README.md

metadata

license: mit
language:
  - en
tags:
  - gesture-recognition
  - sensor-data
  - flex-sensors
  - accelerometer
size_categories:
  - n<1K
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
dataset_info:
  features:
    - name: label
      dtype: string
    - name: batch
      list:
        list: int64
  splits:
    - name: train
      num_examples: 180
    - name: test
      num_examples: 48

Gesture Recognition Dataset

Dataset Structure

• Labels: ['Good', 'Null', 'Thirsty', 'Bad', 'Me', 'Hungry']
• Format: Each record contains a 'label' and a 'batch' field
• Batch Size: 30 rows per batch (30 time steps)
• Features: 15 columns per row
• Selection Method: cosine_similarity - Files selected based on similarity to majority pattern

Column Information

Each row in a batch contains 15 values in this order:

1. Timestamp - Timestamp
2. F1 - Flex sensor 1
3. F2 - Flex sensor 2
4. F3 - Flex sensor 3
5. F4 - Flex sensor 4
6. F5 - Flex sensor 5
7. Acc_Fin_x - Accelerometer Fin x axis
8. Acc_Fin_y - Accelerometer Fin y axis
9. Acc_Fin_z - Accelerometer Fin z axis
10. Acc_Palm_x - Accelerometer Palm x axis
11. Acc_Palm_y - Accelerometer Palm y axis
12. Acc_Palm_z - Accelerometer Palm z axis
13. Acc_Arm_x - Accelerometer Arm x axis
14. Acc_Arm_y - Accelerometer Arm y axis
15. Acc_Arm_z - Accelerometer Arm z axis

Data Format

{
    'label': 'gesture_name',  # One of: ['Good', 'Null', 'Thirsty', 'Bad', 'Me', 'Hungry']
    'batch': [
        [Timestamp,F1, F2, F3, F4, F5, Acc_Fin_x, Acc_Fin_y, Acc_Fin_z, Acc_Palm_x, Acc_Palm_y, Acc_Palm_z, Acc_Arm_x, Acc_Arm_y, Acc_Arm_z],  # Row 1
        [Timestamp,F1, F2, F3, F4, F5, Acc_Fin_x, Acc_Fin_y, Acc_Fin_z, Acc_Palm_x, Acc_Palm_y, Acc_Palm_z, Acc_Arm_x, Acc_Arm_y, Acc_Arm_z],  # Row 2
        ...  # 30 rows total
    ]
}

Sensors

• F1-F5: Flex sensors measuring finger bend (5 sensors)
• Acc_Fin: Accelerometer on finger (x, y, z axes)
• Acc_Palm: Accelerometer on palm (x, y, z axes)
• Acc_Arm: Accelerometer on arm (x, y, z axes)

Data Quality

Files were selected using cosine_similarity to ensure the most representative samples for each gesture class.