Skip to content

README.md

Latest commit

 

History

History
81 lines (53 loc) · 2.63 KB

README.md

File metadata and controls

81 lines (53 loc) · 2.63 KB

Audio Classification Model

This repository contains a project to classify audio files as either real or faked using MFCC (Mel-frequency cepstral coefficients) features and a neural network model built with TensorFlow and Keras.

📂 Dataset

The dataset should be organized into two directories:

  • dataset/real: Contains real audio samples in .wav format.
  • dataset/fake: Contains faked audio samples in .wav format.

🛠️ Setup

Prerequisites

You can install the required packages using pip in env:

pip install numpy tensorflow keras librosa scikit-learn

Project Structure

├── dataset
│   ├── real
│   │   └── *.wav
│   └── fake
│       └── *.wav
├── model.py
├── model.h5
└── README.md

🚀 Getting Started

1. Data Preparation

Ensure your dataset is correctly placed in the dataset/real and dataset/fake directories. Each .wav file will be processed to extract MFCC features.

2. Training the Model

Run the model.py script to train the model:

python model.py

This script performs the following steps:

  1. Extract MFCC Features: From the audio files in the dataset.
  2. Create Dataset: Combine the features and labels for real and faked audio.
  3. Train the Model: Train a neural network model on the extracted features.
  4. Save the Model: The trained model is saved as model.h5.

3. Analyzing New Audio Files

After training the model, you can use it to classify new audio files:

When prompted, enter the path of the .wav file you want to analyze. The model will predict whether the audio is real or faked.

📊 Evaluation

The model's performance is evaluated using accuracy and a confusion matrix, which are printed during training:

Accuracy: 0.95

🔧 Functions Overview

  • extract_mfcc_features(audio_path, n_mfcc=13, n_fft=2048, hop_length=512): Extracts MFCC features from the given audio file.
  • create_dataset(directory, label): Creates a dataset by extracting MFCC features from all .wav files in the given directory.
  • build_model(input_shape): Builds a neural network model with the specified input shape.
  • train_model(X, y): Trains the neural network model on the provided features and labels.
  • analyze_audio(input_audio_path): Analyzes a given audio file to predict whether it is real or faked.

📝 Notes

  • Ensure the audio files are in .wav format and 22500hH.
  • Adjust the parameters such as n_mfcc, n_fft, and hop_length if necessary to better suit your data.