maxF6YsK/non_speech_human_sounds

Non-Speech Human Sounds Dataset

A comprehensive collection of human-produced sounds that are not speech, recorded in different indoor environments. The dataset includes a wide range of bodily sounds, emotional expressions, and physical actions, each recorded in both quiet and noisy indoor settings.

Dataset Description

This dataset contains 92 audio recordings of various non-speech human sounds, with each sound type recorded in two different environmental conditions. All sounds are performed by male subjects and include detailed environmental context and background noise information.

Key Features

• Total Samples: 92 recordings
• Audio Format: M4A
• Gender: Male
• Environments: 2 types (indoor_quiet, indoor_noisy)
• Sound Categories: 46 unique sound types
• Background Variations: Multiple noise types in noisy environment

Sound Categories

1. Emotional Expressions:

   • Laughing, Crying, Sobbing
   • Screaming, Shouting
   • Sighing, Moaning, Groaning

2. Bodily Functions:

   • Breathing, Coughing, Sneezing
   • Hiccuping, Snoring, Yawning
   • Burping, Retching, Spitting

3. Physical Actions:

   • Clapping hands, Finger snapping
   • Foot stomping, Foot dragging
   • Slapping, Chest beating

4. Mouth Sounds:

   • Whistling, Humming
   • Tongue clicking, Lips smacking
   • Shushing

5. Other Sounds:

   • Hair brushing, Hand rubbing
   • Joint cracking, Nail tapping
   • Head scratching, Eye rubbing

Data Format

Each entry contains:

{
  "audioFile": "string",
  "soundType": "string",
  "environment": "string",
  "backgroundNoiseType": "string",
  "gender": "string"
}

Environmental Conditions

Indoor Quiet

• Clean recordings with no background noise
• Controlled environment
• Optimal recording conditions

Indoor Noisy

Background noise types include:

• Keyboard typing
• Speech
• Appliances
• Mixed ambient sounds
• Phone sounds

Usage

This dataset is particularly useful for:

• Sound recognition systems
• Human activity recognition
• Ambient sound detection
• Healthcare monitoring applications
• Emotion detection systems
• Background noise filtering

Loading the Dataset

from datasets import load_dataset

dataset = load_dataset("maxF6YsK/non_speech_human_sounds")

Dataset Statistics

Distribution

• 46 unique sound types
• Each sound recorded in 2 environments
• Total of 92 recordings
• 100% male performers
• Even distribution between quiet and noisy environments

Sound Categories Distribution

• Emotional expressions: ~20%
• Bodily functions: ~30%
• Physical actions: ~25%
• Mouth sounds: ~15%
• Other sounds: ~10%

Applications

1. Healthcare:

   • Sleep apnea detection
   • Respiratory condition monitoring
   • Cough detection systems

2. Security:

   • Anomaly detection
   • Emergency sound detection
   • Environmental monitoring

3. Human-Computer Interaction:

   • Gesture recognition
   • Non-verbal command systems
   • Ambient awareness

4. Audio Processing:

   • Noise cancellation
   • Sound classification
   • Environmental sound separation