Unveiling Digital Drugs Through Artificial Intelligence: An Audio Feature Analysis Approach

Read full text Summarise with VIRAT
Authors: Karam Muhammed Mahdi Salih, Abdulmajeed Sulaiman, Nora Rashid Najem, Abdulrahman Mohammed
Cited by: 0
References: 24

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Binaural beat audio is common nowadays, and some clips are promoted as "digital drugs," creating a need for simple, reliable detection for moderation and forensics.This study builds a small, reproducible system that separates such audio from normal content using compact tabular features and standard classifiers.The dataset includes 700 clips from public sources all audio clips are standardized to 44.1 kHz stereo and -23 loudness units relative to full scale LUFS, with exact and near-duplicate items removed before any split.After removing Labels and tags a singleindex threshold acts as a cross-check.Features include band powers and ratios (δ-γ), basic spectral and time statistics (RMS, centroid, roll-off, Zero-Crossing Rate (ZCR)), and one scalar index Fs that summarizes slow vs fast band energy.Models are multiple statistical fingerprints, Random Forest, XGBoost, LightGBM, a convolutional neural network (CNN), and a Mel-frequency cepstral coefficient support vector machine (MFCC-SVM) baseline.Evaluation uses stratified 5-fold cross-validation with creator-level grouping; we report Accuracy, ROC-AUC, macro-F1 as mean±SD across folds.The best model reaches ROC-AUC ≈ 0.95 with strong Accuracy and macro-F1, and an ablation shows marginal changes when forensic slow-fast energy index Fs is added.The pipeline is lightweight and easy to reproduce; limits include operational labels rather than clinical labels, and next steps include testing a spectrogram-based CNN under the same split policy.

Fields of Study & Subjects

Music and Audio Processing Speech Recognition and Synthesis Wireless Signal Modulation Classification

Publication details

Journal / source
International journal of intelligent engineering and systems
Volume
19
Issue
4
Pages
1022–1038
Publication date
23 March 2026
Language
EN