Motion Primitive Discovery in a Humanoid Robot via Self-Organising Maps for Phase Recognition

Igor Farka\v{s}; Radovan Gregor

REVIEW 4 major objections 5 minor 19 references

Motion Primitive Discovery in a Humanoid Robot via Self-Organising Maps for Phase Recognition

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Self-organising map trajectories, temporally integrated by an echo state network, alone recognise the current phase of a humanoid robot's movement at 93.9% accuracy; adding world-state context adds only one percentage point.

desk verdict SOMs were trained on all 4,200 trials including the 840 'test' trials, so the 93.9% held-out number only holds out the readout, not the primitives; fix that and the causal windowing and the paper becomes credible. read the letter →

arxiv 2607.18737 v1 pith:Z5J44YS2 submitted 2026-07-21 cs.RO

Radovan Gregor , Igor Farkaš This is my paper · ORCID

classification cs.RO

keywords motionprimitiveself-organisingmapechostatenetworkphaserecognitionmirrorneuronsystemhumanoidrobotonline

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 asks how much of a robot's ongoing movement can be recognised from the temporal sequence of learned motor primitives alone, and how much requires world-state context. It answers, on NICO humanoid simulation data, that two self-organising maps—one for arm kinematics, one for hand configuration—yield best-matching-unit trajectories that an echo state network reads to 93.9% phase-recognition accuracy across 27 phase labels. Adding action identity, palm–object distance, and contact state raises accuracy by only one percentage point. The authors take this as evidence that self-organised motor representations, when temporally integrated, carry the dominant phase-discriminative structure; context refines but does not drive recognition. The result matters because it suggests an unsupervised motor vocabulary can support online action understanding without end-to-end training on labelled sequences.

What carries the argument

The load-bearing object is the pair of best-matching-unit (BMU) weight vectors per timestep. Two Kohonen self-organising maps are trained on non-redundant kinematic features selected by hierarchical correlation clustering (22 of 57 candidates). At inference, each raw frame windows (0.5s, 50% overlap, labeled by central frame) maps to a BMU on each map; the BMU's weight vector is the primitive descriptor. An echo state reservoir (300 neurons, spectral radius 0.99, leaking rate 0.5) accumulates these descriptors through its state and a ridge-regression readout classifies the current phase. The machinery does two things: it discretises continuous kinematics into a reusable primitive vocabulary,

What would settle it

Compute the BMU for each timestep using only a causal window ending at that timestep (or instantaneous frames), retrain the readout, and compare the phase-recognition accuracy against 93.9%. If the accuracy drops materially, the claim that SOM activation trajectories alone suffice for online recognition is weakened; if it stays near 94%, the causal window objection is moot.

Watch

Extended reading notes

Core claim

The central claim is that SOM activation trajectories, not contextual cues, carry the phase-discriminative structure of movement. For each timestep, the ESN receives the weight vectors of the best-matching units on A-SOM and H-SOM; from these 22 values it predicts one of 27 movement phases online. On 840 held-out trials, this primitive-only input reaches 93.9% accuracy, and injecting a seven-dimensional action one-hot, a distance, and a contact flag adds at most 1.0 percentage point. Hyperparameter runs show the context gain shrinks as reservoir size grows—from +4.1pp at 30 neurons to +0.6pp at 1,000—indicating context compensates for limited memory rather than carrying independent phase inf

Load-bearing premise

The whole online-recognition claim rests on the assumption that the best-matching-unit for timestep t is computed from data available up to t; the paper's SOM windows are 0.5s and labelled at the centre, which may silently give the classifier about 0.25s of future frames.

Editorial extensions

If this is right

Unsupervised primitive learning can supply phase labels: no end-to-end training on labelled phase sequences is needed to reach ~94% online recognition.
Context is a refinement channel, not a core signal; at larger reservoirs, adding world-state context buys less than one percentage point.
Arm and hand should be represented as separate primitive spaces; together they resolve each other's ambiguity (H-SOM disambiguates finger-opening transitions A-SOM cannot see).
The approach is offline-trained but online-operating; the same reservoir state could be retargeted to predict next primitive instead of classifying current phase.
Phase recognition accuracy rises with reservoir capacity, so the bottleneck is temporal memory, not primitive quality at these map sizes.

Reading between the lines

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

The reported 93.9% should be read as the performance of a window-centred descriptor; if BMU lookup were restricted to causal windows, accuracy might be lower.
If the context gain truly vanishes with larger reservoirs, then the phase-recognition problem is essentially solved by the primitive sequence; future work could test whether the same holds for cross-embodiment observation, where context might matter more.
A testable extension: feed raw 22 kinematic features directly to the same ESN (no SOM) and compare; the paper explicitly does not claim to outperform end-to-end, but such a comparison would isolate what the topographic discretisation contributes.
Another extension: use the SOM+ESN as a generative model by training the reservoir to predict the next BMU, then decode imagined trajectories; the paper mentions this as future work but does not test it.

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.

Desk Editor's Note

read the letter

Quick take: the dual SOM+ESN setup is a reasonable integration, and the finding that context adds only ~1pp is worth taking seriously. But the stress-test note is right: Section 4.2 says the split comes from 'the original 4,200 simulated trials from SOM training,' which means the A-SOM and H-SOM were trained on all 4,200 trials, including the 840 test trials. The 93.9% figure therefore holds out only the ESN readout, not the SOM feature extractor. Retraining the SOMs on the 3,360 training trials could shift the test BMU trajectories and change PRA; the paper never reports that comparison. That is load-bearing because the contribution is that self-organized primitives generalize to unseen trials.

There are two smaller issues. Section 2.3 windows are labeled by the central frame, so if the ESN's per-timestep BMU lookup uses those windows, each decision uses about 0.25s of future kinematics. The authors need to state whether inference uses instantaneous frames or window means; as written, the 'online, available up to that timestep' claim is not clean. Also, there are no repeated-seed variances, no raw-kinematics baseline, and the hyperparameter grid is only partially reported. Those are addressable but they affect how much of the accuracy should be credited to the SOM representation versus reservoir size.

Credit: the research question is explicit; the readout is analytic ridge regression rather than iterative fitting; the complementary A-SOM/H-SOM analysis is clear; the single-trial account of how the two maps resolve ambiguity is the best part. The finding that context gain shrinks with larger reservoirs is a nice way to separate memory from information. The work extends the authors' earlier ESN and SOM work, but the dual-SOM primitive representation plus the context-economics result is a legitimate increment, not a repackaging.

Verdict: the paper deserves a serious referee, but the authors should be asked to retrain the SOMs on the training split only, fix the causal windowing, report seed variances, and release code/data. I would not desk-reject; I would send it out with a request for major revision. The core question -- whether unsupervised primitives plus a reservoir can carry phase recognition with little context -- is still open, and this paper, once cleaned up, is a reasonable step toward answering it.

Referee Report

4 major / 5 minor

Summary. The paper proposes a two-level, MNS-inspired architecture for online movement-phase recognition on the simulated NICO humanoid. Level one trains two Self-Organising Maps (A-SOM on arm kinematics, H-SOM on hand kinematics) from 4,200 simulated trials covering seven actions and 27 phases. Level two feeds the BMU weight vectors into an Echo State Network whose readout is trained by ridge regression to predict the current phase. The headline result is that the context-free primitive descriptor achieves 93.9% phase-recognition accuracy on the ESN test split, with the full context configuration adding only +1.0 pp. The paper concludes that SOM activation trajectories carry the dominant phase-discriminative structure and that world-state context is a secondary refinement.

Significance. If the reported numbers are trustworthy, the paper offers a compact demonstration that an unsupervised topographic representation, temporally integrated by a reservoir, can support high-accuracy phase recognition on a humanoid robot, and that contextual information plays a minor role. The use of an analytic ridge readout, a stratified action split for the ESN, and a hyperparameter sensitivity table are concrete strengths. However, the central claim is currently compromised by two load-bearing evaluation issues: the SOMs are trained on all 4,200 trials before the ESN split, and the ESN input construction appears to use future information under the stated windowing scheme. The limited-context conclusion also lacks repeated-seed variance. These issues are fixable, but the headline figure is not yet established as a clean held-out, causal, online result.

major comments (4)

[§4.2 / §3.1] The reported 93.9% PRA is not a fully held-out result. Section 4.2 states: 'The dataset consists of the original 4,200 simulated trials from SOM training' and then applies the 80:20 split only for the ESN readout. Since the A-SOM and H-SOM in §3.1 were trained on all 4,200 trials, the BMU weight vectors b_A(t), b_H(t) used as ESN inputs for the 840 'test' trials come from feature extractors that have already seen those same trials. The abstract and Table 3 therefore overstate the generalization claim. The authors should retrain the SOMs on the 3,360 training trials only, compute BMU trajectories for the 840 left-out trials, and report the resulting PRA. Without this, the contribution that unsupervised primitives generalize to novel trials is not established.
[§2.3 / §4.2] The online/causal claim is not supported by the described pipeline. Section 2.3 forms windows of W=30 frames (0.5 s) with stride S=15 and labels them by the central frame; the ESN input in §4.2 is the BMU of this window representation at timestep t. If the ESN queries the SOM with the window mean, then each input u(t) incorporates kinematic information from approximately t-0.25 s to t+0.25 s, i.e., about 0.25 s of future data. This contradicts the 'available up to that timestep' framing in Section 1. If instead the ESN queries with instantaneous raw frames, the input distribution differs from the window-mean training distribution of the SOMs, creating a train/inference mismatch. The authors should specify the exact per-timestep input and, if causal, recompute with causal features only.
[§5.1 / §5.2] The central 'context is secondary' claim is based on single ESN runs with no repeated-seed variance. The reservoir matrices are randomly initialized; the reported +1.0 pp difference between ctx-no and ctx-all at N=300 could easily be within run-to-run variability, as is the 0.6 pp difference at N=1000. The sensitivity table shows that reservoir capacity dominates, but it lists only selected runs and no error bars. The authors should run multiple seeds (e.g., 10) for each configuration and report mean ± std, particularly for ctx-no versus ctx-all.
[§4.2 / §5.1] There is no baseline against raw kinematics, so the specific contribution of the SOM stage is not isolated. The paper explicitly says it does not aim to outperform a direct-kinematics ESN, but the claim that 'SOM activation trajectories preserve the dominant phase-discriminative structure' (Abstract, §5.1) still requires a control: an ESN trained directly on the raw 22D kinematic features (or on random projections) would show whether the 93.9% figure depends on the SOM topology or simply reflects the ease of the phase-classification task. Adding this baseline is important for the stated claim about motor primitives.

minor comments (5)

[§5.2, Table 4] The caption says '9 run grid omitted for brevity', but only four rows (R1, R5, R8, R9) are shown. The claim that accuracy increases monotonically with N is not verifiable from the selected rows; please provide the full grid.
[§6] The limitation paragraph acknowledges that the model is trained and evaluated on NICO's own actions. This is appropriate, but the same paragraph should also acknowledge the SOM split issue and the causal input issue, since both directly affect the validity of the current numeric claims.
[§3.2 / §4.1] The term 'motion primitive' is used for SOM prototype nodes, but no quantitative evidence is provided that the same primitive is reused across different actions or phases beyond the single-trial trajectory in Figure 4. A cross-action BMU reuse analysis would support the 'reusable building block' definition in §2.1.
[§4.2] The notation 'stride S=5' for ESN subsampling is confusing because Section 2.3 already uses stride S=15 for windowing. Consider using different symbols to distinguish the two subsampling operations.
[§5.1] The sentence 'All results are reported on the held-out test set' is misleading given that the SOMs were trained on all trials. It should be rephrased to 'held-out with respect to the ESN readout' or the experiment must be changed accordingly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: SOMs are unsupervised, the ESN readout is trained on held-out phase targets, and the reported accuracy is a genuine held-out measurement.

full rationale

The derivation chain is self-contained. The SOMs are trained on kinematic windows using the Kohonen rule (Eq. 1) without phase labels; phase/action labels are used only for post-hoc purity analysis, not for updating weights. The ESN input u(t)=[b_A(t), b_H(t)] is just the BMU weight vectors, and the reservoir update (Eqs. 3-4) integrates them temporally. The readout W_out is the only supervised component and is fitted by ridge regression (Eq. 5) on the 3,360 training trials; the 840 test trials are used only to measure accuracy (Eq. 6). No fitted parameter is renamed as a prediction: the 93.9% PRA is a measured transfer of the trained readout to a separated trial split. Self-citations ([7], [13], [14]) are background references and are not load-bearing; no uniqueness theorem or ansatz is imported from them. One should note a separate evaluation-validity concern: Section 4.2 says the 840 test trials are drawn from the same 4,200 trials used to train the SOMs, so the SOM feature extractor has seen the test kinematics, and Section 2.3's windowing with central-frame labels could introduce non-causal information if BMU lookup uses window means. These are correctness/causality issues, not circularity: the phase-recognition result is not forced by construction, and no equation reduces to its own input. Hence the circularity score is 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 1 invented entities

The central claim depends on a handful of user-chosen quantities: the τ cut, windowing, SOM grid sizes, reservoir hyperparameters, and the retained feature set. No repeated-seed variance is given. The main axioms are that the simulation is a faithful proxy for NICO and that the 4-parameter hand controller is a valid stand-in for hand shaping; the latter is directly responsible for the near-perfect H-SOM grasp purity. The causal-availability of ESN inputs is an implicit assumption contradicted by the central-frame labeling unless windows are causal.

free parameters (6)

Dendrogram cut τ = 0.7
Converts 57 candidate kinematic features into 22 SOM inputs; changing the cut changes the representation and all downstream accuracy.
SOM window length / stride = W=30 frames (0.5s), S=15 (50% overlap)
Windowing and central-frame labels determine the temporal smoothing of the primitive descriptors.
SOM grid sizes = A-SOM 35×35; H-SOM 18×18
A-SOM grid chosen by grid search over phase purity (+5.5pp vs 25×25); H-SOM size reflects observed hand-state count.
ESN reservoir hyperparameters = N=300, ρ=0.99, α=0.50, λ=1e-4 (main result)
Main accuracy numbers are single configurations; partial grid shows N dominates but scatter across random seeds is not reported.
ESN stride = S=5 frames (83 ms)
Subsampling rate for the temporal sequence fed to the reservoir.
Retained feature set = 22 features (13 arm, 9 hand)
Result of functional exclusions plus τ=0.7 clustering; a modeling choice with direct effect on what the SOMs see.

assumptions (5)

standard math Standard SOM and ESN mathematics (Kohonen update, reservoir state equations, ridge-regression readout).
Used as accepted background; not the paper's contribution.
domain assumption Unity physics simulation with Barracuda ONNX IK faithfully represents NICO kinematics.
All data are simulated; no real-robot validation is reported (Section 2.1, Conclusion).
domain assumption The 4-parameter hand synergy controller adequately captures human-like hand shaping.
H-SOM's 97.2% grasp purity and low utilization may be artifacts of the controller's ~6 discrete poses (Sections 2.1, 3.2).
domain assumption Ground-truth phase labels from the action state machines are correct and unambiguous.
Phase purity and PRA are defined relative to these labels (Section 2.1).
ad hoc to paper The ESN input at timestep t contains no information from after time t.
Required for the 'online ... available up to that timestep' claim but not guaranteed by the central-frame windowing in Section 2.3.

invented entities (1)

SOM prototype nodes as reusable 'motion primitives'
purpose: Discrete vocabulary for phase recognition, discovered unsupervised from arm/hand kinematics.
The paper operationalizes prior notion of motor primitives as BMU neurons, but provides no independent evidence (e.g., transfer to new actions/embodiments) that these nodes correspond to generalizable primitives.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Motion Primitive Discovery in a Humanoid Robot via Self-Organising Maps for Phase Recognition." pith.science (2026). https://pith.science/paper/Z5J44YS2

@misc{pith2026260718737,
  author       = {Pith},
  title        = {Pith review of: Motion Primitive Discovery in a Humanoid Robot via Self-Organising Maps for Phase Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z5J44YS2}},
  note         = {Machine review of arXiv:2607.18737}
}

read the original abstract

Understanding the computational basis of action recognition is a central challenge in social cognition as well as in human-robot interaction. Inspired by the Mirror Neuron System (MNS), we propose a two-level architecture for motor primitive discovery and online phase recognition applied to the NICO humanoid robot. At the first level, two Self-Organising Maps (SOMs) learn topographic representations of arm kinematics (A-SOM) and hand kinematics (H-SOM) from simulated trials covering seven motor actions. The maps are trained on non-redundant features identified through hierarchical correlation analysis of motion trajectories. The results show that the two SOMs encode complementary aspects of motor behaviour. At the second level, an Echo State Network (ESN) evaluates whether temporal trajectories of SOM activations, represented by consecutive best-matching units, are sufficient for online recognition of the currently executed movement phase. The results show that SOM-based trajectories preserve the dominant phase-discriminative structure of the movement, while contextual information provides only a secondary refinement. Our contribution is the integration of established SOM and ESN methods within an MNS-inspired architecture for motor primitive representation and online phase recognition. The results are compatible with the computational hypothesis that self-organised motor representations, when temporally integrated, can support accurate online recognition of ongoing movement phases.

Figures

Figures reproduced from arXiv: 2607.18737 by the authors.

**Figure 1.** Processing pipeline of the SOM+ESN motion primitive recognition system. 2 Methods Here we describe the whole data processing pipeline illustrated in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** A-SOM dominant phase per neuron. Neuron colour indicates the phase whose training samples most frequently activated that neuron. The map exhibits clear topographical organisation (Mean Phase Purity 73.9%). 3.2 Results A-SOM achieves 90.0% neuron utilisation, with dead neurons concentrating at the map periphery as expected. Phase purity (73.9%) is the strongest label dimension, with large coherent zones emerging wi… view at source ↗

**Figure 3.** H-SOM dominant grasp type per neuron. Neuron colour indicates the dominant grasp synergy with its purity percentage. (Mean Purity 97.2%). a temporal one – each hand pose occupies its own island with no kinematic path between them, reflecting the hard discreteness of the synergy controller. Action purity (72.9%) and phase purity (76.1%, exceeding A-SOM) confirm that finger-level signals carry complementary discrimin… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Single-trial trajectory on A-SOM representing Eat action traces a smooth leftto-right path. Across 68 timesteps, the trajectory visits 46 unique neurons, indicating that the movement progresses through a structured sequence of arm-motion primitives. and hand configura…

**Figure 5.** Figure 5: ESN phase recognition timeline for representative Eat trial. Top: groundtruth phase sequence. Middle: phase distribution of the active A-SOM neuron at each timestep. Bottom: phase distribution of the active H-SOM neuron. The complementary coverage of A-SOM and H-SOM i…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references

[1]

Bertasius, G., Wang, H., Torresani, L.: Is space-time attention all you need for video understanding? In: ICML. pp. 813–824 (2021)

2021
[2]

PLOS ONE6(11), 1–15 (2011)

Chersi, F., Ferrari, P.F., Fogassi, L.: Neuronal chains for actions in the parietal lobe: A computational model. PLOS ONE6(11), 1–15 (2011)

2011
[3]

Annual Review of Neuroscience23, 613–647 (2000)

deCharms, R.C., Zador, A.: Neural representation and the cortical code. Annual Review of Neuroscience23, 613–647 (2000)

2000
[4]

Brain Research1083, 174–188 (2006)

Erlhagen,W.,Mukovskiy,A.,Bicho,E.:Adynamicmodelforactionunderstanding and goal-directed imitation. Brain Research1083, 174–188 (2006)

2006
[5]

Science308(2005)

Fogassi, L., Ferrari, P.F., Gesierich, B., Rozzi, S., Chersi, F., Rizzolatti, G.: Parietal lobe: From action organization to intention understanding. Science308(2005)

2005
[6]

Brain119, 593–609 (1996)

Gallese, V., Fadiga, L., Fogassi, L., Rizzolatti, G.: Action recognition in the pre- motor cortex. Brain119, 593–609 (1996)

1996
[7]

In: ICACR (2025)

Gregor, R., Farkaš, I., Malinovská, K., Sobota, B., Kerzel, M.: Mirroring-based prediction of robot arm movements using an echo state network. In: ICACR (2025)

2025
[8]

Frontiers in Psychology1, 243 (2011)

Hemeren, P.E., Thill, S.: Deriving motor primitives through action segmentation. Frontiers in Psychology1, 243 (2011)

2011

Show all 19 references

[9]

In: RO-MAN

Kerzel, M., et al.: NICO – Neuro-Inspired COmpanion: A developmental humanoid robot platform for multimodal interaction. In: RO-MAN. pp. 113–120 (2017)

2017
[10]

Bio- logical Cybernetics43(1), 59–69 (1982)

Kohonen, T.: Self-organized formation of topologically correct feature maps. Bio- logical Cybernetics43(1), 59–69 (1982)

1982
[11]

Computer Science Review3(3), 127–149 (2009)

Lukoševičius, M., Jaeger, H.: Reservoir computing approaches to recurrent neural network training. Computer Science Review3(3), 127–149 (2009)

2009
[12]

Neural Networks19, 254–271 (2006)

Oztop, E., Kawato, M., Arbib, M.A.: Mirror neurons and imitation: A computa- tionally guided review. Neural Networks19, 254–271 (2006)

2006
[13]

In: ICDL (2019)

Pospíchal, J., Farkaš, I., Pecháč, M., Malinovská, K.: Modeling self-organized emer- gence of perspective in/variant mirror neurons in a robotic system. In: ICDL (2019)

2019
[14]

Rebrová, K., Pecháč, M., Farkaš, I.: Towards a robotic model of the mirror neuron system. In: Int. Conf. on Development and Learning and on Epig. Robotics (2013)

2013
[15]

Annual Review of Neu- roscience27, 169–192 (2004)

Rizzolatti, G., Craighero, L.: The mirror-neuron system. Annual Review of Neu- roscience27, 169–192 (2004)

2004
[16]

Journal of Neuroscience18(23), 10105–10115 (1998)

Santello, M., Flanders, M., Soechting, J.F.: Postural hand synergies for tool use. Journal of Neuroscience18(23), 10105–10115 (1998)

1998
[17]

Cognitive Computation3(4), 525–538 (2011)

Thill, S., Svensson, H., Ziemke, T.: Modeling the development of goal-specificity in mirror neurons. Cognitive Computation3(4), 525–538 (2011)

2011
[18]

Foundations and Trends in Robotics4, 105–223 (2013)

Thomaz, A., Hoffman, G., Cakmak, M.: Computational human-robot interaction. Foundations and Trends in Robotics4, 105–223 (2013)

2013
[19]

Philosophical Transactions of the Royal So- ciety B358(1431), 593–602 (2003)

Wolpert, D.M., Doya, K., Kawato, M.: A unifying computational framework for motor control and social interaction. Philosophical Transactions of the Royal So- ciety B358(1431), 593–602 (2003)

2003

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Bertasius, G., Wang, H., Torresani, L.: Is space-time attention all you need for video understanding? In: ICML. pp. 813–824 (2021)

2021

[2] [2]

PLOS ONE6(11), 1–15 (2011)

Chersi, F., Ferrari, P.F., Fogassi, L.: Neuronal chains for actions in the parietal lobe: A computational model. PLOS ONE6(11), 1–15 (2011)

2011

[3] [3]

Annual Review of Neuroscience23, 613–647 (2000)

deCharms, R.C., Zador, A.: Neural representation and the cortical code. Annual Review of Neuroscience23, 613–647 (2000)

2000

[4] [4]

Brain Research1083, 174–188 (2006)

Erlhagen,W.,Mukovskiy,A.,Bicho,E.:Adynamicmodelforactionunderstanding and goal-directed imitation. Brain Research1083, 174–188 (2006)

2006

[5] [5]

Science308(2005)

Fogassi, L., Ferrari, P.F., Gesierich, B., Rozzi, S., Chersi, F., Rizzolatti, G.: Parietal lobe: From action organization to intention understanding. Science308(2005)

2005

[6] [6]

Brain119, 593–609 (1996)

Gallese, V., Fadiga, L., Fogassi, L., Rizzolatti, G.: Action recognition in the pre- motor cortex. Brain119, 593–609 (1996)

1996

[7] [7]

In: ICACR (2025)

Gregor, R., Farkaš, I., Malinovská, K., Sobota, B., Kerzel, M.: Mirroring-based prediction of robot arm movements using an echo state network. In: ICACR (2025)

2025

[8] [8]

Frontiers in Psychology1, 243 (2011)

Hemeren, P.E., Thill, S.: Deriving motor primitives through action segmentation. Frontiers in Psychology1, 243 (2011)

2011