hzyhpp123/audio_noise

Audio Noise Recognition Dataset

Overview

This dataset is designed for evaluating AI models' ability to identify and classify various types of noise in audio recordings. It contains 28 carefully curated audio samples covering diverse noise scenarios, including both stationary and non-stationary noise types.

Dataset Structure

audio_noise/
├── test/
│   ├── audio/
│   │   ├── NS_001.wav
│   │   ├── NS_002.wav
│   │   └── ...
│   │   └── NS_028.wav
│   └── data.jsonl
└── README.md

Data Format

Each sample in data.jsonl contains:

• audio: Audio file path (relative to test/audio/)
• task: Task information
  • name: Task description ("识别音频噪声" - Identify audio noise)
  • label: Noise category ("平稳噪声" for stationary noise or "非平稳噪声" for non-stationary noise)
  • answer: Detailed noise description in Chinese

Noise Categories

Stationary Noise (平稳噪声) - 6 samples

• White noise (白噪声)
• Air conditioner noise (空调噪声)
• Electric fan noise (电风扇噪声)
• Fan noise (风扇噪声)
• Wind noise (风声)

Non-Stationary Noise (非平稳噪声) - 22 samples

Human-related:

• Background voices (背景人声)
• Baby crying (婴儿啼哭声)
• Station announcement (车站大厅报幕声)
• Open space background voices (空旷环境的背景人声)

Music-related:

• Rap music BGM (说唱音乐BGM)
• Game music (游戏音乐)
• Electronic music (电子音乐)

Environmental:

• Subway noise (地铁噪声)
• Street noise (街道噪声)
• Station noise (车站噪声)
• Rain sound (下雨声)
• Wind noise (风声)

Animal:

• Cat meowing (猫叫声)
• Dog barking (狗叫声)

Mechanical/Electronic:

• Alarm sound (警报声)
• Car horn (汽车喇叭声)
• Keyboard noise (键盘噪声)
• Mouse noise (鼠标噪声)
• Electric buzzing (滋滋噪声)
• Plastic rustling (捏塑料包装纸的噪声)
• Nose blowing (擤鼻涕声)
• Popping/Breath noise (喷麦噪声)
• Knocking noise (敲击噪声)
• Metal collision (金属碰撞声)

Dataset Statistics

• Total samples: 28
• Stationary noise: 6 samples
• Non-stationary noise: 22 samples
• Audio format: WAV files
• Sample rate: 24,000 Hz
• Duration: ≤30 seconds per sample

Usage

This dataset can be used to evaluate audio understanding models' ability to:

1. Detect presence of noise in audio recordings
2. Classify noise types (stationary vs. non-stationary)
3. Identify specific noise characteristics
4. Provide natural language descriptions of noise content

Sample Entry

{
    "audio": "audio/NS_001.wav",
    "task": {
        "name": "识别音频噪声",
        "label": "平稳噪声",
        "answer": "我觉得这段音频存在白噪声。"
    }
}

Task

• Objective: Identify and describe audio noise
• Input: Audio file
• Output: Natural language description of detected noise
• Labels: Binary classification (stationary/non-stationary) + detailed description