Speaker Identification System using Machine Learning

roject Overview

This project is a machine learning-based speaker identification system that identifies a speaker from an audio recording.

The system is trained on voice samples from 15 university students. Due to privacy constraints, the dataset is not included in this repository.

Privacy Notice

The original audio dataset is not publicly available due to privacy and ethical considerations.

This repository contains:

• Source code
• Data preprocessing pipeline
• Feature extraction methods
• Model training and prediction scripts

Objective

To classify speakers based on voice characteristics extracted from audio signals using machine learning techniques.

Methodology

Audio Input → Data Augmentation → Preprocessing → Feature Extraction (MFCC) → Model Training (SVM) → Prediction

Dataset Description

• 15 speakers
• 3 recordings per speaker
• Augmented using signal processing techniques

Data Augmentation Methods:

• Pitch shifting
• Time stretching
• Amplitude scaling

Preprocessing Steps

• Noise reduction using pre-emphasis filtering
• Fixed-length trimming/padding (3 seconds)
• Signal normalization

Feature Extraction

• Method: MFCC (Mel-Frequency Cepstral Coefficients)
• Features: 6 coefficients
• Aggregation: Mean over time frames
• Library: librosa

Model

• Algorithm: Support Vector Machine (SVM)
• Kernel: Linear
• Library: scikit-learn

Results

Model Accuracy: 96.77%

The model shows strong performance in distinguishing between 15 speakers.

How to Run

Install dependencies

pip install -r requirements.txt

Train model

python src/train.py

Predict speaker

python src/predict.py path/to/audio.wav

Run interactive mode

python main.py

Project Structure

speaker-identification-project/

src/

• augmentation.py
• preprocessing.py
• feature_extraction.py
• train.py
• predict.py

notebooks/

• data_augmentation_demo.ipynb
• model_training.ipynb

models/

• speaker_model.pkl (optional / not included due to privacy)

requirements.txt README.md .gitignore main.py

👥 Authors

• Fateme Khosravi
• Fateme Hajizadeh
• Ali Khosrozadeh