Pith. sign in

REVIEW 5 major objections 7 minor 63 references

Finger in Camera Speaks Everything: Unconstrained Air-Writing for Real-World

T0 review · 5 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that air-writing Chinese characters can be recognized from ordinary RGB camera video using a two-stage fingertip-trajectory model, and supports the claim with a new 102,688-video benchmark covering all 3,755 GB1…

desk verdict Genuine dataset, but VCRec's stroke-identity feature is undefined and may give privileged information; the central accuracy claim needs a fix before publication. read the letter →

arxiv 2412.19537 v1 pith:AAAE53JU submitted 2024-12-27 cs.CV cs.HC

Meiqi Wu , Kaiqi Huang , Yuanqiang Cai , Shiyu Hu , Yuzhong Zhao , Weiqiang Wang This is my paper · ORCID
classification cs.CVcs.HC
keywords air-writingRGBvideorecognitionChinesecharacterfingertiptrajectorybenchmarkdatasetspatio-temporalmodelinggraphattentionnetwork
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

This paper tries to establish that air-writing Chinese characters can be recognized from ordinary RGB camera video, without radar, EEG, depth sensors, or wearable motion sensors. To show this, it introduces AWCV-100K-UCAS2024, a dataset of 102,688 videos and 8.8 million frames covering all 3,755 GB2312-80 level-1 Chinese characters, captured by 211 participants under varied lighting and backgrounds. It also proposes VCRec, a two-stage recognizer that first converts sparse video frames into fingertip trajectories, then models those trajectories with temporal convolution plus a stroke-based graph attention network. On this benchmark VCRec reaches 52.43% accuracy, which is 29.03 percentage points above the best one-stage video model tested. If these results hold, the field gains a large public RGB-only benchmark and evidence that fingertip trajectory features, rather than raw video, are the right representation for this task.

What carries the argument

The load-bearing mechanism is the Fingertip Feature Extractor plus the Spatio-Temporal Sequence Module. The extractor compresses sparse visual frames into trajectory point features, each represented as an eight-dimensional vector of position offsets, writing direction ($\sin\alpha,\cos\alpha$), curvature ($\sin\beta,\cos\beta$), and two stroke-boundary indicators; this is what lets 30 FPS RGB video behave like a trajectory signal. The temporal encoder is a hierarchical 1D CNN with residual blocks that captures finger movement over time, and the spatial encoder, StrokeGAT, is a graph attention network whose nodes are temporal clip features, modeling the stroke structure of the Chinese character. Their features are added and decoded by a two-layer fully connected head (with a CTC head used for English word sequences).

What would settle it

Run the fingertip tracker on a random sample of AWCV-100K-UCAS2024 videos and measure the proportion of frames where the detected keypoint is missing, jumps discontinuously, or visibly does not follow the writing fingertip; if that proportion is large, or if retraining VCRec on hand-verified trajectories instead of tracker outputs changes the reported 52.43% accuracy substantially, the two-stage advantage is an artifact of the tracker rather than evidence about air-writing recognition.

Watch

Extended reading notes

Core claim

The central claim is that a two-stage architecture built on sparse fingertip features solves video-based air-writing recognition far better than one-stage video encoders, and that this can be demonstrated at scale on a new benchmark covering the complete GB1 Chinese character set. The paper's VCRec pipeline extracts an eight-dimensional per-frame fingertip representation (XY offsets, writing direction and curvature sine/cosine terms, and stroke-identity change indicators) using a pretrained fingertip tracker, then passes it through a hierarchical 1D convolutional temporal encoder and a graph attention encoder called StrokeGAT that treats trajectory clips as nodes of a character-structure graph. The decoder fuses the two feature streams and outputs character probabilities. VCRec reports 52.43% accurate rate on the AWCV-100K-UCAS2024 test split, a 29.03-point improvement over the previous video-based state of the art (ST-R3D), and a 4.92-point improvement over the same two-stage pipeline without the spatial encoder, which the paper reads as evidence that modeling the spatial structure of logograms matters.

Load-bearing premise

The whole two-stage claim depends on the pretrained fingertip tracker producing accurate fingertip keypoints on essentially every frame of the 30 fps RGB videos under the dataset's lighting, motion blur, and cluttered backgrounds, since the paper reports no tracker failure rate and no rejection of frames where tracking fails.

Editorial extensions

If this is right

  • If the benchmark stands, RGB-only air-writing becomes a testable public task with a corpus covering 99.7% of daily-used Chinese characters, removing the sensor barrier for phones, laptops, and smart TVs.
  • The reported 29.03-point gap over ST-R3D implies that for this task, fingertip trajectory features are far more informative than learned spatiotemporal video features at 30 FPS.
  • The 4.92-point gain from adding StrokeGAT indicates Chinese character structure contributes signal beyond raw temporal dynamics, supporting further work on structure-aware recognizers.
  • VCRec's comparable performance on trajectory datasets (IAHCC-UCAS2016, IAHEW-UCAS2016) and on the English video dataset WiTA suggests the two-stage recipe transfers across languages and input modalities.

Reading between the lines

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

  • The paper leaves unmeasured how much of its accuracy comes from the tracker; a useful next experiment is ablating the tracker (e.g., using ground-truth or multi-tracker trajectories) and reporting per-frame tracking failure, which would tell whether the benchmark's difficulty is in tracking or in recognition.
  • The eight-dimensional local trajectory representation (offsets, direction, curvature, stroke boundaries) is task-agnostic and could be applied to other sparse RGB-input tasks such as gesture commands, sign-language subword recognition, or online handwritten text, independent of the Chinese character decoder.
  • Because the dataset splits by person, it tests writer generalization; an unstated extension is measuring per-character confusability and stroke-order sensitivity, which would identify which GB1 characters are intrinsically hard to write in the air.
  • If the public release matches the described splits and environment metadata, follow-up work could train one-stage video models on this larger corpus and directly test whether the two-stage advantage persists with more data and higher frame rates.
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

5 major / 7 minor

Summary. The paper introduces AWCV-100K-UCAS2024, a video-based air-writing dataset of Chinese characters collected with ordinary RGB cameras, covering all 3,755 GB2312-80 level-1 characters through 102,688 videos and 8.8 million frames. The authors argue this is the first logogram-based video air-writing benchmark with a comprehensive corpus and diverse real-world conditions. They also propose VCRec, a two-stage recognizer that first extracts fingertip trajectories with MediaPipe and then encodes them with a temporal CNN plus a graph-attention spatial encoder (StrokeGAT), reporting 52.43% recognition accuracy on the new dataset. The paper includes ablations of the temporal encoder, spatial structure model, decoder, and fusion strategies, as well as experiments on existing trajectory- and video-based datasets in Chinese and English.

Significance. If the dataset is released as claimed, it would be a valuable new resource for video-based air-writing research: it is the largest RGB-only video dataset for Chinese air-writing, covers a complete character set, and includes substantial participant and environment diversity. The paper also provides a reproducible baseline (VCRec) and a clear evaluation protocol, which is useful for future comparisons. The algorithmic novelty of VCRec is modest — it combines MediaPipe, a temporal CNN, and graph attention — but the dataset contribution alone is significant for the community. However, the performance claims, especially the reported superiority over prior methods, are not fully supported by the current experiments and notation, and the role of stroke-identity information in VCRec must be clarified before the results can be trusted.

major comments (5)
  1. [IV-B, Eq. (3)] The fingertip feature vector in Eq. (3) includes the two terms I(st = st+1) and I(st ≠ st+1), where st is a stroke identity. The paper never states how st is obtained from the RGB video. Section IV-B describes only that the Fingertip Tracker (MediaPipe) yields fingertip trajectories; MediaPipe provides 21 hand keypoints and does not segment strokes. If st comes from ground-truth stroke annotations (as the stroke-labeled examples in Figs. 3 and 4 suggest), then VCRec uses privileged information at training and, more importantly, at inference time a deployed system cannot compute Eq. (3). The comparison against one-stage models would then be unfair, and the reported 52.43% accuracy would not represent a genuine video-based recognition result. If st is produced by an undocumented heuristic, the method is not reproducible. The authors must specify the source of st, remove the stroke-identity terms, or demonstrate a video-derived estimate of st, and re-run the experiments accordingly.
  2. [Abstract and Section V-C] The abstract and introduction claim a '4.92% accuracy improvement' over existing methods, but the 4.92% figure in Table IV is the gain of VCRec over 1DCNN, a temporal-only ablation variant of the authors' own method, not over an existing method from the literature. The actual comparison against the prior video SOTA, ST-R3D, shows a 29.03% gap (Table III), but that comparison depends on the one-stage baselines being properly trained. The paper should state unambiguously what 4.92% is relative to and avoid the implication that it is an improvement over prior published methods.
  3. [V-C, Table III] The one-stage baselines in Table III (CNN+LSTM, TwoStream, C3D, ST-MC, ST-rMC, ST-R(2+1)D, ST-R3D, ViT) are reported with no training details on AWCV-100K-UCAS2024. The paper does not specify whether these models were trained from scratch on the same train split, with the same image size, optimizer, epochs, or any hyperparameter tuning. Given the very large accuracy gap between VCRec (52.43%) and the best one-stage method (ST-R3D at 23.40%), it is plausible that the baselines were undertrained or used suboptimal settings, making the claimed 'significant outperformance' over one-stage approaches not established. The authors should provide the full training protocol for each baseline and, ideally, multiple trials with variance.
  4. [V-E, Tables X and XI] The generalization results are reported as 'comparable' but are actually numerically worse than prior methods. On WiTA (Table X), VCRec+CTC has CER 30.12%, which is higher (worse) than ST-R3D (29.24%) and TR-AWR (29.86%). On IAHEW-UCAS2016 (Table XI), VCRec's CAR of 96.51% is below 1DCNN+Decoder (97.45%) and LSTM+CTC (97.13%). These negative results are acknowledged only implicitly through the word 'comparable'; the paper should discuss them directly, as they weaken the claim that VCRec generalizes across languages and forms better than existing methods.
  5. [III-C and V-B] The entire two-stage pipeline assumes MediaPipe provides accurate fingertip trajectories on every frame, including the motion-blur and overexposure cases highlighted in Fig. 3 and Section III-C. The paper reports no tracking failure rate, no rejection of low-confidence frames, and no analysis of how tracker errors propagate to the final recognition accuracy. Without such an analysis, the 'robustness in real-world scenarios' claim is not supported. At minimum, the authors should report the fraction of frames or videos where MediaPipe fails and show either a fallback strategy or the recognition accuracy on the successfully tracked subset.
minor comments (7)
  1. [V-C] The text says 'Compared with ST-R3D [35]' but Table III lists ST-R3D with reference [17]; reference [35] is TR-AWR in Table X. The citation should be corrected.
  2. [IV-C, Eq. (4)] Equation (4) is written as p = fc(fg(fr(x)), fr(x)), but the following paragraph defines p = fc(Z ⊕ Z) where ⊕ denotes the add operator. This is ambiguous: fc appears to take two arguments in Eq. (4) and one fused argument in the text. Please clarify the intended formulation.
  3. [III-E] The evaluation protocol defines CR and AR using sentence-level errors (De, Se, Ie) but then states that the sequence length is set to 1. It should be clarified how CR is computed when each test item is a single character rather than a sentence, and whether the metrics reduce to simple classification accuracy.
  4. [Table II] The WiTA row lists Illumination as '-', but the text of Section V-A says WiTA data were captured from an RGB camera; the Illumination column should indicate whether illumination information is available or just not reported.
  5. [IV-B, Fig. 7] In Fig. 7, the trajectory points (pt, qt) are shown with subscripts t-2 to t+2, but the definitions of α and β (writing direction and curvature) are not given explicitly in the text. Please define these quantities or refer to a standard formulation.
  6. [VI] The conclusion says 'The dataset, toolkit, and experimental results will be released', while the abstract says 'code and data examples will be made public'. Please clarify the exact release plan, including the license and whether the full dataset or only examples will be available.
  7. [III-A] The text states 'Each participant has then composed approximately 500 words in Chinese'. Since the task writes characters, 'words' should likely be 'characters'; please correct the wording.

Circularity Check

1 steps flagged · score 5.0 of 10

VCRec's headline 52.43%-vs-23.40% result is partly self-definitional: the stroke-identity terms I(st=st+1), I(st≠st+1) in Eq. (3) have no video-derived source, and the only exhibited source is the dataset's ground-truth stroke annotation; the dataset and cross-dataset comparisons are otherwise non-circular.

  1. self definitional [Section IV-B (Fingertip Feature Extractor), Eq. (3); loaded in Section V-C, Table III]
    "the video is first inputted by the Fingertip Tracker [40], [41] to obtain fingertip trajectories and encoded into fingertip features by Fingertip Representation... the following representations are calculated for the t-th point (pt, qt, st) of the trajectory: ... (4) the change of the stroke identity. As a result, each point (pt, qt, st) is represented as an eight-dimensional vector xt at time step t, t∈ Z, i.e., xt = [∆pt, ∆qt, sin α, cos α, sin β, cos β, I (st = st+1), I (st ≠ st+1)]."

    Eq. (3) feeds two stroke-identity indicators into the recognizer, but the described chain — 'video → Fingertip Tracker (MediaPipe) → fingertip trajectories → Fingertip Representation → features' — has no stroke-segmentation stage. The dataset labels do contain strokes (Fig. 3 'Stroke: 2'; Fig. 4B 'the first two frames of the video correspond to the first stroke'), so ground-truth stroke annotation is the only stated source of st. The input is thus partly defined in terms of the target character's own stroke segmentation; the headline comparison (Table III: VCRec 52.43% vs ST-R3D 23.40%) credits VCRec with label-derived stroke-boundary information unavailable to the one-stage models. The 29.03-point gap is not a clean video-to-character result, and at deployment Eq.

full rationale

The dataset contribution is non-circular: AWCV-100K-UCAS2024 is a new collection with its own protocol (Sec. III), and its statistics are reported, not derived from a fitted model. Cross-dataset evaluations (WiTA, IAHCC-UCAS2016, IAHEW-UCAS2016) use external published benchmarks; despite several trajectory datasets and the 1DCNN line coming from the same UCAS group ([12], [14], [29], [30], [58]), these are used as comparisons, and the architecture is fully specified (Fig. 8), so the self-citations are credit/benchmarks, not load-bearing. 'StrokeGAT' adapts the externally cited GAT [52]; it is not a renamed known result. The one load-bearing gap is Eq. (3): two of eight input dimensions need a stroke identity st that no described component computes from RGB video, and the only source the paper exhibits is the dataset's ground-truth stroke annotation (Fig. 3, Fig. 4B). This partly self-defines the central comparison: the two-stage model receives label-derived stroke boundaries while one-stage baselines do not, so the 29.03-point superiority in Table III is not an unprivileged video-recognition result; notably, on the external video benchmark WiTA (Table X) VCRec+CTC (30.12 CER) does not beat ST-R3D (29.24). Score 5: the benchmark and external evaluations are independent, but the headline claim reduces in part to a label-derived input feature.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The paper introduces no fitted physical constants. The main supporting assumptions are about tracker reliability, stroke annotation, corpus coverage, and split fairness. The stroke-identity feature in Eq. (3) is the largest unexamined burden because the data collection section never explains how stroke boundaries are obtained.

free parameters (1)
  • VCRec hyperparameter set (dropout, attention heads, learning rate, batch size) = dropout 0.2, heads 8, GAT layers 1, LR 0.001, batch 8
    Chosen by hand in Section V-B; the reported 52.43% accuracy is conditional on these settings, but they are conventional and not presented as fitted physical quantities.
assumptions (4)
  • domain assumption Fingertip Tracker (MediaPipe) yields accurate 21-point hand keypoints on 30 fps general RGB frames under the dataset's illumination and clutter conditions.
    Invoked in Section IV-B and Fig. 7; no failure-rate or quality validation is reported on the new dataset.
  • ad hoc to paper Each trajectory point carries a stroke identity st used in the fingertip feature vector.
    Equation (3) in Section IV-B uses I(st=st+1) and I(st!=st+1), but the data collection section never describes how stroke boundaries are segmented or annotated.
  • domain assumption The GB2312-80 level-1 set covers 99.7% of daily-used Chinese characters, as reported by Liu et al. [18].
    Used in Section III-C to justify the corpus comprehensiveness claim; the paper does not independently recompute this coverage.
  • domain assumption A person-disjoint 8:1:1 train/validation/test split provides a fair generalization test.
    Stated in Section V-A; no leakage analysis, repeated splits, or confidence intervals are provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Finger in Camera Speaks Everything: Unconstrained Air-Writing for Real-World." pith.science (2026). https://pith.science/paper/AAAE53JU

@misc{pith2026241219537,
  author       = {Pith},
  title        = {Pith review of: Finger in Camera Speaks Everything: Unconstrained Air-Writing for Real-World},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AAAE53JU}},
  note         = {Machine review of arXiv:2412.19537}
}
read the original abstract

Air-writing is a challenging task that combines the fields of computer vision and natural language processing, offering an intuitive and natural approach for human-computer interaction. However, current air-writing solutions face two primary challenges: (1) their dependency on complex sensors (e.g., Radar, EEGs and others) for capturing precise handwritten trajectories, and (2) the absence of a video-based air-writing dataset that covers a comprehensive vocabulary range. These limitations impede their practicality in various real-world scenarios, including the use on devices like iPhones and laptops. To tackle these challenges, we present the groundbreaking air-writing Chinese character video dataset (AWCV-100K-UCAS2024), serving as a pioneering benchmark for video-based air-writing. This dataset captures handwritten trajectories in various real-world scenarios using commonly accessible RGB cameras, eliminating the need for complex sensors. AWCV-100K-UCAS2024 includes 8.8 million video frames, encompassing the complete set of 3,755 characters from the GB2312-80 level-1 set (GB1). Furthermore, we introduce our baseline approach, the video-based character recognizer (VCRec). VCRec adeptly extracts fingertip features from sparse visual cues and employs a spatio-temporal sequence module for analysis. Experimental results showcase the superior performance of VCRec compared to existing models in recognizing air-written characters, both quantitatively and qualitatively. This breakthrough paves the way for enhanced human-computer interaction in real-world contexts. Moreover, our approach leverages affordable RGB cameras, enabling its applicability in a diverse range of scenarios. The code and data examples will be made public at https://github.com/wmeiqi/AWCV.

Figures

Figures reproduced from arXiv: 2412.19537 by the authors.

Figure 1
Figure 1. Comparing Our Work and Conventional Air-Writing [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison between AWCV-100K-UCAS2024 with [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Examples of AWCV-100K-UCAS2024. The figure shows a comparison of data under different lighting intensities and [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: More Complex and Comprehensive Corpus. The figure shows the characteristics of AWCV-100K-UCAS2024. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Statistical Analysis Environments in AWCV-100K-UCAS2024. (A) Depicts the diverse environmental distributions [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Overview of the Video-based Character Recognizer (VCRec). The VCRec comprises a Fingertip Feature Extractor, a [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Fingertip Feature Extractor. The video is first inputted [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 10
Figure 10. Figure 10: Architectures of Air-Writing Recognition Method. (A) [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 9
Figure 9. Figure 9: Spatial Feature Encoder. It models the spatial structure [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 11
Figure 11. Figure 11: Feature Fusion Strategy. (B), (C) and (D) are three different architectures of feature fusion strategies. (A) is the model [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 12
Figure 12. Figure 12: The Visualization of StrokeGAT. (A), (B) and (C) [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 52 canonical work pages

  1. [1]

    Radar-based air-writing gesture recognition using a novel multistream cnn approach,

    S. Ahmed, W. Kim, J. Park, and S. H. Cho, “Radar-based air-writing gesture recognition using a novel multistream cnn approach,” IEEE Internet of Things Journal , vol. 9, no. 23, pp. 23869–23880, 2022

    work page 2022

  2. [2]

    Nntrak: A neural network approach towards calculating air-writing trajectories in real-time with a smartwatch,

    V . Chandel and A. Ghose, “Nntrak: A neural network approach towards calculating air-writing trajectories in real-time with a smartwatch,” in 2023 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Work- shops), pp. 374–379, 2023

    work page 2023

  3. [3]

    A unified cnn-rnn approach for in- air handwritten english word recognition,

    J. Gan, W. Wang, and K. Lu, “A unified cnn-rnn approach for in- air handwritten english word recognition,” in 2018 IEEE International Conference on Multimedia and Expo (ICME) , pp. 1–6, IEEE, 2018

    work page 2018

  4. [4]

    Neuroair: Deep learning framework for airwriting recognition from scalp-recorded neural signals,

    A. Tripathi, A. Gupta, A. P. Prathosh, S. P. Muthukrishnan, and L. Ku- mar, “Neuroair: Deep learning framework for airwriting recognition from scalp-recorded neural signals,” 2023

    work page 2023

  5. [5]

    A wearable real-time character recognition system based on edge computing-enabled deep learning for air-writing,

    H. Zhang, L. Chen, Y . Zhang, R. Hu, C. He, Y . Tan, J. Zhang, et al. , “A wearable real-time character recognition system based on edge computing-enabled deep learning for air-writing,” Journal of Sensors , vol. 2022, 2022

    work page 2022

  6. [6]

    Language mod- els are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, et al., “Language mod- els are few-shot learners,” Advances in neural information processing systems, vol. 33, pp. 1877–1901, 2020

    1901

  7. [7]

    Glm: General language model pretraining with autoregressive blank infilling,

    Z. Du, Y . Qian, X. Liu, M. Ding, J. Qiu, Z. Yang, and J. Tang, “Glm: General language model pretraining with autoregressive blank infilling,” arXiv preprint arXiv:2103.10360 , 2021

    arXiv 2021

  8. [8]

    A new writing experience: Finger writing in the air using a kinect sensor,

    X. Zhang, Z. Ye, L. Jin, Z. Feng, and S. Xu, “A new writing experience: Finger writing in the air using a kinect sensor,” IEEE MultiMedia , vol. 20, no. 4, pp. 85–93, 2013

    work page 2013

Show all 63 references
  1. [9]

    Air-writing recognition—part i: Modeling and recognition of characters, words, and connecting motions,

    M. Chen, G. AlRegib, and B.-H. Juang, “Air-writing recognition—part i: Modeling and recognition of characters, words, and connecting motions,” IEEE Transactions on Human-Machine Systems, vol. 46, no. 3, pp. 403–413, 2015

    2015

  2. [10]

    A pointing gesture based egocentric interaction system: Dataset, approach and application,

    Y . Huang, X. Liu, X. Zhang, and L. Jin, “A pointing gesture based egocentric interaction system: Dataset, approach and application,” in Proceedings of the IEEE conference on computer vision and pattern recognition workshops, pp. 16–23, 2016

    2016

  3. [11]

    Air-writing with sparse network of radars using spatio-temporal learning,

    M. Arsalan, A. Santra, K. Bierzynski, and V . Issakov, “Air-writing with sparse network of radars using spatio-temporal learning,” in 2020 25th International Conference on Pattern Recognition (ICPR) , pp. 8877– 8884, IEEE, 2021

    2020

  4. [12]

    Recognition of in-air handwritten chinese character based on leap motion controller,

    N. Xu, W. Wang, and X. Qu, “Recognition of in-air handwritten chinese character based on leap motion controller,” in Image and Graphics: 8th International Conference, ICIG 2015, Tianjin, China, August 13– 16, 2015, Proceedings, Part III , pp. 160–168, Springer, 2015

    2015

  5. [13]

    Data augmentation and direc- tional feature maps extraction for in-air handwritten chinese character recognition based on convolutional neural network,

    X. Qu, W. Wang, K. Lu, and J. Zhou, “Data augmentation and direc- tional feature maps extraction for in-air handwritten chinese character recognition based on convolutional neural network,” Pattern recognition letters, vol. 111, pp. 9–15, 2018

    2018

  6. [14]

    In-air handwritten english word recognition using attention recurrent translator,

    J. Gan and W. Wang, “In-air handwritten english word recognition using attention recurrent translator,” Neural Computing and Applications , vol. 31, pp. 3155–3172, 2019

    2019

  7. [15]

    Compressing the cnn architecture for in-air handwritten chinese character recognition,

    J. Gan, W. Wang, and K. Lu, “Compressing the cnn architecture for in-air handwritten chinese character recognition,” Pattern Recognition Letters, vol. 129, pp. 190–197, 2020

    2020

  8. [16]

    Fin- gertip detection and tracking for recognition of air-writing in videos,

    S. Mukherjee, S. A. Ahmed, D. P. Dogra, S. Kar, and P. P. Roy, “Fin- gertip detection and tracking for recognition of air-writing in videos,” Expert Systems with Applications , vol. 136, pp. 217–229, 2019

    2019

  9. [17]

    Writing in the air: Unconstrained text recognition from finger movement using spatio- temporal convolution,

    U.-H. Kim, Y . Hwang, S.-K. Lee, and J.-H. Kim, “Writing in the air: Unconstrained text recognition from finger movement using spatio- temporal convolution,” IEEE Transactions on Artificial Intelligence , 2022

    2022

  10. [18]

    Word naming and psycholinguistic norms: Chinese,

    Y . Liu, H. Shu, and P. Li, “Word naming and psycholinguistic norms: Chinese,” Behavior research methods, vol. 39, no. 2, pp. 192–198, 2007

    2007

  11. [19]

    A novel vision- based finger-writing character recognition system,

    L. Jin, D. Yang, L.-X. Zhen, and J.-C. Huang, “A novel vision- based finger-writing character recognition system,” Journal of Circuits, Systems, and Computers , vol. 16, no. 03, pp. 421–436, 2007

    2007

  12. [20]

    Vision-based handwriting recognition for unrestricted text input in mid-air,

    A. Schick, D. Morlock, C. Amma, T. Schultz, and R. Stiefelhagen, “Vision-based handwriting recognition for unrestricted text input in mid-air,” in Proceedings of the 14th ACM international conference on Multimodal interaction, pp. 217–220, 2012

    2012

  13. [21]

    Writing in the air with wifi signals for virtual reality devices,

    Z. Fu, J. Xu, Z. Zhu, A. X. Liu, and X. Sun, “Writing in the air with wifi signals for virtual reality devices,” IEEE Transactions on Mobile Computing, vol. 18, no. 2, pp. 473–484, 2018

    2018

  14. [22]

    Glove-talk: A neural network interface between a data-glove and a speech synthesizer,

    S. S. Fels and G. E. Hinton, “Glove-talk: A neural network interface between a data-glove and a speech synthesizer,” IEEE transactions on Neural Networks, vol. 4, no. 1, pp. 2–8, 1993

    1993

  15. [23]

    Twostick: Writing with a game controller,

    T. K ¨oltringer, P. Isokoski, and T. Grechenig, “Twostick: Writing with a game controller,” in Proceedings of Graphics Interface 2007 , pp. 103– 110, 2007

    2007

  16. [24]

    Airwriting recognition using wearable motion sensors,

    C. Amma, D. Gehrig, and T. Schultz, “Airwriting recognition using wearable motion sensors,” in Proceedings of the 1st Augmented Human international Conference, pp. 1–8, 2010

    2010

  17. [25]

    C. Amma, M. Georgi, and T. Schultz, “Airwriting: Hands-free mobile text input by spotting and continuous recognition of 3d-space handwrit- IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY , VOL. X, NO. X, X X 12 ing with inertial sensors,” in 2012 16th Internatio...

    2012

  18. [26]

    Airwriting: a wearable handwrit- ing recognition system,

    C. Amma, M. Georgi, and T. Schultz, “Airwriting: a wearable handwrit- ing recognition system,” Personal and ubiquitous computing , vol. 18, pp. 191–203, 2014

    2014

  19. [27]

    Writing in the air with wifi signals for virtual reality devices,

    Z. Fu, J. Xu, Z. Zhu, A. X. Liu, and X. Sun, “Writing in the air with wifi signals for virtual reality devices,” IEEE Transactions on Mobile Computing, vol. 18, p. 473–484, feb 2019

    2019

  20. [28]

    Study of text segmentation and recognition using leap motion sensor,

    P. Kumar, R. Saini, P. P. Roy, and D. P. Dogra, “Study of text segmentation and recognition using leap motion sensor,” IEEE Sensors Journal, vol. 17, no. 5, pp. 1293–1301, 2016

    2016

  21. [29]

    In-air handwritten english word recognition using attention recurrent translator,

    J. Gan and W. Wang, “In-air handwritten english word recognition using attention recurrent translator,” Neural computing & applications , 2019

    2019

  22. [30]

    In-air handwritten chinese text recognition with temporal convolutional recurrent network,

    J. Gan, W. Wang, and K. Lu, “In-air handwritten chinese text recognition with temporal convolutional recurrent network,” Pattern Recognition , vol. 97, p. 107025, 2020

    2020

  23. [31]

    Microsoft kinect sensor and its effect,

    Z. Zhang, “Microsoft kinect sensor and its effect,” IEEE multimedia , vol. 19, no. 2, pp. 4–10, 2012

    2012

  24. [32]

    Characters as graphs: Interpretable handwritten chinese character recognition via pyramid graph transformer,

    J. Gan, Y . Chen, B. Hu, J. Leng, W. Wang, and X. Gao, “Characters as graphs: Interpretable handwritten chinese character recognition via pyramid graph transformer,” Pattern Recognition, vol. 137, p. 109317, 2023

    2023

  25. [33]

    In-air handwritten chinese text recognition with attention convolutional recurrent network,

    Z. Wu, X. Qu, J. Huang, and X. Wu, “In-air handwritten chinese text recognition with attention convolutional recurrent network,” in MultiMedia Modeling: 29th International Conference, MMM 2023, Bergen, Norway, January 9–12, 2023, Proceedings, Part II , pp. 695– 707, Springer, 2023

    2023

  26. [34]

    Casia online and offline chinese handwriting databases,

    C.-L. Liu, F. Yin, D.-H. Wang, and Q.-F. Wang, “Casia online and offline chinese handwriting databases,” in 2011 International Conference on Document Analysis and Recognition , pp. 37–41, 2011

    2011

  27. [35]

    An end-to-end air writing recognition method based on transformer,

    X. Tan, J. Tong, T. Matsumaru, V . Dutta, and X. He, “An end-to-end air writing recognition method based on transformer,” IEEE Access, 2023

    2023

  28. [36]

    3d fingertip and palm tracking in depth image sequences,

    H. Liang, J. Yuan, and D. Thalmann, “3d fingertip and palm tracking in depth image sequences,” in Proceedings of the 20th ACM international conference on Multimedia , pp. 785–788, 2012

    2012

  29. [37]

    Multi-touchless: Real-time fingertip de- tection and tracking using geodesic maxima,

    P. Krejov and R. Bowden, “Multi-touchless: Real-time fingertip de- tection and tracking using geodesic maxima,” in 2013 10th IEEE International Conference and Workshops on Automatic Face and Gesture Recognition (FG), pp. 1–7, IEEE, 2013

    2013

  30. [38]

    Model-based 3d hand pose estimation from monocular video,

    M. de La Gorce, D. J. Fleet, and N. Paragios, “Model-based 3d hand pose estimation from monocular video,” IEEE transactions on pattern analysis and machine intelligence , vol. 33, no. 9, pp. 1793–1805, 2011

    2011

  31. [39]

    Spatio- temporal hough forest for efficient detection–localisation–recognition of fingerwriting in egocentric camera,

    H. J. Chang, G. Garcia-Hernando, D. Tang, and T.-K. Kim, “Spatio- temporal hough forest for efficient detection–localisation–recognition of fingerwriting in egocentric camera,” Computer Vision and Image Understanding, vol. 148, pp. 87–96, 2016

    2016

  32. [40]

    Medi- apipe: A framework for building perception pipelines,

    C. Lugaresi, J. Tang, H. Nash, C. McClanahan, E. Uboweja, M. Hays, F. Zhang, C.-L. Chang, M. G. Yong, J. Lee, et al. , “Medi- apipe: A framework for building perception pipelines,” arXiv preprint arXiv:1906.08172, 2019

    1906 arXiv

  33. [41]

    Mediapipe hands: On-device real-time hand tracking,

    F. Zhang, V . Bazarevsky, A. Vakunov, A. Tkachenka, G. Sung, C.-L. Chang, and M. Grundmann, “Mediapipe hands: On-device real-time hand tracking,” arXiv preprint arXiv:2006.10214 , 2020

    2006 arXiv

  34. [42]

    Dynamic attention guided multi-trajectory analysis for single object tracking,

    X. Wang, Z. Chen, J. Tang, B. Luo, Y . Wang, Y . Tian, and F. Wu, “Dynamic attention guided multi-trajectory analysis for single object tracking,” IEEE Transactions on Circuits and Systems for Video Tech- nology, vol. 31, no. 12, pp. 4895–4908, 2021

    2021

  35. [43]

    DFAT: dynamic feature- adaptive tracking,

    W. Zhang, L. Jiao, F. Liu, S. Yang, and J. Liu, “DFAT: dynamic feature- adaptive tracking,” IEEE Trans. Circuits Syst. Video Technol. , vol. 33, no. 1, pp. 43–58, 2023

    2023

  36. [44]

    Siamon: Siamese occlusion-aware network for visual tracking,

    C. Fan, H. Yu, Y . Huang, C. Shan, L. Wang, and C. Li, “Siamon: Siamese occlusion-aware network for visual tracking,” IEEE Trans. Circuits Syst. Video Technol., vol. 33, no. 1, pp. 186–199, 2023

    2023

  37. [45]

    Siamese-based twin attention network for visual tracking,

    H. Bao, P. Shu, H. Zhang, and X. Liu, “Siamese-based twin attention network for visual tracking,” IEEE Trans. Circuits Syst. Video Technol., vol. 33, no. 2, pp. 847–860, 2023

    2023

  38. [46]

    Switch and refine: A long-term tracking and segmentation framework,

    X. Xu, J. Zhao, J. Wu, and F. Shen, “Switch and refine: A long-term tracking and segmentation framework,” IEEE Trans. Circuits Syst. Video Technol., vol. 33, no. 3, pp. 1291–1304, 2023

    2023

  39. [47]

    Icdar 2013 chinese handwriting recognition competition,

    F. Yin, Q.-F. Wang, X.-Y . Zhang, and C.-L. Liu, “Icdar 2013 chinese handwriting recognition competition,” in 2013 12th international con- ference on document analysis and recognition , pp. 1464–1470, IEEE, 2013

    2013

  40. [48]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , pp. 770–778, 2016

    2016

  41. [49]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift,

    S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in Proceedings of the 32nd International Conference on Machine Learning (F. Bach and D. Blei, eds.), vol. 37 of Proceedings of Machine Learning Research , (...

    2015

  42. [50]

    Dropout: a simple way to prevent neural networks from overfit- ting,

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhut- dinov, “Dropout: a simple way to prevent neural networks from overfit- ting,” The journal of machine learning research, vol. 15, no. 1, pp. 1929– 1958, 2014

    1929

  43. [51]

    Delving deep into rectifiers: Surpassing human-level performance on imagenet classification,

    K. He, X. Zhang, S. Ren, and J. Sun, “Delving deep into rectifiers: Surpassing human-level performance on imagenet classification,” in Proceedings of the IEEE international conference on computer vision , pp. 1026–1034, 2015

    2015

  44. [52]

    Graph attention networks,

    P. Velickovic, G. Cucurull, A. Casanova, A. Romero, P. Lio, Y . Bengio, et al., “Graph attention networks,” stat, vol. 1050, no. 20, pp. 10–48550, 2017

    2017

  45. [53]

    Pytorch: Tensors and dynamic neural networks in python with strong gpu acceleration,

    A. Paszke, S. Gross, S. Chintala, and G. Chanan, “Pytorch: Tensors and dynamic neural networks in python with strong gpu acceleration,” PyTorch: Tensors and dynamic neural networks in Python with strong GPU acceleration, vol. 6, no. 3, p. 67, 2017

    2017

  46. [54]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  47. [55]

    The kinetics human action video dataset,

    W. Kay, J. Carreira, K. Simonyan, B. Zhang, and A. Zisserman, “The kinetics human action video dataset,” 2017

    2017

  48. [57]

    Deep lstm networks for online chinese handwriting recognition,

    L. Sun, T. Su, C. Liu, and R. Wang, “Deep lstm networks for online chinese handwriting recognition,” in 2016 15th International Conference on Frontiers in Handwriting Recognition (ICFHR) , pp. 271–276, 2016

    2016

  49. [58]

    A new perspective: Recognizing online handwritten chinese characters via 1-dimensional cnn,

    J. Gan, W. Wang, and K. Lu, “A new perspective: Recognizing online handwritten chinese characters via 1-dimensional cnn,” Information Sciences, vol. 478, pp. 375–390, 2019

    2019

  50. [59]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” CoRR, vol. abs/1706.03762, 2017

    2017 arXiv

  51. [60]

    Connection- ist temporal classification: labelling unsegmented sequence data with recurrent neural networks,

    A. Graves, S. Fern ´andez, F. Gomez, and J. Schmidhuber, “Connection- ist temporal classification: labelling unsegmented sequence data with recurrent neural networks,” in Proceedings of the 23rd international conference on Machine learning , pp. 369–376, 2006

    2006

  52. [61]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” arXiv preprint arXiv:1609.02907 , 2016

    2016 arXiv

  53. [62]

    Drawing and recognizing chinese characters with recurrent neural network,

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

    2018

  54. [63]

    An image is worth 16x16 words: Trans- formers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Trans- formers for image recognition at scale,” CoRR, vol. abs/2010.11929, 2020

    2010 arXiv

  55. [64]

    Neural machine translation by jointly learning to align and translate,

    D. Bahdanau, K. Cho, and Y . Bengio, “Neural machine translation by jointly learning to align and translate,” 2016. Meiqi Wu received the MSc Degree from the University of Science and Technology of China In 2021 and continued to study for her doctorate at the School of Compute...

    2016

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.