MATLAB-Simulated Dataset for Automatic Modulation Classification in Wireless Fading Channels
Description
This dataset contains feature-extracted numerical representations of five digital modulation schemes—BPSK, QPSK, 16-QAM, 64-QAM, and 256-QAM—generated under Rayleigh and Rician fading channels at different sampling frequencies (1 MHz, 10 MHz, 100 MHz, 500 MHz, and 1 GHz). The dataset consists of 10 CSV files, each corresponding to a unique combination of channel conditions and sampling rates. A separate folder contains MATLAB scripts for dataset generation, ensuring reproducibility and enabling customization according to user requirements. The modulation signals were generated and transmitted through Rayleigh and Rician fading channels to simulate real-world wireless conditions. Each received signal underwent feature extraction through three primary approaches: 1. Signal Processing-Based Features – Extracted from both time and frequency domains, capturing statistical properties (mean, variance, skewness), spectral attributes (FFT-based spectral centroid, flatness, phase-based metrics), and signal measures (RSCR, PAPR, SCF). These features provide a comprehensive representation of modulation characteristics for classification and analysis. 2. Spectrogram-Based Features – Derived from the time-frequency representation of signals using STFT. Statistical measures (entropy, variance, skewness, kurtosis), spectral attributes (centroid, spread, flatness), and amplitude-based features (contrast, homogeneity, range) were extracted to quantify spectral energy distribution and signal consistency. 3. Image Processing-Based Features – Utilizing Binary Robust Invariant Scalable Keypoints (BRISK), Maximally Stable Extremal Regions (MSER), and Gray-Level Co-occurrence Matrix (GLCM) techniques to analyze spectrogram-based signal representations. Each dataset file contains 202 extracted features, providing a high-dimensional numerical representation of modulation signals. The dataset holds significant value across multiple domains, particularly in machine learning and deep learning, where it facilitates modulation classification using ML/DL models to develop robust classifiers for wireless communication systems. In wireless signal processing, researchers can utilize the extracted features for channel estimation, adaptive modulation, and interference analysis. Additionally, the dataset is highly relevant for cognitive radio and LTE/5G research, supporting spectrum sensing, signal recognition, and intelligent radio technologies. It also serves as a benchmarking resource for testing feature selection methods, dimensionality reduction techniques, and classification algorithms, making it a valuable asset for both academic and industrial applications.