Fourier-Classification 1.0.0

A package for classifying 1D signals using Fourier Series and Machine Learning

Fourier Series Classification

A package for classifying 1D signals using Fourier Series and Machine Learning.

Overview

This package implements the methodology described in the paper "Using Fourier Series and Machine Learning to Classify 1D-Signals". It provides tools for:

• Generating various types of 1D signals (Box, Sawtooth, Exponential, Sinusoidal, Gaussian)
• Computing Fourier series and coefficients
• Detecting jumps in signals using concentration factors
• Training neural network models for signal classification
• Visualizing signals, Fourier coefficients, and classification results

Installation

Requirements

• Python 3.6 or higher
• NumPy
• TensorFlow 2.4 or higher
• Matplotlib
• Plotly
• Pandas
• scikit-learn

Install from source

git clone https://github.com/abbass12/FourierSeriesClassification.git
cd FourierSeriesClassification
pip install -e .

Usage

Basic Example

import numpy as np
from fourier_classification.signals import box_signal
from fourier_classification.fourier import fourier_series
from fourier_classification.visualization import plot_signal_and_fourier

# Create domain
x = np.linspace(-np.pi, np.pi, 1500)

# Generate a box signal
signal = box_signal(x, a=2, b=5, normalized=True)

# Compute Fourier coefficients
fourier_coeffs = box_signal(x, a=2, b=5, normalized=True, fourier=True, n_modes=40)

# Plot signal and Fourier coefficients
fig = plot_signal_and_fourier(x, signal, fourier_coeffs, title="Box Signal")
fig.savefig("box_signal.png")

Signal Classification

import numpy as np
from fourier_classification.utils import create_domain, prepare_dataset
from fourier_classification.models import (
    create_feed_forward_model, 
    train_model, 
    evaluate_model,
    prepare_data_for_model_b
)

# Create domain
domain = create_domain(start=-np.pi, end=np.pi, num_points=1500)

# Prepare dataset
signal_types = ['Box', 'Saw', 'Exp', 'Sin', 'Gaus']
signals, labels = prepare_dataset(
    signal_types, 
    num_per_type=100, 
    domain=domain, 
    fourier=True, 
    n_modes=40
)

# Prepare data for Model B (Fourier coefficients)
x_train, x_test, y_train, y_test = prepare_data_for_model_b(signals, labels)

# Create and train model
model = create_feed_forward_model(input_shape=(x_train.shape[1],))
model, history = train_model(model, x_train, y_train, epochs=100)

# Evaluate model
results = evaluate_model(model, x_test, y_test, class_names=signal_types)
print(f"Test accuracy: {results['accuracy']:.4f}")

Module Structure

• signals.py: Functions for generating various types of 1D signals
• fourier.py: Functions for Fourier series calculations and transformations
• operations.py: Functions for signal operations (noise, jumps)
• models.py: Neural network models for signal classification
• visualization.py: Functions for visualizing signals and results
• utils.py: Utility functions for data handling and preprocessing

License

This project is licensed under the MIT License - see the LICENSE file for details.

Citation

If you use this code in your research, please cite:

@article{srour2023fourier,
  title={Using Fourier Series and Machine Learning to Classify 1D-Signals},
  author={Srour, Abbass},
  journal={},
  year={2023}
}