-length series {li}, both obtained from the Beta-elliptic segmentation that constitutes the network's input representation. This makes the target a deterministic function of the input, so the LSTM can learn to approximate a distance-to-prototype mapping without necessarily learning the semantic perceptual-code recognition the paper claims. The headline 98.5% MAYASTROUN result in Table 4 is exactly the Fuzzy Ground Truth condition, and no independent label source, split, or prototype construction details are given; this circularity is load-bearing and must be resolved.
  3. [Section 5.1 and all experimental tables] The experimental methodology is critically underspecified. Section 5.1 states only that 'we use 24861 samples for training and test' without giving the train/test split, writer independence, number of runs, or hyperparameter values; no error bars or confidence intervals are provided. Tables 4-9 compare the proposed systems with prior work without standardizing train/test protocols, database subsets, or evaluation measures, making the numerical comparisons unreliable.
  4. [Sections 3.1.1.2 and 3.1.1.3] The load-bearing assumption that 'we use only 9 basic perceptual shapes which can form any script' is asserted without proof, coverage analysis, or an error analysis of the segmentation stage. The EPC assignment also depends on a hand-chosen overlap constant 'cst equal to /16' and on hand-defined direction-region boundaries, with no sensitivity analysis. As written, the reported rates cannot be expected to transfer to new writing styles or scripts.
  5. [Tables 6, 8, and 9 and Section 5.2] The tables contain unresolved internal contradictions. Tables 6 and 9 both report IRONOFF results but give different OnHSR-LSTM rates (97.5% vs 93%) and appear to refer to different tasks (digits vs words). Section 5.2 says IRONOFF digit results are illustrated in Table 8, but Table 8 actually reports ADAB word recognition, while Section 5.3 refers to Table 8 for the mobile database when the mobile results are in Table 7. These inconsistencies make the reliability of the reported numbers impossible to assess.
  6. [Section 7, Table 10] The noise robustness experiment reports that OnHR-convLSTM with raw (x,y,z) input drops to 50.45% accuracy, which is near chance for a digit task, while the same architecture under other conditions reaches 96-98%. No noise level, noise generation procedure, number of trials, or train/test protocol is given, so the robustness claim is not substantiated and the 50.45% figure suggests a possible training or evaluation problem that needs explanation.
minor comments (6)
  1. [Section 3.1.1.3] The overlap constant is written as 'cst equal to /16'; this should be cst = π/16, and the use of π should be made explicit.
  2. [Throughout] Table numbering is inconsistent: the text refers to 'Table 2' for the comparison with Tagougui et al., while the actual comparison is in Table 4; several other table references (e.g., Section 5.2 citing Table 8 for IRONOFF) are mismatched and must be corrected.
  3. [Section 3.1.1.2, Equation (1)] The Beta function in Equation (1) contains likely typographical errors in the exponents and in the definition of tc; please verify the equation against the standard Beta-elliptic model references.
  4. [Section 5.1] The statement 'We used 54 shapes instead of 56 Shapes' is unexplained; if two LMCA letter shapes were excluded from training or testing, the authors must state which shapes and why.
  5. [Equations (3)-(7)] The notation uses p for both a Beta parameter and the number of strokes (e.g., Equations (5)-(6)), which is ambiguous; please use distinct symbols.
  6. [Figures 15 and 16] Figure 16 is referenced twice with different content (noise sample and perceptual illusions); the figures need to be renumbered coherently.

Circularity Check

1 steps flagged · score 6.0 of 10

Fuzzy ground-truth membership is computed by Euclidean distance on the same stroke-length series fed to the LSTM; the paper's headline 98.5% result is thus partly a prediction of a target that is a function of the input.

  1. fitted input called prediction [Section 4.3.2 (Algorithm 2), evaluated in Table 4 (OnHSR-LSTM MAYASTROUN, 98.5%)]
    "To define the value of membership, we calculate the segmented data which has the longest series of elliptic-strokes{li}. Then, we calculate the Euclidian distance between the length of class of reference and the actual data segmented{lr} and{li}."

    The fuzzy ground-truth targets used to train OnHSR-LSTM are derived from the Euclidean distance between the sample's own stroke-length series {li} and the class reference series {lr}. The same {li} are the LSTM's input representation, so the training target is a deterministic function of the input plus a per-class prototype. The reported recognition rate with fuzzy ground truth (98.5% on MAYASTROUN, vs 94.5% framewise) measures the LSTM's ability to reproduce a distance-to-prototype mapping rather than to assign independent class labels. It is therefore a fitted target renamed as a prediction, and the abstract's 'recognition rate exceeds 98%' rests in part on this construction.

full rationale

The most load-bearing reported number, 98.5% on MAYASTROUN for OnHSR-LSTM, comes from the Fuzzy Ground Truth protocol: Algorithm 2 constructs the target membership from a Euclidean distance between the input stroke-length series and class-reference series. Because the target is computed from the same features used as input, high agreement with this target does not independently validate the perceptual-code recognition claim; the framewise baseline on the same database is 94.5%, and many reported rows in Tables 4-10 are below 98%. The second system (OnHR-convLSTM) achieves 98% on IRONOFF words without this fuzzy-target construction, so the paper is not wholly circular. The claim that '9 basic perceptual shapes can form any script' (Section 3.1.1.2) is an unsupported coverage assumption rather than a circular step, and the many self-citations to prior REGIM work define the feature vocabulary but are not invoked as a uniqueness theorem. Because one central prediction reduces, by the paper's own equations, to a target computed from its input, the partial circularity score is 6.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities. The ledger is dominated by domain assumptions about segmentation, the sufficiency of nine perceptual codes, and an ad hoc fuzzy ground truth target computed from the same features used for recognition.

free parameters (3)
  • Fuzzy overlap constant cst for EPC indecision intervals = pi/16 (text renders it as '/16')
    Hand-chosen interval that defines when a stroke belongs to more than one Elementary Perceptual Code; its value directly affects the fuzzy codes fed to the LSTM.
  • EPC direction region boundaries on the trigonometric circle = not stated
    Angular sectors mapping deviation angle theta to the four EPCs are defined geometrically but not justified, and no sensitivity analysis is given.
  • LSTM and convLSTM hyperparameters = not reported
    Number of layers, number of units, dropout rate, learning rate, and optimizer are not stated, so the reported results cannot be tied to a specific architecture.
assumptions (4)
  • domain assumption Nine basic perceptual codes suffice to compose any Arabic or Latin script.
    Section 3.1.1.2 asserts "we use only 9 basic perceptual shapes which can form any script" without a coverage analysis over the tested alphabets.
  • domain assumption Beta-elliptic segmentation identifies the same strokes a human perceiver would.
    Segmentation quality is never measured and is the input to all downstream perceptual codes; failure here would invalidate the reported pipeline results.
  • ad hoc to paper Fuzzy membership derived from stroke-length Euclidean distance forms a valid training target.
    Algorithm 2 defines memberships using distances between reference and actual stroke lengths without validating that these soft labels correspond to human class judgments.
  • standard math LSTM, CTC, and convLSTM work as described in the cited literature.
    Standard sequence learning machinery is assumed from references; the paper does not re-derive it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neural Architecture based on Fuzzy Perceptual Representation For Online Multilingual Handwriting Recognition." pith.science (2026). https://pith.science/paper/OUVTO2X6

@misc{pith2026190800634,
  author       = {Pith},
  title        = {Pith review of: Neural Architecture based on Fuzzy Perceptual Representation For Online Multilingual Handwriting Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OUVTO2X6}},
  note         = {Machine review of arXiv:1908.00634}
}
read the original abstract

Due to the omnipresence of mobile devices, online handwritten scripts have become the most important feeding input to smartphones and tablet devices. To increase online handwriting recognition performance, deeper neural networks have extensively been used. In this context, our paper handles the problem of online handwritten script recognition based on extraction features system and deep approach system for sequences classification. Many solutions have appeared in order to facilitate the recognition of handwriting. Accordingly, we used an existent method and combined with new classifiers in order to get a flexible system. Good results are achieved compared to online characters and words recognition system on Latin and Arabic scripts. The performance of our two proposed systems is assessed by using five databases. Indeed, the recognition rate exceeds 98%.

Figures

Figures reproduced from arXiv: 1908.00634 by the authors.

Figure 1
Figure 1. Architecture OnHS-LSTM. TABLE 2 Basic Perceptual Codes Number Basic Perceptual Codes Abbreviation Shape 1 Valley V 2 Left-oblique-shaft L-O-S 3 Shaft S 4 Right-oblique-shaft R-O-S 5 Right-Half-Occlusion R-H-O 6 Left-Half-Occlusion L-H-O 7 Up-Half-Occlusion U-H-O 8 Down-Half-Occlusion D-H-O 9 Occlusion Occ 3.1 Online Handwriting Segmentation & Recognition using LSTM, OnHSR-LSTM Our first model OnHSR-LSTM, is based on… view at source ↗
Figure 2
Figure 2. Illustrated example of composed forms. order to obtain better recognition rate. The principles of pre￾processing steps are used to eliminate trembles in writing , to reduce noise and to remove the hardware imperfections. The preprocessing operations used in our system are: • Interpolation: adding missing points caused by the variation of writing velocity. • De-hooking: eliminating trembles in writing due to inaccura… view at source ↗
Figure 3
Figure 3. Segmentation Step of Arabic Letter ”yaa”. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: Segmentation and Detection of Beta-points of Arabic word ”Kalaa” . [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Directions: (a): Freeman0s chain code, (b): the proposed regions, (c):Deviation angles regions and EPCs on the trigonometric cir￾cle,(d):Elementary Perceptual Codes [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 7
Figure 7. Figure 7: Architecture of LSTM cells. to the 4 EPCs. A segment is a group of strokes, and stroke is composed of 4 EPCs with degree of membership. Each stroke is classified into 4 EPCs, and a segment is a group of n strokes. A BPC is a group of n segments or a script(character, d…
Figure 6
Figure 6. Figure 6: Example of perceptual problem decision using fuzzy logic. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 8
Figure 8. Figure 8: Some samples of BPC’s combination. performance of our network, we use the CTC as output layer after softmax layer in system 1 which pertains to Perceptual Codes-LSTM for Online handwriting recognition. In fact, it is a layer of labelling. The CTC algorithm employs a ma…
Figure 9
Figure 9. Figure 9: (a)Architecture of OnHR-convLSTM, (b) convLSTM cell. [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Data Augmentation of Arabic letter ”haa” [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 11
Figure 11. Figure 11: Distortion of digit 7. 4.3 Training Network To improve the performance of two our proposed systems, various techniques are used that we will illustrate below. 4.3.1 Framewise/Crisper Training We opt this technique of training with two proposed sys￾tems OnHRS-LSTM and …
Figure 12
Figure 12. Figure 12: Arabic Alphabets and their forms at different positions. [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]
Figure 13
Figure 13. Figure 13: Classes Predicted and Correct ones. 5.3 Our Mobile Database Based on mobile application of [7], we created a mobile database for acquired data from different persons with dif￾ferent styles of writing in order to examine the performance of our system. The (x, y, z) coo…
Figure 14
Figure 14. Figure 14: Interfaces of Mobile Database. TABLE 10 Recognition Rate with LMCA and manually segmented from ADAB Database with noise . System Methods Recognition Rate OnHSR-LSTM BCP+LSTM 96.00 % OnHR-convLSTM (x,y,z)+ConvLSTM 50.45% OnHR-convLSTM (x,y,z)+ConvLSTM +teta+4 EPCs 74.4…
Figure 15
Figure 15. Figure 15: Sample of input noise & output [PITH_FULL_IMAGE:figures/full_fig_p012_15.png]
Figure 16
Figure 16. Figure 16: Sample of input noise & output word dhahab I. ë [PITH_FULL_IMAGE:figures/full_fig_p012_16.png]
Figure 17
Figure 17. Figure 17: Handwriting perceptual illusions : (a): the word he or be, (b): the [PITH_FULL_IMAGE:figures/full_fig_p013_17.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 39 canonical work pages

  1. [1]

    Recognizing online arabic handwritten characters using a deep architecture,

    N. Tagougui and M. Kherallah, “Recognizing online arabic handwritten characters using a deep architecture,” Proc.SPIE, vol. 10341, pp. 1– 5, 2017. [Online]. Available: https://doi.org/10. 1117/12.2268419

  2. [2]

    A Hybrid NN/HMM Modeling Technique for Online Arabic Handwriting Recognition

    N. Tagougui, H. Boubaker, M. Kherallah, and A. M. Alimi, “A hybrid NN/HMM modeling technique for online arabic handwriting recognition,” CoRR, vol. abs/1401.0486, 2014. [Online]. Available: http://arxiv.org/abs/1401.0486

  3. [3]

    Multi-language online handwriting recognition,

    D. Keysers, T. Deselaers, H. A. Rowley, L. L. Wang, and V . Car- bune, “Multi-language online handwriting recognition,” IEEE T ransactions on Pattern Analysis and Machine Intelligence , vol. 39, no. 6, pp. 1180–1194, June 2017

  4. [4]

    Online farsi handwritten character recognition using hidden markov model,

    V . Ghods and M. K. Sohrabi, “Online farsi handwritten character recognition using hidden markov model,” JCP, vol. 11, pp. 169– 175, 2016

  5. [5]

    Off- lexicon online arabic handwriting recognition using neural net- work,

    Y. Hamdi, A. Chaabouni, B. Houcine, and A. M. Alimi, “Off- lexicon online arabic handwriting recognition using neural net- work,” pp. 1–5, 2016

  6. [6]

    Improved deep convolutional neural network for online handwritten chinese character recogni- tion using domain-specific knowledge,

    W. Yang, L. Jin, Z. Xie, and Z. Feng, “Improved deep convolutional neural network for online handwritten chinese character recogni- tion using domain-specific knowledge,” pp. 551–555, Aug 2015

  7. [7]

    A gru-based encoder-decoder ap- proach with attention for online handwritten mathematical ex- pression recognition,

    J. Zhang, J. Du, and L. Dai, “A gru-based encoder-decoder ap- proach with attention for online handwritten mathematical ex- pression recognition,” 12 2017

  8. [8]

    Design of a very compact cnn classifier for online handwritten chinese character recognition using dropweight and global pooling,

    X. Xiao, Y. Yang, T. Ahmad, L. Jin, and T. Chang, “Design of a very compact cnn classifier for online handwritten chinese character recognition using dropweight and global pooling,” 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR), vol. 01, pp. 891–895, 2017

Show all 40 references
  1. [9]

    Drawing and recognizing chinese characters with recurrent neural network,

    X. Zhang, F. Yin, Y. Zhang, C. Liu, and Y. Bengio, “Drawing and recognizing chinese characters with recurrent neural network,” IEEE T ransactions on Pattern Analysis and Machine Intelligence , vol. 40, no. 4, pp. 849–862, April 2018

  2. [10]

    Real time handwriting recognition for mathematic expressions using hidden markov model,

    Y. M.Pranoto, E. Setyati, E. Pramana, Y. Kristian, and R. Budiman, “Real time handwriting recognition for mathematic expressions using hidden markov model,” pp. 1–6, 07 2016

  3. [11]

    Fast multi-language lstm-based online handwriting recognition,

    V . Carbune, P . Gonnet, T. Deselaers, H. A. Rowley, A. N. Daryin, M. Calvo, L. Wang, D. Keysers, S. Feuz, and P . Gervais, “Fast multi-language lstm-based online handwriting recognition,” CoRR, vol. abs/1902.10525, 2019. [Online]. Available: http://arxiv.org/abs/1902.10525

  4. [12]

    Recognition of online handwriting with variability on smart devices,

    G. Ren and V . Ganapathy, “Recognition of online handwriting with variability on smart devices,” pp. 7605–7609, May 2019

  5. [13]

    Online arabic handwriting recognition: A survey,

    N. Tagougui, M. Kherallah, and A. M. Alimi, “Online arabic handwriting recognition: A survey,” Int. J. Doc. Anal. Recognit. , vol. 16, no. 3, pp. 209–226, Sep. 2013. [Online]. Available: http://dx.doi.org/10.1007/s10032-012-0186-8

  6. [14]

    The state of the art in online handwriting recognition,

    C. C. Tappert, C. Y. Suen, , and T. Wakahara, “The state of the art in online handwriting recognition,” IEEE T rans. Pattern Anal. Mach. Intell., vol. 12, p. 787808, August 1990

  7. [15]

    Online hand- writing recognition: The npen++ recognizer,

    S. Jaeger, S. Manke, J. Reichert, and A. Waibel, “Online hand- writing recognition: The npen++ recognizer,” Int. J. Doc. Anal. Recognit., vol. 3, p. 169180, 2001

  8. [16]

    Novel hybrid nn/hmm modelling tech- niques for on-line handwriting recognition,

    J. Schenk and G. Rigoll, “Novel hybrid nn/hmm modelling tech- niques for on-line handwriting recognition,” in Proc. Int. Workshop Frontiers Handwriting Recognit. , p. 619623, 2006

  9. [17]

    A novel connectionist system for unconstrained handwriting recognition,

    A. Graves, M. Liwicki, S. Fernandez, R. Bertolami, and H. B. andJ. Schmidhuber, “A novel connectionist system for unconstrained handwriting recognition,” IEEE T rans. Pattern Anal. Mach. Intell. , vol. 31, p. 855868, May 2009

  10. [18]

    A new encoding system: Application to on-line arabic handwriting,

    S. Njah, H. Bezine, and A. M. Alimi, “A new encoding system: Application to on-line arabic handwriting,” pp. 451–456, 2010

  11. [19]

    The pertohs theory for on-line handwriting segmentation,

    S. Njah, M. Ltaief, H. Bezine, and A. M. Alimi, “The pertohs theory for on-line handwriting segmentation,” IJCSI, vol. 9, pp. 142–151, 2012

  12. [20]

    On-line arabic handwriting segmentation via perceptual codes: application to mayastroun database,

    S. Njah, H. Bezine, and A. M. Alimi, “On-line arabic handwriting segmentation via perceptual codes: application to mayastroun database,” pp. 1–5, March 2011

  13. [21]

    Reconnaissance hors-ligne de lcriture cursive par lutilisation de modles perceptifs et neuronaux,

    J.R.Pinales, “Reconnaissance hors-ligne de lcriture cursive par lutilisation de modles perceptifs et neuronaux,” PhD, University of Paris, 2002

  14. [22]

    A fuzzy online hand- writing recognition system: Fohres,

    A.Malaviya, L.Peters, and P .Camposano, “A fuzzy online hand- writing recognition system: Fohres,” Proceedings of the International Conference on fuzzy theory and technology , pp. 1–15, 1993

  15. [23]

    Decision trees based on perceptual codes for on-line arabic char- acter recognition,

    H. Akouaydi, S. Abdelhedi, S. Njah, M. Zaied, and A. M. Alimi, “Decision trees based on perceptual codes for on-line arabic char- acter recognition,” no. 1-5, pp. 153–157, April 2017

  16. [24]

    Genetic algorithms for perceptual codes extraction,

    M. Ltaief, S. Njah, H. Bezine, and A. M. Alimi, “Genetic algorithms for perceptual codes extraction,” International Journal of Intelligent Learning Systems and Applications, JILSA , pp. 256–265, 2012

  17. [25]

    A new approach for the extraction of handwriting perceptual codes using fuzzy logic,

    S. Njah, H. Bezine, and A. M. Alimi, “A new approach for the extraction of handwriting perceptual codes using fuzzy logic,” pp. 302–307, 2008

  18. [26]

    A developmental theory of the relation between geometry and kinematics in handwriting,

    H.Bezine, A.M.Alimi, and N.Derbel, “A developmental theory of the relation between geometry and kinematics in handwriting,” IEEE International Conference Systems, Man and Cybernetics , p. 5, 2002

  19. [27]

    A spiking neural network model for complex handwriting movements generation,

    M.Ltaief, H.Bezine, and A.M.Alimi, “A spiking neural network model for complex handwriting movements generation,” Interna- JOURNAL OF LATEX CLASS FILES, VOL. , NO. , MAY 2019 14 tional Journal of Computer Science and Information Security 14 , p. 7, 2016

  20. [28]

    A fuzzy genetic system for segmentation of on-line handwriting: Application to adab database,

    S. Njah, H. Bezine, and A. M. Alimi, “A fuzzy genetic system for segmentation of on-line handwriting: Application to adab database,” pp. 95–102, April 2011

  21. [29]

    Online arabic writer identification based on beta-elliptic model,

    T.Dhieb, W.Ouarda, H.Boubaker, M.B.Hlima, and A. M. Alimi, “Online arabic writer identification based on beta-elliptic model,” pp. 74–79, 2015

  22. [30]

    Beta-elliptic model for writer identification from online arabic handwriting,

    T.Dhieb, W.Ouarda, H.Boubaker, and A. M. Alimi, “Beta-elliptic model for writer identification from online arabic handwriting,” vol. 11, p. 6, 2016

  23. [31]

    An online writer identification system based on beta- elliptic model and fuzzy elementary perceptual codes,

    T.Dhieb, S.Njah, W.Ouarda, H.Boubaker, M.B.Ayed, and A. M. Alimi, “An online writer identification system based on beta- elliptic model and fuzzy elementary perceptual codes,” p. 12, 2018

  24. [32]

    Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks,

    G. Alex, F. Santiago, G. Faustino, and S. Jurgen, “Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks,” New York, NY, USA, pp. 369–376, 2006. [Online]. Available: http://doi.acm.org/10.1145/ 1143844.1143891

  25. [33]

    Framewise and ctc training of neural networks for handwriting recognition,

    T. Bluche, H. Ney, J. Louradour, and C. Kermorvant, “Framewise and ctc training of neural networks for handwriting recognition,” pp. 81–85, Aug 2015

  26. [34]

    Mayastroun: A multilanguage handwriting database,

    S. Njah, B. B. Nouma, H. Bezine, and A. M. Alimi, “Mayastroun: A multilanguage handwriting database,” pp. 308–312, Sept 2012

  27. [35]

    Recognition of on-line handwritten digits by neural networks using circular and beta approaches,

    M. Kherallah, S. Njah, A. M. Alimi, and N. Derbel, “Recognition of on-line handwritten digits by neural networks using circular and beta approaches,” vol. 2, pp. 164–169, Oct 2002

  28. [36]

    The on/off (lmca) dual arabic handwriting database,

    M.Kherallah, A.Elbaati, H.ElAbed, and A.M.Alimi, “The on/off (lmca) dual arabic handwriting database,” 2008

  29. [37]

    Unipen project of on-line data exchange and recognizer bench- marks,

    I. Guyon, L. Schomaker, R. Plamondon, M. Liberman, and S. Janet, “Unipen project of on-line data exchange and recognizer bench- marks,” vol. 2, pp. 29–33 vol.2, Oct 1994

  30. [38]

    The ireste on/off (ironoff) dual handwriting database,

    V . Gaudin, P . Lallican, S. Knerr, and P . Binter, “The ireste on/off (ironoff) dual handwriting database,” p. 455 458, 1999

  31. [39]

    Android application for handwriting segmentation using pertohs theory,

    H.Akouaydi, S.Njah, and A. M. Alimi, “Android application for handwriting segmentation using pertohs theory,” Proc.SPIE, vol. 10341, pp. 10 341 – 10 341 – 5, 2016. [Online]. Available: https://doi.org/10.1117/12.2269059

  32. [40]

    On-line arabic handwriting recognition competition, adab database and participating systems,

    H. E. Abed, M. Kherallah, V . Mrgner, and A. M. Alimi, “On-line arabic handwriting recognition competition, adab database and participating systems,” International Journal on Document Analysis and Recognition, vol. 14, pp. 15–23, 2011. Hanen Akouaydi currently a PhD student in...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.