Skip to main content
NCBI home page
Scientific Data logoLink to Scientific Data
. 2026 Feb 18;13:474. doi: 10.1038/s41597-026-06843-x

A Psychophysical Dataset for Vibrotactile Augmented Perception

Mostafa Hamidifard 1, Samar Nikfarjad 1, Hosein Pirmohammadi 1, Rezvan Nasiri 1,✉, Majid Nili Ahmadabadi 1
PMCID: PMC13031721  PMID: 41708644

Abstract

Tactile perception modeling has been important research topic for years. However, due to individual differences and complexity of human cognition, the tactile perception modeling is challenging. One of the main challenges in this field is the lack of a rich dataset with sufficient inter-subject, inter-session, and intra-session diversities. Here, for the first time, we present a dataset of human position and intensity perception for 51 different vibration patterns. The dataset includes the experimental results for 40 individuals (20 female and 20 male) in two different sessions. The experimental results include the perceived intensity and position of each vibration pattern as well as selection time and confidence level. After each session the participants were also asked to fill a questionnaire file. We also collected the anthropometric and demographic data, and the participants underwent Bioelectrical Impedance Analysis (BIA) to measure body composition indicators. The detailed results for each participant is located in the dataset. This dataset can be used to develop tactile perception models, study individuals’ perception differences, and design tactile sensory feedback.

Subject terms: Data publication and archiving, Computational models

Background & Summary

Only in USA, the number of amputee individuals exceeded 1.6 million in 2005, and it is estimated that this number will be more than twice by 20501. When we upscale these numbers in worldwide level, we can understand that there are a huge number of amputee individuals that require a prosthetic limb to continue their daily life. One of the best solutions to assist the amputee individuals is using neuroprosthetics2,3. The conventional neuroprosthetics (e.g., prosthetic hands) solely rely on the motor command decoding to control the prosthetic functions. However, despite the achievements on the motor command decoding, the performance of these devices are highly limited due to the lack of sensory tactile feedback4,5. Without tactile sensory feedback, the visual attention is required to control the prosthetic, which is not only cognitively demanding6,7, but also very slow and cannot function as tactile sensation8. Providing sensory feedback through non-invasive means is therefore a critical research direction. Among these, vibrotactile feedback via vibration motors is a favorable solution due to its cost efficiency and fewer implementation challenges9 compared to other methods such as pressure mechanisms10 or electrical stimulation11.

In recent years, there have been enormous efforts to design perceivable tactile feedback using vibration motors to encode position12–14, force13,15–19, and state14,17,18,20,21. Notable advancements include compact proprioceptive feedback for prostheses via spatial encoding and Gaussian interpolation22, navigation systems using phantom tactile sensations to guide users with directional cues23, and wrist-based vibrotactile path-following for motor learning24. Applications extend to obstacle detection in dynamic environments through haptic vests employing vibrotactile patterns25, object recognition using evoked multimodal feedback26, comparative studies of vibrotactile and joint-torque feedback27, and stroke rehabilitation via supplemental kinesthetic feedback28. Haptic guidance strategies utilizing optimal reciprocal collision avoidance have further enhanced interaction in dynamic settings29. Investigations into spatial acuity on the back30, force-modulated grasp control31, and stiffness discrimination improvements in prostheses further highlight the versatility of vibrotactile solutions. Reviews synthesizing non-invasive methods9 underscore its potential across prosthetics, rehabilitation, and assistive navigation. As a conclusion, we refer to the work32 done by Hong Z Tan et al.

Due to the significant diversity among individuals’ tactile perception33, proposing effective vibrotactile pattern requires vast datasets with inter-subject33, inter-session17, and intra-session17 variability. On the other hand, data collection for tactile perception modeling and recovery is a time and energy consuming task34. To address this gap, we present, for the first time, a comprehensive tactile sensory perception dataset recorded using a five-motor vibrotactile array with 51 vibration patterns (comprising single- and dual-motor activations at three intensity levels). The dataset was collected from 40 healthy adult participants (20 male and 20 female) across 255 trials per participant (five repetitions for each pattern) in two separate sessions spaced 0 to 34 days apart (mean interval about 12 days).

The collected dataset includes the perceived vibration position and intensity in both sessions as well as the participants’ selection time and confidence level, alongside their answers to a questionnaire file containing questions such as the physical demand, cognitive load, and fatigue. The dataset also includes their anthropometric and demographic data such as gender, weight, height, and waist circumference alongside their Bioelectrical Impedance Analysis (BIA) results of body composition indicators such as body fat percentage, subcutaneous fat percentage, and visceral fat index. This unique dataset can be used to extract cognitive tactile perception models35 using machine learning methods36. Using this dataset the inter-subject, inter-session, and intra-session variability in tactile sensory perception can also be modeled and studied37. In addition to basic science application, the proposed dataset can be used to propose perceivable vibration patterns for human-machine interface including the sensory feedback for prosthetic hand38, deep sensing for prosthetic leg39, situational awareness40, guiding blind individuals41, navigation42, and alarming systems25.

Methods

Participants

Forty healthy individuals (20 males and 20 females, age 24.5  ± 3.7 years, weight 71.8  ± 14.0 kg, and height 171.1  ± 10.9 cm; mean  ± s.t.d.) participated in the experimental study. The participants’ detailed anthropometric and demographic data is presented in Table 1 and Fig. 1.

Table 1.

Anthropometric and demographic data of participants. To maintain anonymity, height, weight, and waist circumference numbers have been rounded.

Subject ID Gender Session gap [day] Weight [kg] Height [cm] Waist circumference [cm] Hand preference Ticklishness
1 M 8 69 180 81 R 0
2 M 7 74 183 94 R 0
3 F 6 77 159 83 R 0
4 F 9 52 158 70 R 0
5 M 10 88 168 107 R 0
6 M 7 74 181 81 R 1
7 M 13 88 172 90 R 0
8 M 13 75 162 97 L 0
9 M 11 78 170 97 R 0
10 F 33 75 160 88 R 0
11 F 5 84 181 90 R 0
12 F 12 54 153 72 R 0
13 M 9 88 183 98 R 0
14 M 6 96 178 107 R 0
15 F 13 59 165 76 R 0
16 F 8 55 158 75 R 1
17 F 6 64 163 67 L 0
18 F 29 60 162 69 R 0
19 M 14 78 187 87 R 0
20 M 7 97 187 96 R 0
21 F 7 66 163 80 R 0
22 F 13 49 170 66 R 0
23 M 8 73 178 89 R 0
24 F 7 54 150 83 R 0
25 F 32 52 152 71 R 0
26 M 7 101 182 108 L 1
27 F 24 66 167 77 R 1
28 F 34 60 163 77 R 1
29 M 8 77 181 85 R 1
30 M 20 99 194 103 L 0
31 F 0 61 172 74 R 1
32 F 7 63 159 87 R 0
33 M 12 73 180 80 R 0
34 F 8 69 164 84 R 0
35 F 28 68 175 71 R 1
36 M 7 78 184 86 R 0
37 M 6 89 181 96 R 0
38 M 15 75 175 86 R 1
39 M 8 59 172 75 R 1
40 F 14 58 170 70 R 0
Open in a new tab

Fig. 1.

Fig. 1

Open in a new tab

Bar and box plot of anthropometric and demographic data presented at Table 1.

Participants were screened to be generally healthy. However, the following conditions were reported: one participant had a history of treated depression, two reported regular use of blood-pressure medication, and three had undergone abdominal surgery. Additionally, eleven participants exhibited sensitivity to ticklish stimuli.

Also, the participants underwent Bioelectrical Impedance Analysis (BIA) to measure body composition indicators including body fat percentage, subcutaneous fat percentage, visceral fat index, muscle mass percentage, basal metabolic rate, bone mass, body water percentage, metabolic age, body protein percentage. The detailed results for each participant are located in the dataset.

All participants received full details regarding the study’s purpose and procedures, including their right to withdraw at any time, and provided written informed consent. By signing, participants agreed to the publication of anonymized anthropometric, demographic, and study-related data, and consented to its storage and use in current and future research without any personally identifiable information. The consent form emphasized that the experiment posed no foreseeable risk, but would be halted immediately if any participant reported discomfort or harm. No adverse events or harms were reported during data collection. The study was conducted in accordance with the Declaration of Helsinki and approved by the Research Ethics Committee of Faculty of Physical Education and Sport Sciences University of Tehran (IR.UT.SPORT.REC.1403.039).

Experiment setup

This experiment was conducted using a vibration belt equipped with five similar vibration motors, attached to the participant’s abdomen, fastened around the front of the abdomen as can be seen in Fig. 2. The middle vibration motor was placed 2 cm above the navel. The left and right vibration motors were positioned on the very left and right sides of the participants’ torsos at the points they perceived as the boundary between the front and back of their bodies. Two other vibration motors were placed at intermediate position between middle, left, and right vibration motors. Medical-grade adhesive tapes were used to attach the vibration motors, ensuring a secure fit without causing skin irritation; see Fig. 2. All tactile stimuli were generated using the attached vibration motors (coin-type, model: VC1030A002F, manufacturer: Vybronics). The vibration motors were controlled via Pulse Width Modulation (PWM) with an 8-bit resolution, offering a value range from 0 to 255. In this setup, a value of 0 represented a 0% duty cycle (no vibration), and 255 represented a 100% duty cycle (vibration with maximum intensity). The vibration intensity is calibrated for each participant and each motor; see Familiarization subsection. The vibration motors’ circuit can be seen in Fig. 3.

Fig. 2.

Fig. 2

Open in a new tab

Overview of the experimental procedure. (A) Pre-experiment setup, including participant preparation and calibration. (B) Vibration stimuli presentation, where participants experienced 255 trials and recorded their responses. (C) Post-experiment questionnaire assessing participant experience. The lower section details the trial structure, including response selection, confidence level rating, and rest intervals. Note that the participant in (C) viewed this image and allowed us to use this image inside our manuscript.

Fig. 3.

Fig. 3

Open in a new tab

The experiment setup. The vibration level of the motors was controlled via PWM (ranging from 0 to 255). The output could take one of four values: zero (0), low (participant’s minimum perceivable intensity computed during calibration), medium (average of low and high), or high (255).

The experiment took place in a specially designed, acoustically isolated approximately 2 × 2 meters test room, where a monitor was positioned in front of the participant with a gray background. During the test, participants maintained a 40 cm distance from the monitor sitting on a fixed chair. To ensure full concentration on the experiment and reducing external interference, the participants were hearing a soft gray noise using a headphone, preventing them from hearing external sounds and the potential sound/noise produced by the vibration motors.

Experiment protocol

The overall experiment protocol includes two experimental sessions; the gap between the experiment sessions is presented in Table 1, where the zero gap indicates that the sessions are performed in the same day. Each session consisted of three main phases: familiarization (pre-experiment), the main test, and filling the questionnaire (post-experiment). In the main test, participants experienced 51 different vibration patterns, each pattern 5 times, in a randomized sequence; a total of 255 trials. Between two consecutive trials, there was a one-second interval which all vibration motors were turned off. After every 100 trials, the participants had a 3-minute break. Finally, at the end of each test session, participants completed a questionnaire regarding their experience. The full procedure is summarized in Fig. 4.

Fig. 4.

Fig. 4

Open in a new tab

This diagram illustrates the overall structure and stages of the experiment. On the left axis, the sequence of three experimental blocks is shown, consisting of two 100-trial blocks and one 55-trial block, separated by three-minute breaks to prevent participant’s fatigue and maintain data quality. On the right axis, the step-by-step process of each trial is presented. First, a vibrotactile stimulus is applied to the participant; second, the participant selects the perceived position of the vibration on the half-circle interface, rates their confidence on a five-grade scale, and finally proceed to the next trial at their own pace. This self-paced design allows for a detailed investigation of sensory localization, decision-making processes, and participant’s subjective confidence levels in response to tactile feedback.

Familiarization

The familiarization phase included seven steps: (1) attaching the vibration belt, (2) ensuring the belt is properly positioned on the participant’s abdomen, (3) checking if all vibration motors are functioning properly, (4) playing an instructional video explaining the experiment’s objective, procedure, and rules, (5) collecting anthropometric and demographic data such as gender, weight, height, and waist circumference, (6) calibrating the minimum detectable vibration intensity, and (7) checking the vibration levels.

Calibration: To determine the minimum detectable motor vibration intensities for participant, each of five vibration motors was randomly vibrated three times (a total of 15 times) with linearly increasing intensity from 0%. The participant was instructed to immediately click when the vibration was detected. Once all responses were recorded, the maximum value among the three minimum detected intensities for each vibration motor was selected as the minimum vibration intensity. For each motor, the minimum(maximum) vibration intensity is called as the low(high) vibration level, and the average of minimum and maximum vibration intensities is also called as the medium vibration level.

Check: To familiarize the participants with the vibration levels, the vibration motors were sequentially vibrated from the very left to the very right side of the abdomen at low, medium, and high levels, by a self selected pace during which the respective point was shown on the half-circle. This procedure was repeated three times after calibration.

Main test

During the main test, for each trial, participants experienced a vibration pattern. The vibration patterns were single and dual combinations of different vibration levels including 15 single-motor vibration patterns (low, medium, and high vibration levels for five vibration motors) and 36 dual-motor vibration patterns; vibration of four adjacent pairs of vibration motors for nine combined vibration levels.

The participants were asked to select the perceived vibration intensity and position as radius and angle of a half-circle on the screen. The highest and lowest perceived magnitudes were mapped to the outer and inner radius of the half-circle. Also, the very left and right perceived positions were mapped to 0 and 180 degrees on the half-circle. The vibration patterns along with the distribution of the perceived point selections on the screen for each pattern are illustrated in Fig. 5. After selecting the perceived points, the participants should also select the confidence level of their selection.

Fig. 5.

Fig. 5

Open in a new tab

Probability distribution of points selected by all participants in both sessions, grouped by unique vibration patterns. Each half-circle represents the front of the human abdomen, with red circles indicating vibration activation. Specifically, one, two, and three circles correspond to low, medium, and high vibration intensities, respectively.

Questionnaire

The questionnaire included eleven different questions including six scale questions, two yes/no questions, and three opinion questions. In scale questions, the participants were asked to rate their fatigue at the end of experiment, the physical difficulty of the experiment, and the required concentration between 0 and 10, where 0(10) refers to the least(highest) value. Additionally, they assessed the quality of information (angle and magnitude) encoding through the combination of vibration intensity and position. Also, there were two other scale questions to evaluate their perceptual accuracy, they specified the degree resolution at which they could distinguish vibration locations and the number of intensity levels they could differentiate. The overall responses for scaling questions are illustrated in Fig. 6. In yes/no questions, we asked the participants if this method could be potentially effective for (1) navigation in visually impaired individuals or (2) encoding force-position data to restore tactile feedback for amputee individuals.

Fig. 6.

Fig. 6

Open in a new tab

Questionnaire results of the (A) first and (B) second sessions. Q1: How fatigued did you feel during the experiment? (0, 1, …, 10) Q2: To what extent was the proposed method successful in encoding positional information and accurately combining vibration activations? (0, 1, …, 10) Q3: How physically demanding was this experiment for you? (0, 1, …, 10) Q4: To what extent did this experiment require focus and attention? (0, 1, …, 10) Q5: How accurately were you able to distinguish the positional information of the effective vibration center? (2, 4, …, 20 degrees) Q6: At how many levels were you able to distinguish the vibration intensity of the effective vibration center? (2, 3, …, 10 levels).

At the end of the questionnaire, participants provided any suggestions for improving the information integration process, regarding the two potential applications; assisting visually impaired individuals in navigation and restoring sensory feedback for amputee individuals. Moreover, participants were encouraged to propose alternative methods for aiding visually impaired and amputee individuals using vibrotactile feedback.

Data acquisition

At the beginning of the experiment, anthropometric and demographic data such as gender, weight, height, and waist circumference were recorded. The detailed results for each participant are located in the dataset. During the experiment, for each trial, the point selected by the participant on the screen along with the duration of each trial and the confidence level of selections were recorded. At the end of experiment, the questionnaire results were also recorded. After the second session of the experiment, for each participant, 9 body composition indicators in Bioelectrical Impedance Analysis (BIA) including different types of body part mass and weight, different types of body fat, basal metabolic rate, biological age, and body protein percentage were measured.

All experimental data were initially recorded in JSON format. The subsequent preprocessing involved several steps to prepare the data. For each trial, the selected point on the screen was represented in pixel coordinates. This point was first adjusted by subtracting the screen offset and then repositioned relative to the screen center. The magnitude, angle, and (x,y) coordinates of the shifted point were then computed. The magnitude and (x,y) coordinates were subsequently normalized using the outer radius of the half-circle. To ensure complete participant anonymity, each participant was assigned a unique identifier, these identifiers were randomly shuffled, and all personal information, including names and dates, was removed. Additionally, data such as height, weight, and waist circumference were rounded to further protect participant anonymity. Finally, all processed experimental data were aggregated into a single CSV file. The overall workflow of the experiment data processing is illustrated in Fig. 7.

Fig. 7.

Fig. 7

Open in a new tab

Overview of the experimental data preparation pipeline and anonymization procedure.

Data Records

The dataset is stored at Vibrotactile Dataset43 and consists of tabular-structured files in CSV format. Within the directory named Minimum Vibration Intensity Calibration, for each participant and session, the minimum perceivable vibration intensities are stored and separated into two files corresponding to the sessions. The Questionnaire directory contains questionnaire results. Data collected during the main experiment phase from each participant in the first and second sessions are stored in the Experiment Data file, with corrected radius and angles. Finally, the Participant Info file contains anthropometric and demographic data such as gender, weight, height, and waist circumference along with BIA results. The hierarchical structure of the dataset, including column headings for each file, is illustrated in Fig. 8.

Fig. 8.

Fig. 8

Open in a new tab

The hierarchical overview of the experimental dataset, showing directory organization and data tables corresponding to each file. The Session 1 and Session 2 files in the Minimum Vibration Calibration directory share the same table structure. The tables are shown in summarized format to illustrate the overall structure.

Technical Validation

This section concentrates on validation of dataset reproducibility, reliability, accuracy, and consistency. We conducted a comprehensive validation procedure, which will be described in the following subsections.

Hardware and logging

As described in Experimental setup subsection, all tactile stimuli were generated using five identical vibration motors. Since motors of this type exhibit slight differences in vibration frequency at the same applied voltage, we performed a pre-experimental characterization to ensure hardware consistency and select the most similar units. Sixteen candidate motors were tested in an acoustically isolated room of approximately 2 × 2 meters using a spectrum analyzer (Decibel X mobile application) across multiple voltage levels. From this pool, we selected five motors with the most similar average frequencies to ensure comparable performance across all vibration patterns. All experiments were conducted using these five selected motors and the same experimental setup (see Figure 3). In addition, all experimental data are stored in JSON format with a uniform numerical precision of nine decimal places.

Correction

By mistake, a small number of the points selected by participants were slightly outside the half-circle (about 2.94%), because the participants could select points outside the circle as well. So, we provided the corrected magnitude, angle, x, and y coordinate of the points, as well as the raw ones in the dataset to prevent any bias and make the dataset ready for further analysis. More specifically, some points had a radius higher than the half-circle outer radius, which has been corrected to the outer radius of the half-circle before normalization. Also, some selected points had the angle less(greater) than 0(180) degrees, which were saturated on 0(180) degrees. With this correction, all data points have a normalized magnitude between 0 and 1 and an angle between 0 and 180 degrees. It should be noted that for better illustrations in the paper we used the corrected data.

Consistency and validation

All records contained valid values, and no missing data were found in any fields. The dataset was collected from 40 healthy adult participants, each performing 255 trials across two sessions spaced 0 to 34 days apart; i.e., mean ( ± s.d.) interval of 12.3 ± 8.3. We first assessed temporal and spatial consistency. Similarity in the data distributions between the two sessions would demonstrate reproducibility. Figures 9 and 10 show comparable distributions of captured data across sessions, supporting the idea of temporal consistency. Additionally, the symmetry of x and angle distributions about 0 and 90 provides evidence of spatial consistency.

Fig. 9.

Fig. 9

Open in a new tab

Histograms of decision (time starting from the beginning of the trial to the first click on the graph), delay (the difference between decision and trial time), and trial time separated by session.

Fig. 10.

Fig. 10

Open in a new tab

Histograms of x, y, magnitude, and angle (the cartesian and polar coordinate components) separated by session. To normalize coordinate variables, they are divided by the half-circle radius as described in Data Acquisition.

Next, we examined demographic representativeness. The average adult male and female heights in Iran are approximately 161 cm to 175 cm respectively, with corresponding weights of 72 kg to 80 kg44. Fig. 1 shows that our participants’ height and weight distributions are consistent with these statistics.

Finally, we assessed time consistency. As shown in Fig. 11, the response times across trials fall within expected ranges, and the observed pattern-trial time > decision time > delay time-is also consistent with expectations. Violin plots in Figs. 12 and 13 indicate that similar patterns (e.g., symmetric patterns) exhibit similar distributions of response time and confidence level.

Fig. 11.

Fig. 11

Open in a new tab

Trend of decision time, delay time, and trial time.The very first trial response time values may not be proper due to different reasons, so it should be noted that all plots over trials are started from second trial.

Fig. 12.

Fig. 12

Open in a new tab

Variation of trial time among different patterns is shown via violin plot. Single or dual patterns are separated.

Fig. 13.

Fig. 13

Open in a new tab

Variation of confidence level among different patterns is shown via violin plot. Single or dual patterns are separated.

All these visualizations collectively demonstrate the dataset’s temporal and spatial consistency, its demographic representativeness, and the consistency of participant responses across different patterns.

Usage Notes

The files are stored in CSV format with UTF-8 encoding. Therefore, they can be accessed and processed as tables in MATLAB or opened as data frames using the pandas package in Python with the mentioned encoding.

Acknowledgements

This work was supported by the Iran National Science Foundation (INSF) under Grant 4020634. The authors would like to thank the University of Tehran for providing support with this study and the members of RAIIS lab, especially Diba Ravanshid for her support during dataset collection.

Author contributions

Mostafa Hamidifard: Conceptualization, Literature Review, Experiment Design, Data Collection, Writing, and Visualization. Samar Nikfarjad: Data Collection, Writing, Visualization, Analyzes, and Validation. Hosein Pirmohammadi: Experiment Design, Data Collection, Writing, and Visualization. Rezvan Nasiri: Supervision, Conceptualization, Visualization, Writing, Review, and Editing, and Proofreading. Majid Nili Ahmadabadi: Supervision, Conceptualization, Review, Editing, and Proofreading.

Data availability

The dataset and all source code are publicly available under a CC BY open license and can be accessed at the Figshare repository43.

Code availability

The source code for the Arduino hardware and the custom GUI used to collect data from experiment participants is available in the repository. The usage_example directory within the repository also provides a concise tutorial demonstrating how to access the dataset files and run a sample statistical test, together with detailed explanations to help users become familiar with the dataset workflow. In addition, all scripts used to generate the figures included in the Technical Validation section are provided in the Figshare repository43.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  • 1.Ziegler-Graham, K., MacKenzie, E. J. & Ephraim, P. L. Estimating the prevalence of limb loss in the United States: 2005 to 2050. Archives of physical medicine and rehabilitation89(3), 422–429 (2008). [DOI] [PubMed] [Google Scholar]
  • 2.Zelechowski, M., Valle, G. & Raspopovi, S. A computational model to design neural interfaces for lower-limb sensory neuroprostheses. Journal of neuroengineering and rehabilitation17(1), 24 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Borton, D., Micera, S., Millán, J. D. R. & Courtine, G. Personalized neuroprosthetics. Science translational medicine5(210), 210rv2–210rv2 (2013). [DOI] [PubMed] [Google Scholar]
  • 4.Dosen, S. et al. Multichannel electrotactile feedback with spatial and mixed coding for closed-loop control of grasping force in hand prostheses. IEEE Transactions on Neural Systems and Rehabilitation Engineering25(3), 183–195 (2016). [DOI] [PubMed] [Google Scholar]
  • 5.Ajoudani, A. et al. Exploring teleimpedance and tactile feedback for intuitive control of the pisa/iit softhand. IEEE transactions on haptics7, 203–215 (2014). [DOI] [PubMed] [Google Scholar]
  • 6.Engeberg, E. D. & Meek, S. Enhanced visual feedback for slip prevention with a prosthetic hand. Prosthetics and orthotics international36(4), 423–429 (2012). [DOI] [PubMed] [Google Scholar]
  • 7.Parr, J. V. V., Vine, S. J., Wilson, M. R., Harrison, N. R. & Wood, G. Visual attention, EEG alpha power and T7-Fz connectivity are implicated in prosthetic hand control and can be optimized through gaze training. Journal of neuroengineering and rehabilitation16, 1–20 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Eiber, C. D., Lovell, N. H. & Suaning, G. J. Attaining higher resolution visual prosthetics: a review of the factors and limitations. Journal of neural engineering10(1), 011002 (2013). [DOI] [PubMed] [Google Scholar]
  • 9.Stephens-Fripp, B., Alici, G. & Mutlu, R. A review of non-invasive sensory feedback methods for transradial prosthetic hands. IEEE Access6, 6878–6899 (2018). [Google Scholar]
  • 10.Stephens-Fripp, B., Mutlu, R. & Alici, G. Applying mechanical pressure and skin stretch simultaneously for sensory feedback in prosthetic hands. In 2018 7th IEEE International Conference on Biomedical Robotics and Biomechatronics (Biorob), 230–235 (2018).
  • 11.Szeto, A. Y. J. & Riso, R. R. Sensory feedback using electrical stimulation of the tactile sense. Rehabilitation engineering, 29–78 (2018).
  • 12.Risi, N., Shah, V., Mrotek, L. A., Casadio, M. & Scheidt, R. A. Supplemental vibrotactile feedback of real-time limb position enhances precision of goal-directed reaching. Journal of neurophysiology122(1), 22–38 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Ratschat, A. L., van Rooij, B. M., Luijten, J. & Marchal-Crespo, L. Evaluating tactile feedback in addition to kinesthetic feedback for haptic shape rendering: a pilot study. Frontiers in Robotics and AI11, 1298537 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Schätzle, S., Ende, T., Wüsthoff, T. & Preusche, C. Vibrotac: An ergonomic and versatile usable vibrotactile feedback device. In 19th International Symposium in Robot and Human Interactive Communication, 670–675 (2010).
  • 15.Buratti, S. et al. Effect of vibrotactile feedback on the control of the interaction force of a supernumerary robotic arm. Machines11(12), 1085 (2023). [Google Scholar]
  • 16.Liu, S., He, J., Jiang, S., Zhang, Z. & Lv, M. MatStick: Changing the material sensation of objects upon impact. Virtual Reality & Intelligent Hardware6(6), 486–501 (2024). [Google Scholar]
  • 17.Khaleq, A. A., More, Y., Skaufel, B. & Al Borno, M. The Effects of vibrotactile stimulation of the upper extremity on sensation and perception. bioRxiv, 2024–05 (2024).
  • 18.Sandykbayeva, D., Kappassov, Z. & Orazbayev, B. Vibrotouch: Active tactile sensor for contact detection and force sensing via vibrations. Sensors22(17), 6456 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Smith, A., Ward-Cherrier, B., Etoundi, A. & Pearson, M. J.Feeling the pressure: The influence of vibrotactile patterns on feedback perception. In 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 634–640 (2022).
  • 20.Hossain, M. T., Noghani, M. A., Sidaway, B. & Hejrati, B. Investigating the efficacy of a tactile feedback system to increase the gait speed of older adults. Human movement science90, 103103 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Seim, C. et al. Perception in hand-worn haptics: placement, simultaneous stimuli, and vibration motor comparisons. (2015).
  • 22.Marinelli, A. et al. A compact solution for vibrotactile proprioceptive feedback of wrist rotation and hand aperture. Journal of NeuroEngineering and Rehabilitation21(1), 142 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Liao, Z., Luces, J. V. S. & Hirata, Y. Human navigation using phantom tactile sensation based vibrotactile feedback. IEEE Robotics and Automation Letters5(4), 5732–5739 (2020). [Google Scholar]
  • 24.Salazar, J., Okabe, K. & Hirata, Y. Path-following guidance using phantom sensation based vibrotactile cues around the wrist. IEEE Robotics and Automation Letters3(3), 2485–2492 (2018). [Google Scholar]
  • 25.Samsel, Z., Gunia, A., Jäger, M. I. & Schöning, J. A comparison of vibrotactile patterns in an early warning system for obstacle detection using a haptic vest. Applied Ergonomics122, 104396 (2025). [DOI] [PubMed] [Google Scholar]
  • 26.Vargas, L., Huang, H., Zhu, Y. & Hu, X. Object recognition via evoked sensory feedback during control of a prosthetic hand. IEEE robotics and automation letters7(1), 207–214 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Thomas, N., Ung, G., McGarvey, C. & Brown, J. D. Comparison of vibrotactile and joint-torque feedback in a myoelectric upper-limb prosthesis. Journal of NeuroEngineering and Rehabilitation16, 1–18 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Ballardini, G. et al. Vibrotactile feedback for improving standing balance. Frontiers in bioengineering and biotechnology8, 94 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Baldi, T. L., Scheggi, S., Aggravi, M. & Prattichizzo, D. Haptic guidance in dynamic environments using optimal reciprocal collision avoidance. IEEE Robotics and Automation Letters3(1), 265–272 (2017). [Google Scholar]
  • 30.Plaisier, M. A., Sap, L. I. N. & Kappers, A. M. L. Perception of vibrotactile distance on the back. Scientific Reports10(1), 17876 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Pena, A. E., Rincon-Gonzalez, L., Abbas, J. J. & Jung, R. Effects of vibrotactile feedback and grasp interface compliance on perception and control of a sensorized myoelectric hand. PloS one14(1), e0210956 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Tan, H. Z., Choi, S., Lau, F. W. Y. & Abnousi, F. Methodology for maximizing information transmission of haptic devices: A survey. Proceedings of the IEEE108(6), 945–965 (2020). [Google Scholar]
  • 33.Kawai, A. & Tanaka, Y. Individual differences in skin vibration characteristics and vibrotactile sensitivity at fingertip. In 2022 IEEE Haptics Symposium (HAPTICS), 1–6 (2022).
  • 34.Cao, G. Jiang, J. Bollegala, D. & Luo, S. Learn from incomplete tactile data: Tactile representation learning with masked autoencoders. In 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 10800–10805 (2023).
  • 35.Devillard, A. W. M., Ramasamy, A., Cheng, X., Faux, D. & Burdet, E. Tactile, Audio, and Visual Dataset During Bare Finger Interaction with Textured Surfaces. Scientific Data12(1), 484 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Maeda, Y., Tanimoto, K., Sasayama, K. & Takao, H. Neural-Network-Based Tactile Perception System Using Ultrahigh-Resolution Tactile Sensor. IEEE Transactions on Haptics16(4), 504–510 (2023). [DOI] [PubMed] [Google Scholar]
  • 37.Voudouris, D., Georgiadis, P., Fiehler, K. & Arikan, B. E. Reliability of tactile perception and suppression measurements. bioRxiv, 2024–08 (2024). [DOI] [PubMed]
  • 38.Pastor, F., Lin-Yang, D., Gómez-de-Gabriel, J. M. & García-Cerezo, A. J. Dataset with tactile and kinesthetic information from a human forearm and its application to deep learning. Sensors22(22), 8752 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Chen, B., Feng, Y. & Wang, Q. Combining Vibrotactile Feedback with Volitional Myoelectric Control. Peripheral Nervous System-Machine Interfaces, (2018).
  • 40.Pollini, A. et al. Human-Machine Interaction Design in Adaptive Automation. Procedia Computer Science253, 1034–1044 (2025). [Google Scholar]
  • 41.Tang, J. et al. Human-centred design and fabrication of a wearable multimodal visual assistance system. Nature Machine Intelligence, 1–12 (2025).
  • 42.Janapati, R., Desai, U., Kulkarni, S. A. & Tayal, S. Human-machine interface technology advancements and applications. (2023).
  • 43.Hamidifard, M., Pirmohammadi, H., Nikfarjad, S., Nasiri, R. & Nili Ahmadabadi, M. Vibrotactile Dataset. Available at 10.6084/m9.figshare.29145557 (2025). [DOI] [PMC free article] [PubMed]
  • 44.WorldData.info. Average height and weight by country. https://www.worlddata.info/average-bodyheight.php?utm_source=chatgpt.com#by-area, 2025.

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The dataset and all source code are publicly available under a CC BY open license and can be accessed at the Figshare repository43.

The source code for the Arduino hardware and the custom GUI used to collect data from experiment participants is available in the repository. The usage_example directory within the repository also provides a concise tutorial demonstrating how to access the dataset files and run a sample statistical test, together with detailed explanations to help users become familiar with the dataset workflow. In addition, all scripts used to generate the figures included in the Technical Validation section are provided in the Figshare repository43.

Articles from Scientific Data are provided here courtesy of Nature Publishing Group

RESOURCES