---
license: unknown
---

## Audio Dataset

> **This raw audio dataset was prepared using my notebook
> *“Building an Audio Classification Pipeline with DL,”* available on my profile.**
> It forms the foundation for all subsequent preprocessing and spectrogram generation.

---

### **Dataset Summary**

| Property                  | Description             |
| ------------------------- | ----------------------- |
| **Number of Classes**     | 13 categories           |
| **Audio Files per Class** | ~40 raw recordings      |
| **Duration**              | ~5 seconds each         |
| **Channels**              | Mono (after processing) |
| **Sampling Rate (final)** | 16 kHz                  |

---

### **Processing Overview**

The raw audio underwent a compact but essential pipeline:

1. **Data Loading & Inspection**
   Imported all recordings and validated metadata (duration, sample rate, SNR).

2. **Cleaning & Normalization**

   * Removed corrupted/silent files
   * Normalized amplitude
   * Trimmed leading/trailing silence
   * Applied noise reduction

3. **Standardization**

   * Converted to mono
   * Resampled to **16,000 Hz**
   * Forced each clip to a **uniform 5-second length**

4. **Augmentation (for balance & variability)**

   * Pitch shift
   * Time stretch
   * Noise injection
   * Time shift

---

### **Final Technical Description**

> **“The raw dataset consists of 13 audio classes with approximately 40 five-second recordings each. All clips were cleaned, normalized, noise-reduced, resampled, and standardized through a custom pipeline implemented in the notebook *‘Building an Audio Classification Pipeline with DL.’* This processed audio served as the basis for generating the Mel-spectrogram dataset used for model training.”**