Coming in handy: CeTI-Age — A comprehensive database of kinematic hand movements across the lifespan

Evelyn Muschter, Jonas Schulz, Máté Tömösközi, Leonie Herbst, Lena Küssel, Merve Sefunç, Stefan Holtzhausen, Stefanie Speidel, Frank H. P. Fitzek, Shu LiScientific 2023 · DOI

Abstract

The Tactile Internet aims to advance human-human and human-machine interactions that also utilize hand movements in real, digitized, and remote environments. Attention to elderly generations is necessary to make the Tactile Internet age inclusive. We present the first age-representative kinematic database consisting of various hand gesturing and grasping movements at individualized paces, thus capturing naturalistic movements. We make this comprehensive database of kinematic hand movements across the adult lifespan (CeTI-Age-Kinematic-Hand) publicly available to facilitate a deeper understanding of intra-individual-focusing especially on age-related differences-and inter-individual variability in hand kinematics. The core of the database contains participants' hand kinematics recorded with wearable resistive bend sensors, individual static 3D hand models, and all instructional videos used during the data acquisition. Sixty-three participants ranging from age 20 to 80 years performed six repetitions of 40 different naturalistic hand movements at individual paces. This unique database with data recorded from an adult lifespan sample can be used to advance machine-learning approaches in hand kinematic modeling and movement prediction for age-inclusive applications.

Introduction

The "Tactile Internet" envisions a future where human-human and human-machine interactions are enhanced by real-time haptic feedback in digital or remote environments. Central to this vision is the ability of machine learning (ML) models to accurately predict and interpret hand movements. However, a significant bottleneck exists: most current kinematic datasets are "age-biased," predominantly featuring young, right-handed adults. As motor control naturally changes with age—characterized by slower movement times and decreased dexterity—AI models trained on youthful data often fail to perform reliably for older populations.

Experimental Design and Methods

To address this gap, researchers developed a protocol focused on capturing naturalistic, self-paced movements across the adult lifespan. The study recruited 63 participants (33 female, 30 male) categorized into three age cohorts: young (20–39), middle-aged (40–59), and old (60–80).

Experimental setup and trial protocol timeline.

Participants wore a CyberGlove III equipped with 22 resistive bend sensors to record joint angles. Crucially, the protocol moved away from real-time mimicry. Instead, participants viewed an instructional video and then performed the gesture at their own pace, allowing the database to capture the intra-individual variability and natural "tempo" inherent to different age groups. This self-paced approach is essential for training robust ML models that can accommodate the diverse movement profiles found across the lifespan.

The CeTI-Age-Kinematic-Hand Database

The resulting database, CeTI-Age-Kinematic-Hand, is the first of its kind to span the entire adult lifespan. It includes raw and re-calibrated sensor data, individual 3D hand models generated via Artec EVA scanners, and the original stimulus videos. The data follows the BIDS (Brain Imaging Data Structure) standard to ensure broad research utility and easy integration into existing pipelines.

Taxonomy of the 40 hand movements in the CeTI-Age database.

The movements are organized into three categories based on complexity:

  • Category A: Basic finger flexions and extensions.
  • Category B: Hand postures (e.g., "thumbs up") and wrist movements.
  • Category C: Functional grasps and object interactions (e.g., tripod grasp, large diameter grasp).

By including 3D hand models, the database also allows researchers to account for anthropometric differences—how physical hand size and shape affect sensor readings—which is a common source of noise in wearable technology data.

Validation and Results

The researchers validated the data quality using four different ML classifiers (Random Forest, KNN, LDA, and SVC). Random Forest (RF) emerged as the top performer across all categories. Classification accuracy demonstrated a clear gradient based on movement complexity:

  • Category A (Basic movements): 94.4% accuracy.
  • Category B (Postures/Wrist): 84.4% accuracy.
  • Category C (Grasps/Functional): 79.3% accuracy.

The decrease in accuracy for Category C reflects the higher complexity of functional grasps, where overlapping joint movements make differentiation more challenging. Furthermore, the analysis confirmed that older participants exhibited significantly greater movement variability and longer durations, validating the study's emphasis on self-paced protocols.

Discussion and Future Outlook

While the CyberGlove III provides high-fidelity data, its "one-size-fits-most" design presents challenges for participants with smaller hands, potentially affecting sensor alignment at the fingertips. Additionally, wearing a glove naturally reduces tactile feedback, which some participants noted made object interactions slightly more difficult than with bare hands.

Despite these hurdles, the CeTI-Age database provides a vital foundation for developing age-inclusive technologies. Future research can leverage this data to develop age-adaptive algorithms capable of automatically adjusting sensitivity or prediction logic based on a user's age-related movement profile. Such advancements are critical to ensuring the Tactile Internet remains accessible and effective for all generations.

Reference: Muschter, E., et al. (2023). Coming in handy: CeTI-Age — A comprehensive database of kinematic hand movements across the lifespan. Scientific Data. https://doi.org/10.1038/s41597-023-02738-3

Summary updated

Abstract

The Tactile Internet aims to advance human-human and human-machine interactions that also utilize hand movements in real, digitized, and remote environments. Attention to elderly generations is necessary to make the Tactile Internet age inclusive. We present the first age-representative kinematic database consisting of various hand gesturing and grasping movements at individualized paces, thus capturing naturalistic movements. We make this comprehensive database of kinematic hand movements across the adult lifespan (CeTI-Age-Kinematic-Hand) publicly available to facilitate a deeper understanding of intra-individual-focusing especially on age-related differences-and inter-individual variability in hand kinematics. The core of the database contains participants' hand kinematics recorded with wearable resistive bend sensors, individual static 3D hand models, and all instructional videos used during the data acquisition. Sixty-three participants ranging from age 20 to 80 years performed six repetitions of 40 different naturalistic hand movements at individual paces. This unique database with data recorded from an adult lifespan sample can be used to advance machine-learning approaches in hand kinematic modeling and movement prediction for age-inclusive applications.
BibTeX
@inproceedings{muschter2023coming,
  author = {Evelyn Muschter and Jonas Schulz and Máté Tömösközi and Leonie Herbst and Lena Küssel and Merve Sefunç and Stefan Holtzhausen and Stefanie Speidel and Frank H. P. Fitzek and Shu Li},
  title = {{Coming in handy: CeTI-Age — A comprehensive database of kinematic hand movements across the lifespan}},
  booktitle = {Scientific},
  year = {2023},
  doi = {10.1038/s41597-023-02738-3},
}
Figure