https://github.com/nourine-nadir/speech_processing

This repository explores speech processing techniques like noise cancellation and speech segmentation through Python code.(Speech recognition soon)
https://github.com/nourine-nadir/speech_processing

artificial-intelligence noise-cancellation speech-processing speech-segmentation

Last synced: about 2 months ago
JSON representation

This repository explores speech processing techniques like noise cancellation and speech segmentation through Python code.(Speech recognition soon)

• Host: GitHub
• URL: https://github.com/nourine-nadir/speech_processing
• Owner: Nourine-Nadir
• Created: 2024-04-20T17:46:29.000Z (9 months ago)
• Default Branch: master
• Last Pushed: 2024-09-09T17:14:12.000Z (4 months ago)
• Last Synced: 2024-09-09T21:28:10.456Z (4 months ago)
• Topics: artificial-intelligence, noise-cancellation, speech-processing, speech-segmentation
• Language: Jupyter Notebook
• Homepage:
• Size: 8.39 MB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: ReadMe.md

Awesome Lists containing this project

README

        
## Speech Processing

This repository contains code for various speech processing tasks.

**Subfolders:**

* **speech_preprocessing:**  Contains functions for preprocessing speech audio, including:

    * Spectrogram calculation

    * Noise cancellation (using spectral subtraction)

* **wavs:** Stores audio files used in the examples.

**Files:**

* **Noise_cancelling.py:** Implements functions for calculating spectrograms, noise cancellation, and audio reconstruction.

* **speech_segmentation.py:** Implements a basic speech segmentation engine using energy-based detection.

**Noise Cancellation**

The `Noise_cancelling.py` file provides functions for:

* Spectrogram calculation (`spectrogram` and `spectrogram2wav`) for converting between time and frequency domains.

* Noise cancellation (`NoiseCancelling`) using spectral subtraction to remove noise from an audio signal.

**Speech Segmentation**

The `speech_segmentation.py` file demonstrates a simple speech segmentation engine based on energy levels. The `Engine` class implements methods for:

* Calculating energy of the audio signal.

* Updating a dynamic threshold based on background noise.

* Detecting speech segments based on energy exceeding the threshold for a minimum duration.

**Running the Examples**

This repository requires libraries like `numpy`, `matplotlib.pyplot`, `scipy.io.wavfile`, and `IPython.display` (for audio playback). Make sure you have them installed before running the scripts.

The provided code snippets showcase the functionalities. You can modify them to suit your specific needs.

**Future Work**

This repository serves as a starting point for speech processing tasks. Here are some potential areas for future development:

* Implementing more sophisticated noise cancellation techniques.

* Utilizing feature extraction methods for speaker recognition or speech classification.

* Refining the speech segmentation engine for better accuracy.

Feel free to explore, modify, and contribute to this repository!