MMH: A multimodal dataset of whole-body kinematics, bilateral ground reaction forces, and lower-limb surface electromyography signals during load lifting and lowering

The MMH dataset is a multimodal collection of whole-body kinematics, ground reaction forces, and lower-limb electromyography signals recorded during various lifting and lowering techniques, designed to support research in biomechanics, ergonomics, and musculoskeletal modeling.

The Gist

The Big Idea: The "Human Motion Movie" Dataset

Imagine you are trying to teach a robot how to pick up a heavy box without hurting its back. You can’t just give it a manual; you have to show it exactly how a human body moves, how much pressure the feet put on the floor, and how the muscles "fire" like tiny engines to get the job done.

The researchers who created the MMH dataset have essentially created a high-tech, multi-layered "instruction manual" for human movement. They recorded ten men performing various lifting tasks and captured every single detail of what happened inside and outside their bodies.

The Three Layers of the "Movie"

To understand this dataset, imagine a person lifting a box is like a high-performance car performing a stunt. The researchers recorded three different "camera angles" at once:

1. The Skeleton (Kinematics): The "Chassis" Movement
Think of this as a motion-capture movie. The researchers tracked exactly how the joints (knees, hips, spine) moved through space. It’s like watching a digital blueprint of a car’s frame to see if it bends or twists too much during a turn.

2. The Tires (Ground Reaction Forces): The "Traction"
When a car turns a corner, the tires push against the pavement. In this study, the researchers used special floor sensors to see exactly how much pressure the feet were applying to the ground. It’s like measuring how much "grip" and weight is being pushed into the floor to keep the body stable.

3. The Engine (sEMG): The "Electrical Spark"
Even if you see a car moving, you can't see the electricity jumping between the spark plugs to make the engine run. The researchers used sensors (sEMG) to listen to the "electrical whispers" of 12 different muscles. This tells us exactly which muscles are working hard and which ones are resting.

Why does this matter? (The "Safety Test")

Why go to all this trouble? Because right now, many tools used to prevent workplace injuries are just "educated guesses."

By combining these three layers—the movement, the pressure, and the muscle electricity—scientists can build much better computer simulations.

• The Goal: Instead of waiting for a worker to actually hurt their back, we can use this data to create "Digital Twins." We can test a new way of lifting in a computer program first, using this real-world data, to see if it’s safe.

In short: This dataset is a master toolkit that helps scientists design safer jobs, better ergonomic tools, and smarter robots by giving them a perfect, multi-dimensional look at how humans move.

Technical Summary

Technical Summary: MMH Multimodal Dataset

1. Problem Statement

In the fields of biomechanics and ergonomics, there is a critical need for high-fidelity, synchronized, and multimodal datasets to accurately assess musculoskeletal injury risks during manual material handling (lifting and lowering). Existing datasets often lack the simultaneous integration of whole-body kinematics, ground reaction forces (GRF), and muscle activation patterns (sEMG). Without this multi-layered data, it is difficult to develop robust optimization-based musculoskeletal models or to validate the predictions these models make regarding internal muscle forces and joint loading.

2. Methodology

The researchers developed the MMH dataset through a controlled laboratory study involving the following parameters:

• Participants: Ten healthy, young, normal-weight male adults.
• Task Protocol: Participants performed a total of 72 trials each, involving the lifting and lowering of a 2 kg load.
• Experimental Variables:
  • Lifting Techniques: Three distinct postures were tested: stoop, semi-squat, and full-squat.
  • Handling Modes: Both one-handed and two-handed lifting/lowering.
  • Spatial Variability: Multiple initial and final load locations to simulate real-world movement variability.
• Data Acquisition (Multimodal Integration):
  • Kinematics: Whole-body motion capture to track joint angles and body segments.
  • Kinetics: Three-dimensional (3D) ground reaction forces and two-dimensional (2D) Center of Pressure (CoP) measurements for both feet.
  • Neuromuscular Activity: Surface electromyography (sEMG) recorded from twelve lower-limb muscles (six muscles per leg) to capture muscle activation patterns.

3. Key Contributions

• Multimodal Synchronization: The primary contribution is the simultaneous recording of kinematics, kinetics, and sEMG, providing a holistic view of the human movement system during a functional task.
• Comprehensive Task Coverage: By varying lifting techniques (stoop vs. squat) and hand usage, the dataset captures a wide range of biomechanical strategies and loading conditions.
• High-Resolution Muscle Data: The inclusion of 12 lower-limb sEMG channels provides a granular look at how the legs stabilize the body and manage loads during different lifting postures.

4. Results (Dataset Characteristics)

While the abstract focuses on the nature of the dataset rather than statistical findings, the "results" of the study are the creation of a high-dimensional, synchronized data repository. The dataset provides:

• Rich input for ergonomic risk assessment tools.
• A foundation for optimization-based musculoskeletal modeling.
• A ground-truth reference for EMG-assisted modeling, allowing researchers to compare predicted muscle forces against actual electrical activity.

5. Significance

The MMH dataset serves as a vital resource for several scientific domains:

• Biomechanics & Musculoskeletal Modeling: It enables the development and validation of more accurate models that predict internal joint loads and muscle forces.
• Ergonomics: It provides the empirical data necessary to refine risk assessment protocols (e.g., improving the accuracy of tools like the NIOSH lifting equation or similar computational models).
• Human Movement Research: It offers a standardized benchmark for researchers to test new algorithms in motion analysis, machine learning for movement recognition, and injury prevention strategies.