Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

InterGridNet: An Electric Network Frequency Approach for Audio Source Location Classification Using Convolutional Neural Networks

Christos Korgialas, Ioannis Tsingalis, Georgios Tzolopoulos, Constantine Kotropoulos

TL;DR

This work addresses ENF-based inter-grid geolocation by formulating it as nine-class classification and introducing InterGridNet, a NAS-optimized shallow RawNet that uses ENF preprocessing, residual blocks, and a GRU to produce frame-level embeddings processed through a softmax classifier. Evaluated on the SP Cup 2016 dataset, the method achieves $92\%$ test accuracy, outperforming several baselines, with bandpass ENF filtering providing substantial gains. The approach yields a compact, effective solution for forensic grid localization and points to future directions in transformer-based models and explainable AI to elucidate ENF patterns across grids.

Abstract

A novel framework, called InterGridNet, is introduced, leveraging a shallow RawNet model for geolocation classification of Electric Network Frequency (ENF) signatures in the SP Cup 2016 dataset. During data preparation, recordings are sorted into audio and power groups based on inherent characteristics, further divided into 50 Hz and 60 Hz groups via spectrogram analysis. Residual blocks within the classification model extract frame-level embeddings, aiding decision-making through softmax activation. The topology and the hyperparameters of the shallow RawNet are optimized using a Neural Architecture Search. The overall accuracy of InterGridNet in the test recordings is 92%, indicating its effectiveness against the state-of-the-art methods tested in the SP Cup 2016. These findings underscore InterGridNet's effectiveness in accurately classifying audio recordings from diverse power grids, advancing state-of-the-art geolocation estimation methods.

InterGridNet: An Electric Network Frequency Approach for Audio Source Location Classification Using Convolutional Neural Networks

TL;DR

This work addresses ENF-based inter-grid geolocation by formulating it as nine-class classification and introducing InterGridNet, a NAS-optimized shallow RawNet that uses ENF preprocessing, residual blocks, and a GRU to produce frame-level embeddings processed through a softmax classifier. Evaluated on the SP Cup 2016 dataset, the method achieves test accuracy, outperforming several baselines, with bandpass ENF filtering providing substantial gains. The approach yields a compact, effective solution for forensic grid localization and points to future directions in transformer-based models and explainable AI to elucidate ENF patterns across grids.

Abstract

A novel framework, called InterGridNet, is introduced, leveraging a shallow RawNet model for geolocation classification of Electric Network Frequency (ENF) signatures in the SP Cup 2016 dataset. During data preparation, recordings are sorted into audio and power groups based on inherent characteristics, further divided into 50 Hz and 60 Hz groups via spectrogram analysis. Residual blocks within the classification model extract frame-level embeddings, aiding decision-making through softmax activation. The topology and the hyperparameters of the shallow RawNet are optimized using a Neural Architecture Search. The overall accuracy of InterGridNet in the test recordings is 92%, indicating its effectiveness against the state-of-the-art methods tested in the SP Cup 2016. These findings underscore InterGridNet's effectiveness in accurately classifying audio recordings from diverse power grids, advancing state-of-the-art geolocation estimation methods.

Paper Structure

This paper contains 9 sections, 1 equation, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. The InterGridNet Framework
  4. Dataset Preperation
  5. Classification Architecture
  6. Experimental Results
  7. Model Validation and Testing
  8. Discussion
  9. Conclusions

Figures (5)

  • Figure 1: Flowchart of the proposed InterGridNet framework.
  • Figure 2: Number of audio and power recording frames in each grid.
  • Figure 3: Spectrograms focused on the nominal ENF value for different grids.
  • Figure 4: Architecture of the proposed optimized shallow RawNet model. The operators utilized in the network include Conv1D(kernel size, strides, filters), MaxPool1D(pool size, strides), GRU(units), and Dense(nodes).
  • Figure 5: Confusion matrices predictions on the testing set employing the proposed InterGridNet.