AUST_gait

AUST-gait dataset and introduction.

Introduction

The AUST-gait dataset, collected by Anhui University of Science and Technology, is specifically designed for gait analysis, supporting time-series prediction and classification tasks. The dataset is collected through a wearable hip exoskeleton robot and contains various sensor data, suitable for research in missing value processing and gait recognition.

Equipment and Sensors

• Equipment: Hip exoskeleton robot
• Sensors:
  • 6-axis IMU sensor: Mounted on the back, measuring three-axis acceleration and angular velocity
  • Motor shaft encoder and 12-bit output shaft encoder: Measuring angles and angular velocities of left and right hip joints

Experimental Procedure

Figure 1: On-site photos of data collection

65 healthy adults were recruited (age 23±2 years, height 1.74±0.05 meters, weight 80±8.6 kg), excluding specific conditions. The experiment received ethics committee approval, and participants signed informed consent forms.

Participants performed four walking modes under supervision:

1. Flat Ground Walking (FG): Walking back and forth three times on a 6-meter long, 1.2-meter wide track
2. Stair Ascent (SA): Climbing a 10-step staircase (15cm per step), repeated 3 times
3. Stair Descent (SD): Descending a 10-step staircase, repeated 3 times
4. Inclined Treadmill Walking (IT): Walking on a treadmill inclined at 20° for 45 seconds
   Additionally, fall experiments were conducted to simulate extreme conditions.

Data Description

Figure 2: Illustration of dataset features

Feature Format: Each time point contains 13 features, recorded in Excel cells with20ms intervals.

Features Include:

• Left hip angle
• Right hip angle
• Left hip angular velocity
• Right hip angular velocity
• Back IMU X-axis acceleration
• Back IMU Y-axis acceleration
• Back IMU Z-axis acceleration
• Back IMU X-axis angular velocity
• Back IMU Y-axis angular velocity
• Back IMU Z-axis angular velocity
• Back IMU X-axis Euler angle
• Back IMU Y-axis Euler angle
• Back IMU Z-axis Euler angle

Labels (Walking categories in the last column):

• 0: Flat ground walking
• 1: Stair ascent
• 2: Stair descent
• 3: Inclined treadmill walking
• 4: Fall experiment

Dataset Structure:

• Training set: 75%
• Test set: 25%
• Fall experiment data: Can be combined with other datasets

Tasks:

• Prediction task: Learn from first 10 steps to predict the 11th step
• Recognition task: Classify different walking modes

Missing Value Processing: Users can employ various algorithms to simulate missing values, and the dataset supports experiments with different missing rates.

License：This dataset is released under open data sharing principles. Users can freely use, modify, and distribute the data with proper attribution.

If you have any questions, feel free to raise an issue

---

This README is formatted in Markdown and can be directly uploaded to a GitHub repository.