Skip to content

README.md

Latest commit

 

History

History
101 lines (77 loc) · 3.67 KB

README.md

File metadata and controls

101 lines (77 loc) · 3.67 KB

Speaker Identification and Gender Classification

This repository contains the implementation of a Machine Learning pipeline for Speaker Identification and Gender Classification using audio features.

🚀 Project Overview

The goal of this project is to develop robust models that can:

  1. Classify Gender: Determine whether a speaker is male or female.
  2. Identify Speakers: Distinguish between different speakers based on their voice characteristics.

The project utilizes comprehensive audio signal processing techniques and state-of-the-art machine learning algorithms, ranging from classical classifiers (SVM, KNN, XGBoost) to Neural Networks.

📂 Repository Structure

├── data/                   # Data directory (raw and processed)
├── notebooks/              # Jupyter notebooks for experimentation
├── scripts/                # Executable scripts for training and evaluation
├── src/                    # Source code for the project
│   ├── data/               # Data loading and cleaning
│   ├── features/           # Audio processing and feature extraction
│   ├── models/             # Model definitions (Sklearn, Keras, etc.)
│   └── visualization/      # Plotting and evaluation utilities
├── requirements.txt        # Project dependencies
├── setup.py                # Package setup script
└── README.md               # Project documentation

🛠️ Installation

  1. Clone the repository:

    git clone https://github.com/your-username/Speaker-ID-Gender-Classification.git
cd Speaker-ID-Gender-Classification

  2. Create a virtual environment (recommended):

    python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

  3. Install dependencies:

    pip install -r requirements.txt
pip install -e .

📊 Methodology

Feature Extraction

We extract a rich set of audio features including:

  • Spectral Features: MFCC, Spectral Centroid, Bandwidth, Contrast, Roll-off.
  • Temporal Features: Zero Crossing Rate, RMS Energy.
  • Prosodic Features: Fundamental Frequency (F0), Jitter, Shimmer.

Processing Pipeline

  1. Silence Removal: Trimming silence using spectral centroid based windowing.
  2. Noise Reduction: Spectral subtraction to enhance signal quality.
  3. Filtering: Bandpass filter (80Hz - 5000Hz) to isolate human speech frequencies.
  4. Resampling: Standardizing sample rate to 44.1kHz.

Models

We experiment with multiple architectures:

  • Support Vector Machine (SVM): RBF kernel for non-linear separation.
  • K-Nearest Neighbors (KNN): Baseline distance-based classifier.
  • XGBoost / AdaBoost: Ensemble methods for improved robustness.
  • Multi-Layer Perceptron (MLP): Deep learning approach using Keras/TensorFlow.

🏃‍♂️ Usage

1. Download Data

The dataset is hosted on Google Drive. Run the setup script to download and structure the data:

python src/data/download.py

2. Train Gender Classifier

To train and evaluate the gender classification model:

python scripts/train_gender.py --model svm

Available models: svm, knn, xgboost, adaboost.

📈 Results

Model Accuracy Precision Recall
SVM 0.98 0.98 0.98
XGBoost 0.97 0.97 0.97
KNN 0.96 0.96 0.95

(Note: Results may vary slightly based on random seed and data split)

👥 Contributors

  • Mostafa Kermani Nia - Lead Developer & Researcher

📄 License

This project is licensed under the MIT License.