Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

Morpheus: A Neural-driven Animatronic Face with Hybrid Actuation and Diverse Emotion Control

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper presents Morpheus, an animatronic face whose hybrid rigid-and-tendon actuation, combined with a self-modeling inverse-learning network and a speech-to-blendshape model, generates distinct emotion-specific expressions—happiness…

desk verdict A real hardware integration with a sound inverse-learning idea, undermined on paper by an unspecified landmark coordinate alignment and a flat contradiction about real-time speed. read the letter →

arxiv 2507.16645 v1 pith:ZEANEYMA submitted 2025-07-22 cs.RO

Zongzheng Zhang , Jiawen Yang , Ziqiao Peng , Meng Yang , Jianzhu Ma , Lin Cheng , Huazhe Xu , Hang Zhao
show 1 more author
Hao Zhao
This is my paper
classification cs.RO
keywords animatronicfacehybridactuationspeech-drivenfacialanimationemotiondisentanglementself-modelingnetworkblendshapesinverselearninghuman-robotinteraction
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 presents Morpheus, an animatronic face that combines rigid-driven actuation for the eyes and mouth with tendon-driven actuation for the nose and cheeks, using 33 actuators total. The central claim is that this hybrid hardware, together with a learned inverse model, lets the face speak arbitrary sentences while displaying distinct, emotion-specific expressions—happiness, fear, disgust, and anger. The authors argue this is the first speech-driven animatronic face to demonstrate emotion-specific control across multiple emotions on one platform. They support the claim with a self-modeling network that converts motor commands into facial landmarks, an inverse-learning step that maps virtual blendshape expressions to motor commands, and evaluations of both emotion recognition and lip synchronization.

What carries the argument

The load-bearing machinery has two parts. On the hardware side, the hybrid actuation layout uses 33 actuators: rigid four-bar and five-bar mechanisms drive the eyebrows, eyes, mouth, and neck, while four tendon-driven strings deform the nose and cheeks. On the software side, the self-modeling network is an MLP trained on random motor-command pairs with their observed 2D facial landmarks; once frozen, gradients of the $\ell^1$ distance between real and virtual 468-point landmark sets are backpropagated through it to solve for the 26 motor commands that reproduce a target expression. The speech module uses an emotion-disentangling encoder, dynamic time warping for temporal alignment, and a transformer decoder with emotion-guided attention to output 33 blendshape coefficients per frame, which the inverse loop converts into physical actuation.

What would settle it

Run the inverse-learning loop while recording the physical face with a second, off-axis camera, and compare that camera's landmarks with the virtual face projected from the same viewpoint; if the front-facing $\ell^1$ loss is near zero while the off-axis view shows systematic deviation, then the 2D landmark objective is not uniquely recovering the intended 3D expression.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a speech-to-expression pipeline built on top of a hybrid-actuated face can produce emotionally distinct, speech-synchronized facial expressions. The hybrid actuation is the enabler on the hardware side: rigid mechanisms give precise, large-range control of eyebrows, eyes, and mouth, while tendon-driven strings deform the nose and cheek for subtle microexpressions, all within a compact 33-actuator head. On the software side, the authors learn a forward model—an MLP mapping 26 motor commands to 468 2D facial landmarks—from randomly generated motor commands, then fix that model and backpropagate an $\ell^1$ landmark-distance loss (Eq. 5) to recover motor commands for any desired blendshape expression. A separate transformer-based module disentangles content from emotion in speech, aligns features with dynamic time warping, and outputs 33 blendshape coefficients; feeding these through the inverse model drives the physical face. A video-based expression recognizer classifies the resulting robot faces as happy (90%) and angry (91%), with lower accuracy for disgust (66%) and fear (73%).

Load-bearing premise

The pipeline assumes that minimizing the average $\ell^1$ distance between 2D landmarks of the physical face and the virtual face is enough to recover motor commands that reproduce the intended 3D expression, even though 2D projections are ambiguous and the silicone skin's deformation is only sparsely sampled by those landmarks.

Editorial extensions

If this is right

  • Adding the tendon-driven nose and cheek module raises recognition accuracy across the four emotions from 64.5% to 80%, with the largest gains for happy (72% to 90%) and angry (75% to 91%).
  • The same self-modeling inverse-learning recipe generalizes across four Morpheus hardware versions with 17, 16, 23, and 26 expression degrees of freedom, with landmark distance improving from 1.652 mm to 0.748 mm.
  • On the speech-to-blendshape task, the model reports lower lip-vertex error and emotional-vertex error than prior speech-driven 3D face animation methods on both emotional and neutral speech test sets.
  • A supplementary user study prefers Morpheus over three recent animatronic faces for realistic static expressions in 84% of comparisons and rates its speech-driven clips 7.6 out of 10 for naturalness and emotional clarity.

Reading between the lines

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

  • Because the forward model is trained purely on random motor commands and observed landmarks, the same inverse-learning recipe should transfer to other silicone-skinned animatronic faces, as long as corresponding landmarks can be detected on both the virtual and real faces; the paper demonstrates this only across its own hardware versions.
  • The 'any given sentence' claim is bounded by the training distribution: the speech model is learned from acted emotional speech corpora, so unscripted, non-acted, or non-English speech would test how far the emotion control actually generalizes.
  • The quantitative emotion evaluation relies on a video-based expression recognizer, whose own biases could inflate the reported distinctness; a blind forced-choice human labeling study on the same 400 clips would give a more direct measure of perceptual separability.
  • The hybrid-actuation ablation isolates the tendon module as a whole; ablating the nose tendon and cheek tendon separately would show which subtle region contributes most to each emotion.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. Morpheus presents an animatronic face platform with 33 actuators in a hybrid rigid/tendon configuration, a self-modeling MLP that maps motor commands to MediaPipe facial landmarks, gradient-based inversion that converts blendshape coefficients into motor commands, and a transformer-based speech-to-blendshape module trained on the 3D-ETF dataset. The authors report hardware ablations, a confusion-matrix evaluation of four emotional expressions, LVE/EVE comparisons against prior speech-driven methods, and a supplementary user study. The paper releases both hardware and software repositories.

Significance. If the reported results hold, the main contribution is a complete, reproducible pipeline from emotional speech to an animatronic face, with the inverse-learning bridge being a genuinely learned forward model inverted by gradient descent rather than a circular fit to the target. The hybrid actuation design and the release of hardware/software resources are valuable to the community. However, the central claims of emotion-specific control and cross-domain fidelity currently rest on sparse quantitative evaluations, an unstated coordinate-frame alignment in the inverse-learning loss, and an internally inconsistent real-time statement, so the significance will depend on the revisions addressing these points.

major comments (5)
  1. [III-D, Eq. (5)] The Face Inverse Learning objective in Eq. (5) minimizes the L1 distance between C_r (468 MediaPipe landmarks of the real face, in normalized image coordinates) and C_v (468 landmarks of the virtual face), but the paper does not state how C_v is generated or how the two coordinate frames are aligned. MediaPipe normalized landmarks depend on face scale, position, and camera FOV; without an explicit similarity transform or a common rendering/alignment procedure, the L1 loss contains a large expression-independent offset, and gradient descent can reduce Eq. (5) by matching global geometry rather than the intended expression. This is load-bearing for the self-modeling module and for every claim that the real face reproduces the virtual expression. Please specify the normalization/calibration protocol, or replace the loss with a pose-invariant metric.
  2. [IV-D, Table I] The confusion matrix in Table I is the only quantitative evidence for the central claim that distinct emotional expressions can be generated from arbitrary sentences. It is a single evaluation of 400 videos, with no repeated trials, no error bars or confidence intervals, no chance-level baseline, and no significance test. The high confusion between disgust (66%) and fear (73%) also suggests the emotion-specific controls are not fully disambiguated. The paper should report variability across sentence sets, classifier runs, or hardware resets, and should include a statistical comparison to chance or to a neutral baseline.
  3. [V vs. Supplementary Sec. I] There is a direct internal contradiction about real-time performance. Section V states that the facial expression generation system has 'relatively high inference times' and that future work must achieve 'near real-time dialogue,' while Supplementary Sec. I claims the system 'infers at about 150 FPS on NVIDIA Jetson AGX Xavier, generates virtual expressions at 30 FPS, and controls servos at 50 Hz.' The manuscript needs to reconcile these statements and state clearly whether the full pipeline runs in real time, since the introduction claims real-time capability.
  4. [Supplementary Sec. F] The user study is not a controlled experiment. It reports 50 participants, an 84% preference rate, and a 7.6/10 realism score, but gives no details on stimulus selection, randomization or order balancing, blinding, inter-rater agreement, or confidence intervals. Comparisons against [9], [19], and [32] use materials from published papers and public demos rather than matched, same-protocol stimuli. Because the novelty claim relative to earlier systems is partly supported by this study, the user study should either be substantially strengthened or downgraded to a pilot with explicitly limited conclusions.
  5. [IV-D, Table II] The LVE/EVE comparison in Table II is reported without standard deviations, number of test sequences, or explicit train/test split and evaluation-protocol details. The EVE metric is defined over manually selected eye/forehead vertices, and it is not established that the same vertex selection and preprocessing were used for all compared methods. Without these details, the claimed improvements over EmoTalk, FaceFormer, and MeshTalk are not fully verifiable. Please add error bars, protocol descriptions, and, if possible, evaluation code or a released test set.
minor comments (5)
  1. [III-D, Eq. (5)] The definition of C_v contains a typo: it is written as C^v = (c^v_1, ..., c^r_N), but the last element should be c^v_N.
  2. [IV-E, Table IV] The text refers to 'mAP' when Table IV reports classification accuracy; these are different quantities and the terminology should be corrected.
  3. [IV-C] The section titled 'Quantitative Results' contains only qualitative descriptions; the quantitative results appear in Section IV-D. Consider renaming the sections to avoid confusing the reader.
  4. [Supplementary Sec. D] The phrase 'between between' appears in the description of Fig. 8 and should be corrected to 'between the virtual facial expressions and the real facial expressions.'
  5. [Supplementary Sec. E] The text states that the self-modeling network is evaluated using Mean Absolute Error between predicted and ground-truth landmarks, but Table VI reports 'Landmark Distance (mm)' with no MAE values. Please clarify which quantity is reported and add units or standard deviations.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the pipeline is a supervised speech-to-blendshape model plus a learned forward model inverted by gradient descent, with external FER benchmarking.

full rationale

The derivation chain is self-contained. The Emotional Facial Animation module is trained with standard supervised losses (Eqs. 3-4) against ground-truth blendshape coefficients from the public 3D-ETF dataset, and its quality is reported with LVE/EVE against held-out meshes; no loss term is equal to the claimed output. The Face Inverse Learning module trains an MLP forward map from random motor commands to real MediaPipe landmarks, then inverts that fixed map by gradient descent on Eq. 5 to obtain motor commands for a given virtual landmark target. The physical robot's resulting expressions are then evaluated by an external FER network (Table I) and by landmark displacement, which are genuine generalization checks rather than re-statements of a fitted target. The only overlapping-author citation is EmoTalk/3D-ETF, used as a public dataset and as an external baseline; it does not force the result. The paper's omission of how virtual landmarks are extracted and aligned with the real camera frame is a reproducibility and correctness risk, not a circularity, because it concerns whether Eq. 5 is well-posed rather than whether a result reduces to its inputs by construction.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claim rests on several domain assumptions about sensors, datasets, and learned models that are stated but not independently validated. No new physical entities are introduced. The main free parameters are hand-set engineering choices (motor bounds, skin thickness, loss weights) that shape the achievable expression space.

free parameters (3)
  • Loss balancing weights lambda_1, lambda_2 = 1.0, 1.0
    Set empirically in Sec. III-E; no sensitivity analysis is provided.
  • Per-motor control bounds (33 motors) = lower/start/upper values in Table V
    Hand-defined after testing to avoid structural interference; these bounds define the training distribution for the self-model and limit the achievable expressions.
  • Silicone skin thickness = 4 mm (2 mm at eyelids)
    Selected from qualitative experiments (Supplementary Fig. 10); directly affects tendon-driven deformation and expression quality.
assumptions (6)
  • domain assumption MediaPipe face landmark detection returns accurate 2D landmarks on the robotic face with silicone skin.
    Used in Sec. III-D to create training targets; if detection is biased by the artificial face, the self-model learns a distorted mapping.
  • domain assumption An MLP can learn a sufficiently accurate forward map from 26 motor commands to 468x2 landmarks from 5000 random samples.
    Sec. III-D and IV-A; the inversion via gradient descent inherits all approximation errors of the forward model.
  • domain assumption The 3D-ETF dataset provides valid ground-truth blendshape coefficients for emotional speech.
    Sec. IV-A; the speech-to-blendshape module is trained on this dataset, so its quality bounds the emotional fidelity.
  • domain assumption Virtual blendshape landmarks and real MediaPipe landmarks are commensurable after some unspecified alignment.
    Eq. 5 compares coordinates directly; no calibration or normalization is described, yet the two landmark sets come from different renderers and sensors.
  • domain assumption Backpropagation through the MLP reliably finds motor commands minimizing landmark loss.
    Sec. III-D; non-convex optimization could converge to poor local optima for expressions outside the training distribution.
  • domain assumption A video-based FER network's classifications are a valid proxy for expression quality.
    Sec. IV-D; the network is trained on human faces and used to score robot expressions without validation on robots.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Morpheus: A Neural-driven Animatronic Face with Hybrid Actuation and Diverse Emotion Control." pith.science (2026). https://pith.science/paper/ZEANEYMA

@misc{pith2026250716645,
  author       = {Pith},
  title        = {Pith review of: Morpheus: A Neural-driven Animatronic Face with Hybrid Actuation and Diverse Emotion Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZEANEYMA}},
  note         = {Machine review of arXiv:2507.16645}
}
read the original abstract

Previous animatronic faces struggle to express emotions effectively due to hardware and software limitations. On the hardware side, earlier approaches either use rigid-driven mechanisms, which provide precise control but are difficult to design within constrained spaces, or tendon-driven mechanisms, which are more space-efficient but challenging to control. In contrast, we propose a hybrid actuation approach that combines the best of both worlds. The eyes and mouth-key areas for emotional expression-are controlled using rigid mechanisms for precise movement, while the nose and cheek, which convey subtle facial microexpressions, are driven by strings. This design allows us to build a compact yet versatile hardware platform capable of expressing a wide range of emotions. On the algorithmic side, our method introduces a self-modeling network that maps motor actions to facial landmarks, allowing us to automatically establish the relationship between blendshape coefficients for different facial expressions and the corresponding motor control signals through gradient backpropagation. We then train a neural network to map speech input to corresponding blendshape controls. With our method, we can generate distinct emotional expressions such as happiness, fear, disgust, and anger, from any given sentence, each with nuanced, emotion-specific control signals-a feature that has not been demonstrated in earlier systems. We release the hardware design and code at https://github.com/ZZongzheng0918/Morpheus-Hardware and https://github.com/ZZongzheng0918/Morpheus-Software.

Figures

Figures reproduced from arXiv: 2507.16645 by the authors.

Figure 1
Figure 1. Morpheus: an animatronic face with diverse expressions. (a) Front view. Blue markers indicate the attachment points between the underlying mechanical structure and the soft skin, while yellow arrows denote the directions of movement. (b) Side view highlighting motion capabilities. Blue arrows indicate the three-axis neck movement: nodding, shaking, and rotation. The green arrow illustrates the jaw’s ability for hori… view at source ↗
Figure 2
Figure 2. Overview of the Mechanism for Generating Realistic Speech-driven Facial Expressions. We propose a method that disentangles content and emotion from expressive speech and generates blendshape coefficients using a transformer-based decoder with emotion-guided attention. These coefficients are then mapped to actuator commands through a face inverse learning module, ultimately producing realistic facial expressions that… view at source ↗
Figure 3
Figure 3. Hardware Design of Our Mechanical Face Platform. (a) Front view of the mechanical skeletal structure. (b) Transparent side view of the mechanical structure, highlighting tendon-driven movements in the nose and cheek areas (indicated by blue arrows). (c) Four rigid-driven modules (indicated by green arrows), including the eyebrow module (4 actuators), eye module (6 actuators), mouth module (16 actuators), and neck mo… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: shows 25 different human facial expressions we se￾lected, ranging from basic emotions such as ”angry,” ”happy,” ”sad,” ”fear,” ”disgust,” to more nuanced ones like ”skeptical,” ”shy,” and ”sympathetic.” These expressions are generated through precise control of the eye…
Figure 5
Figure 5. Figure 5: Complete sentences with diverse facial expressions. The first row illustrates the virtual expressions generated by our algorithm rendered in Blender, while the second row displays the corresponding real-world expressions reproduced by the animatronic face. (a) Happy ex…
Figure 6
Figure 6. Figure 6: Comparison of happy expression using different actuation systems. (a) Mouth-Only System. (b) Eyebrow-Only System. (c) Rigid-Driven System w/o tendon driven. (d) Hybrid-Driven System (Ours). high accuracies for happy (90%) and angry (91%) expressions, reflecting its str…
Figure 7
Figure 7. Figure 7: Motor Distribution and Control Areas of the Mechanical Face Hardware Platform. 1 - Left Cheek (tendon drive). 2 - Left Nose (tendon drive). 3 - Left Eyebrow Center. 4 - Left Eyebrow Peak. 5 - Left Upper Eyelid. 6 - Left Lower Eyelid. 7 - Eyes Up/Down Gaze. 8 - Upper Li…
Figure 8
Figure 8. Figure 8: Visualization of Morpheus interpreting the sentence ”look at the sky” with happy, fear, angry and disgust expressions, respectively [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 10
Figure 10. Figure 10: Impact of silicone skin thickness on tendon-driven facial deformation. (a-b) For thinner skins (2mm and 2.5mm), the deformation is unnaturally localized around the pasted region (indicated by the base of the arrow). (c) At 4mm thickness (our selected configuration), t…
Figure 9
Figure 9. Figure 9: Morpheus design evolution across four iterations. (a) Internal skeletal structures. (b) External facial appearances with silicone skin applied. The results demonstrate the self-modeling network per￾formed well across all four versions, showing its ability to adapt to d…

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Automated Synthesis of Facial Mechanisms for Conversational Animatronic Robots

    cs.RO 2026-07 conditional novelty 6.0 of 10

    A parametric linkage face template plus hierarchical collision-driven optimization synthesizes manufacturable facial mechanisms from 2D portraits and runs them with dual-identity conversational motion.

Reference graph

Works this paper leans on

57 extracted references · 53 canonical work pages · cited by 1 Pith paper

  1. [9]

    Smile like you mean it: Driving animatronic robotic face with learned models

    Boyuan Chen, Yuhang Hu, Lianfeng Li, Sara Cummings, and Hod Lipson. Smile like you mean it: Driving animatronic robotic face with learned models. In 2021 IEEE International Conference on Robotics and Automation (ICRA) , pages 2739–

    work page 2021

  2. [19]

    Human-robot facial coexpression

    Yuhang Hu, Boyuan Chen, Jiong Lin, Yunzhe Wang, Yingke Wang, Cameron Mehlman, and Hod Lipson. Human-robot facial coexpression. Science Robotics, 9(88):eadi4724, 2024

    work page 2024

  3. [32]

    Driving Animatronic Robot Facial Expression From Speech

    Boren Li, Hang Li, and Hangxin Liu. Driving Anima- tronic Robot Facial Expression From Speech. arXiv preprint arXiv:2403.12670, 2024

    work page Pith review arXiv 2024

  4. [1]

    The design of an expressive humanlike socially assistive robot

    Brian Allison, Goldie Nejat, and Emmeline Kao. The design of an expressive humanlike socially assistive robot. 2009

    work page 2009

  5. [2]

    Humanoid head face mechanism with expandable facial expressions

    Wagshum Techane Asheber, Chyi-Yeu Lin, and Shih Hsiang Yen. Humanoid head face mechanism with expandable facial expressions. International Journal of Advanced Robotic Sys- tems, 13(1):29, 2016

    work page 2016

  6. [3]

    wav2vec 2.0: A framework for self-supervised learning of speech representations

    Alexei Baevski, Yuhao Zhou, Abdelrahman Mohamed, and Michael Auli. wav2vec 2.0: A framework for self-supervised learning of speech representations. Advances in neural infor- mation processing systems , 33:12449–12460, 2020

    work page 2020

  7. [4]

    Towards a humanoid museum guide robot that interacts with multiple persons

    Maren Bennewitz, Felix Faber, Dominik Joho, Michael Schreiber, and Sven Behnke. Towards a humanoid museum guide robot that interacts with multiple persons. In 5th IEEE- RAS International Conference on Humanoid Robots, 2005. , pages 418–423. IEEE, 2005

    work page 2005

  8. [5]

    Using dynamic time warping to find patterns in time series

    Donald J Berndt and James Clifford. Using dynamic time warping to find patterns in time series. In Proceedings of the 3rd international conference on knowledge discovery and data mining, pages 359–370, 1994

    1994

Show all 57 references
  1. [6]

    Control of facial expressions of the humanoid robot head roman

    Karsten Berns and Jochen Hirth. Control of facial expressions of the humanoid robot head roman. In 2006 IEEE/RSJ inter- national conference on intelligent robots and systems , pages 3119–3124. IEEE, 2006

    2006

  2. [7]

    Emotion and sociable humanoid robots

    Cynthia Breazeal. Emotion and sociable humanoid robots. International journal of human-computer studies , 59(1-2):119– 155, 2003

    2003

  3. [8]

    The Cog project: Building a humanoid robot

    Rodney A Brooks, Cynthia Breazeal, Matthew Marjanovi ´c, Brian Scassellati, and Matthew M Williamson. The Cog project: Building a humanoid robot. In International workshop on computation for metaphors, analogy, and agents , pages 52–87. Springer, 1998

    1998

  4. [10]

    Fully body visual self-modeling of robot morphologies

    Boyuan Chen, Robert Kwiatkowski, Carl V ondrick, and Hod Lipson. Fully body visual self-modeling of robot morphologies. Science Robotics, 7(68):eabn1944, 2022

    2022

  5. [11]

    Capture, learning, and synthesis of 3d speaking styles

    Daniel Cudeiro, Timo Bolkart, Cassidy Laidlaw, Anurag Ran- jan, and Michael J Black. Capture, learning, and synthesis of 3d speaking styles. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 10101–10111, 2019

    2019

  6. [12]

    Faceformer: Speech-driven 3d facial animation with transformers

    Yingruo Fan, Zhaojiang Lin, Jun Saito, Wenping Wang, and Taku Komura. Faceformer: Speech-driven 3d facial animation with transformers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 18770– 18780, 2022

    2022

  7. [13]

    Facially expressive humanoid robotic face

    Zanwar Faraj, Mert Selamet, Carlos Morales, Patricio Torres, Maimuna Hossain, Boyuan Chen, and Hod Lipson. Facially expressive humanoid robotic face. HardwareX, 9:e00117, 2021

    2021

  8. [14]

    Local robust sparse representation for face recognition with single sample per person

    Jianquan Gu, Haifeng Hu, and Haoxi Li. Local robust sparse representation for face recognition with single sample per person. IEEE/CAA Journal of Automatica Sinica, 5(2):547–554, 2017

    2017

  9. [15]

    Learning human-like facial expressions for android phillip k

    Ahsan Habib, Sumit K Das, Ioana-Corina Bogdan, David Han- son, and Dan O Popa. Learning human-like facial expressions for android phillip k. dick. In 2014 IEEE International Con- ference on Automation Science and Engineering (CASE) , pages 1159–1165. IEEE, 2014

    2014

  10. [16]

    Identity emulation (ie): Bio-inspired facial expression interfaces for emotive robots

    David Hanson, Giovanni Pioggia, S Dinelli, Fabio Di Francesco, R Francesconi, and Danilo De Rossi. Identity emulation (ie): Bio-inspired facial expression interfaces for emotive robots. In AAAI Mobile Robot Competition , pages 72–82, 2002

    2002

  11. [17]

    Development of face robot for emotional communication be- tween human and robot

    Takuya Hashimoto, Sachio Hiramatsu, and Hiroshi Kobayashi. Development of face robot for emotional communication be- tween human and robot. In 2006 International Conference on Mechatronics and Automation, pages 25–30. IEEE, 2006

    2006

  12. [18]

    Dynamic display of facial expressions on the face robot made by using a life mask

    Takuya Hashimoto, Sachio Hiramatsu, and Hiroshi Kobayashi. Dynamic display of facial expressions on the face robot made by using a life mask. In Humanoids 2008-8th IEEE-RAS International Conference on Humanoid Robots, pages 521–526. IEEE, 2008

    2008

  13. [20]

    Optimizing android facial expres- sions using genetic algorithms

    Hyun-Jun Hyung, Han Ul Yoon, Dongwoon Choi, Duk-Yeon Lee, and Dong-Wook Lee. Optimizing android facial expres- sions using genetic algorithms. Applied Sciences , 9(16):3379, 2019

    2019

  14. [21]

    Real- istic child robot “affetto” for understanding the caregiver-child attachment relationship that guides the child development

    Hisashi Ishihara, Yuichiro Yoshikawa, and Minoru Asada. Real- istic child robot “affetto” for understanding the caregiver-child attachment relationship that guides the child development. In 2011 ieee international conference on development and learning (icdl), volume 2, pages ...

    2011

  15. [22]

    Identification and evaluation of the face system of a child android robot affetto for surface motion design

    Hisashi Ishihara, Binyi Wu, and Minoru Asada. Identification and evaluation of the face system of a child android robot affetto for surface motion design. Frontiers in Robotics and AI , 5:119, 2018

    2018

  16. [23]

    Mechanical design of emotion expression humanoid robot we-4rii

    Kazuko Itoh, Hiroyasu Miwa, Massimiliano Zecca, Hideaki Takanobu, Stefano Roccella, Maria Chiara Carrozza, Paolo Dario, and Atsuo Takanishi. Mechanical design of emotion expression humanoid robot we-4rii . Springer, 2006

    2006

  17. [24]

    Audio-driven emotional video portraits

    Xinya Ji, Hang Zhou, Kaisiyuan Wang, Wayne Wu, Chen Change Loy, Xun Cao, and Feng Xu. Audio-driven emotional video portraits. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 14080–14089, 2021

    2021

  18. [25]

    Cybernetic human hrp-4c

    Kenji Kaneko, Fumio Kanehiro, Mitsuharu Morisawa, Kanako Miura, Shin’ichiro Nakaoka, and Shuuji Kajita. Cybernetic human hrp-4c. In 2009 9th IEEE-RAS International Conference on Humanoid Robots , pages 7–14. IEEE, 2009

    2009

  19. [26]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph., 42(4):139–1, 2023

    2023

  20. [27]

    Study on face robot for active human interface-mechanisms of face robot and expression of 6 basic facial expressions, 1993

    Hiroshi Kobayashi and Fumio Hara. Study on face robot for active human interface-mechanisms of face robot and expression of 6 basic facial expressions, 1993

    1993

  21. [28]

    Study of a face robot platform as a kansei medium

    Hiroshi Kobayashi, T Tsuji, and K Kikuchi. Study of a face robot platform as a kansei medium. In 2000 26th Annual Con- ference of the IEEE Industrial Electronics Society. IECON 2000. 2000 IEEE International Conference on Industrial Electronics, Control and Instrumentation. 21s...

    2000

  22. [29]

    Realization of realistic and rich facial expressions by face robot

    Hiroshi Kobayashi, Yoshiro Ichikawa, Masaru Senda, and Taichi Shiiba. Realization of realistic and rich facial expressions by face robot. In Proceedings 2003 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2003)(Cat. No. 03CH37453), volume 2, pages 11...

    2003

  23. [30]

    Mu- seum guide robot based on sociological interaction analysis

    Yoshinori Kuno, Kazuhisa Sadazuka, Michie Kawashima, Kei- ichi Yamazaki, Akiko Yamazaki, and Hideaki Kuzuoka. Mu- seum guide robot based on sociological interaction analysis. In Proceedings of the SIGCHI conference on Human factors in computing systems, pages 1191–1194, 2007

    2007

  24. [31]

    Development of an android for emotional expres- sion and human interaction

    Dong-Wook Lee, Tae-Geun Lee, B So, Moosung Choi, Eun- Cheol Shin, K Yang, Moon-Hong Baek, Hong-Seok Kim, and Ho-Gil Lee. Development of an android for emotional expres- sion and human interaction. In Seventeenth world congress the international federation of automatic control,...

    2008

  25. [33]

    The realization of robot theater: Humanoid robots and theatric performance

    Chyi-Yeu Lin, Chang-Kuo Tseng, Wei-Chung Teng, Wei-Chen Lee, Chung-Hsien Kuo, Hung-Yan Gu, Kuo-Liang Chung, and Chin-Shyurng Fahn. The realization of robot theater: Humanoid robots and theatric performance. In 2009 International Confer- ence on Advanced Robotics , pages 1–6. I...

    2009

  26. [34]

    Emotion classification using a cnn lstm- based model for smooth emotional synchronization of the humanoid robot ren-xin

    Ning Liu and Fuji Ren. Emotion classification using a cnn lstm- based model for smooth emotional synchronization of the humanoid robot ren-xin. PloS one, 14(5):e0215216, 2019

    2019

  27. [35]

    Differentiable robot rendering

    Ruoshi Liu, Alper Canberk, Shuran Song, and Carl V on- drick. Differentiable robot rendering. arXiv preprint arXiv:2410.13851, 2024

    2024 arXiv

  28. [36]

    Expression snippet transformer for robust video-based facial expression recognition

    Yuanyuan Liu, Wenbin Wang, Chuanxu Feng, Haoyu Zhang, Zhe Chen, and Yibing Zhan. Expression snippet transformer for robust video-based facial expression recognition. Pattern Recognition, 138:109368, 2023

    2023

  29. [37]

    A facial expression emotion recognition based human-robot interaction system

    Zhentao Liu, Min Wu, Weihua Cao, Luefeng Chen, Jianping Xu, Ri Zhang, Mengtian Zhou, and Junwei Mao. A facial expression emotion recognition based human-robot interaction system. IEEE CAA J. Autom. Sinica , 4(4):668–676, 2017

    2017

  30. [38]

    The ryerson audio- visual database of emotional speech and song (ravdess): A dynamic, multimodal set of facial and vocal expressions in north american english

    Steven R Livingstone and Frank A Russo. The ryerson audio- visual database of emotional speech and song (ravdess): A dynamic, multimodal set of facial and vocal expressions in north american english. PloS one, 13(5):e0196391, 2018

    2018

  31. [39]

    Mediapipe: A framework for building perception pipelines

    Camillo Lugaresi, Jiuqiang Tang, Hadon Nash, Chris McClana- han, Esha Uboweja, Michael Hays, Fan Zhang, Chuo-Ling Chang, Ming Guang Yong, Juhyun Lee, et al. Mediapipe: A framework for building perception pipelines. arXiv preprint arXiv:1906.08172, 2019

    1906 arXiv

  32. [40]

    The bielefeld anthropomorphic robot head “flobi”

    Ingo L ¨utkebohle, Frank Hegel, Simon Schulz, Matthias Hackel, Britta Wrede, Sven Wachsmuth, and Gerhard Sagerer. The bielefeld anthropomorphic robot head “flobi”. In 2010 IEEE international conference on robotics and automation , pages 3384–3391. IEEE, 2010

    2010

  33. [41]

    Hefes: An hybrid engine for facial expressions synthesis to control human-like androids and avatars

    Daniele Mazzei, Nicole Lazzeri, David Hanson, and Danilo De Rossi. Hefes: An hybrid engine for facial expressions synthesis to control human-like androids and avatars. In 2012 4th IEEE RAS & EMBS International Conference on biomedical robotics and biomechatronics (BioRob) , pa...

    2012

  34. [42]

    Creating facial motions of cybernetic human hrp-4c

    Shin’ichiro Nakaoka, Fumio Kanehiro, Kanako Miura, Mit- suharu Morisawa, Kiyoshi Fujiwara, Kenji Kaneko, Shuuji Kajita, and Hirohisa Hirukawa. Creating facial motions of cybernetic human hrp-4c. In 2009 9th IEEE-RAS International Conference on Humanoid Robots , pages 561–567. ...

    2009

  35. [43]

    Design of android type humanoid robot albert hubo

    Jun-Ho Oh, David Hanson, Won-Sup Kim, Young Han, Jung- Yup Kim, and Ill-Woo Park. Design of android type humanoid robot albert hubo. In 2006 IEEE/RSJ International Conference on Intelligent Robots and Systems , pages 1428–1433. IEEE, 2006

    2006

  36. [44]

    Selftalk: A self- supervised commutative training diagram to comprehend 3d talking faces

    Ziqiao Peng, Yihao Luo, Yue Shi, Hao Xu, Xiangyu Zhu, Hongyan Liu, Jun He, and Zhaoxin Fan. Selftalk: A self- supervised commutative training diagram to comprehend 3d talking faces. In Proceedings of the 31st ACM International Conference on Multimedia , pages 5292–5301, 2023

    2023

  37. [45]

    Emotalk: Speech- driven emotional disentanglement for 3d face animation

    Ziqiao Peng, Haoyu Wu, Zhenbo Song, Hao Xu, Xiangyu Zhu, Jun He, Hongyan Liu, and Zhaoxin Fan. Emotalk: Speech- driven emotional disentanglement for 3d face animation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 20687–20697, 2023

    2023

  38. [46]

    Automatic facial expression learning method based on humanoid robot xin-ren

    Fuji Ren and Zhong Huang. Automatic facial expression learning method based on humanoid robot xin-ren. IEEE Trans- actions on Human-Machine Systems , 46(6):810–821, 2016

    2016

  39. [47]

    Meshtalk: 3d face anima- tion from speech using cross-modality disentanglement

    Alexander Richard, Michael Zollh ¨ofer, Yandong Wen, Fernando de la Torre, and Yaser Sheikh. Meshtalk: 3d face anima- tion from speech using cross-modality disentanglement. In 2021 IEEE/CVF International Conference on Computer Vision (ICCV), pages 1153–1162. IEEE, 2021

    2021

  40. [48]

    An android for emotional interaction: spatiotemporal validation of its facial expressions

    Wataru Sato, Shushi Namba, Dongsheng Yang, Shin’ya Nishida, Carlos Ishi, and Takashi Minato. An android for emotional interaction: spatiotemporal validation of its facial expressions. Frontiers in psychology, 12:800657, 2022

    2022

  41. [49]

    Image ratio features for facial expression recognition application

    Mingli Song, Dacheng Tao, Zicheng Liu, Xuelong Li, and Mengchu Zhou. Image ratio features for facial expression recognition application. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics) , 40(3):779–788, 2009

    2009

  42. [50]

    Automatic generation of robot facial expres- sions with preferences

    Bing Tang, Rongyun Cao, Rongya Chen, Xiaoping Chen, Bei Hua, and Feng Wu. Automatic generation of robot facial expres- sions with preferences. In 2023 ieee international conference on robotics and automation (icra) , pages 7606–7613. IEEE, 2023

    2023

  43. [51]

    Development of the humanoid head portrait robot system with flexible face and expression

    Wu Weiguo, Men Qingmei, and Wang Yu. Development of the humanoid head portrait robot system with flexible face and expression. In 2004 IEEE International Conference on Robotics and Biomimetics, pages 757–762. IEEE, 2004

    2004

  44. [52]

    4d gaussian splatting for real-time dynamic scene rendering

    Guanjun Wu, Taoran Yi, Jiemin Fang, Lingxi Xie, Xiaopeng Zhang, Wei Wei, Wenyu Liu, Qi Tian, and Xinggang Wang. 4d gaussian splatting for real-time dynamic scene rendering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20310–20320, 2024

    2024

  45. [53]

    A survey on perception methods for human–robot interaction in social robots

    Haibin Yan, Marcelo H Ang, and Aun Neow Poo. A survey on perception methods for human–robot interaction in social robots. International Journal of Social Robotics , 6:85–119, 2014

    2014

  46. [54]

    Facial Expression Realization of Humanoid Robot Head and Strain-Based Anthropomorphic Evaluation of Robot Facial Expressions

    Zhibin Yan, Yi Song, Rui Zhou, Liuwei Wang, Zhiliang Wang, and Zhendong Dai. Facial Expression Realization of Humanoid Robot Head and Strain-Based Anthropomorphic Evaluation of Robot Facial Expressions. Biomimetics, 9(3):122, 2024

    2024

  47. [55]

    Optimizing facial expres- sions of an android robot effectively: a bayesian optimization approach

    Dongsheng Yang, Wataru Sato, Qianying Liu, Takashi Minato, Shushi Namba, and Shin’ya Nishida. Optimizing facial expres- sions of an android robot effectively: a bayesian optimization approach. In 2022 IEEE-RAS 21st International Conference on Humanoid Robots (Humanoids) , page...

    2022

  48. [56]

    Flow-guided one-shot talking face generation with a high- resolution audio-visual dataset

    Zhimeng Zhang, Lincheng Li, Yu Ding, and Changjie Fan. Flow-guided one-shot talking face generation with a high- resolution audio-visual dataset. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3661–3670, 2021

    2021

  49. [57]

    Visemenet: Audio-driven animator-centric speech animation

    Yang Zhou, Zhan Xu, Chris Landreth, Evangelos Kalogerakis, Subhransu Maji, and Karan Singh. Visemenet: Audio-driven animator-centric speech animation. ACM Transactions on Graphics (TOG), 37(4):1–10, 2018. In this supplementary material, we provide more details about Morpheus, ...

    2018

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.