Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mobile Recording Device Recognition Based Cross-Scale and Multi-Level Representation Learning

Chunyan Zeng, Yuhao Zhao, Zhifeng Wang

TL;DR

This paper introduces a modeling approach that employs multi-level global processing, encompassing both short-term frame-level and long-term sample-level feature scales, and thoroughly investigates the transferability of the model, achieving an 87.9% accuracy in a classification task on a new dataset.

Abstract

This paper introduces a modeling approach that employs multi-level global processing, encompassing both short-term frame-level and long-term sample-level feature scales. In the initial stage of shallow feature extraction, various scales are employed to extract multi-level features, including Mel-Frequency Cepstral Coefficients (MFCC) and pre-Fbank log energy spectrum. The construction of the identification network model involves considering the input two-dimensional temporal features from both frame and sample levels. Specifically, the model initially employs one-dimensional convolution-based Convolutional Long Short-Term Memory (ConvLSTM) to fuse spatiotemporal information and extract short-term frame-level features. Subsequently, bidirectional long Short-Term Memory (BiLSTM) is utilized to learn long-term sample-level sequential representations. The transformer encoder then performs cross-scale, multi-level processing on global frame-level and sample-level features, facilitating deep feature representation and fusion at both levels. Finally, recognition results are obtained through Softmax. Our method achieves an impressive 99.6% recognition accuracy on the CCNU_Mobile dataset, exhibiting a notable improvement of 2% to 12% compared to the baseline system. Additionally, we thoroughly investigate the transferability of our model, achieving an 87.9% accuracy in a classification task on a new dataset.

Mobile Recording Device Recognition Based Cross-Scale and Multi-Level Representation Learning

TL;DR

This paper introduces a modeling approach that employs multi-level global processing, encompassing both short-term frame-level and long-term sample-level feature scales, and thoroughly investigates the transferability of the model, achieving an 87.9% accuracy in a classification task on a new dataset.

Abstract

This paper introduces a modeling approach that employs multi-level global processing, encompassing both short-term frame-level and long-term sample-level feature scales. In the initial stage of shallow feature extraction, various scales are employed to extract multi-level features, including Mel-Frequency Cepstral Coefficients (MFCC) and pre-Fbank log energy spectrum. The construction of the identification network model involves considering the input two-dimensional temporal features from both frame and sample levels. Specifically, the model initially employs one-dimensional convolution-based Convolutional Long Short-Term Memory (ConvLSTM) to fuse spatiotemporal information and extract short-term frame-level features. Subsequently, bidirectional long Short-Term Memory (BiLSTM) is utilized to learn long-term sample-level sequential representations. The transformer encoder then performs cross-scale, multi-level processing on global frame-level and sample-level features, facilitating deep feature representation and fusion at both levels. Finally, recognition results are obtained through Softmax. Our method achieves an impressive 99.6% recognition accuracy on the CCNU_Mobile dataset, exhibiting a notable improvement of 2% to 12% compared to the baseline system. Additionally, we thoroughly investigate the transferability of our model, achieving an 87.9% accuracy in a classification task on a new dataset.

Paper Structure

This paper contains 27 sections, 39 equations, 10 figures, 9 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Information characterization of recording devices based on spectral feature engineering
  4. Recognition of Recording Devices Based on Classical Machine Learning Algorithms
  5. Recognition of recording device based on deep learning technique
  6. Preliminaries
  7. Problem Formalization
  8. Definitions
  9. Proposed Method
  10. Timing Tandem Features Based on MFCC Multi-level Localized Frame-levels
  11. Modeling Mobile Recording Device Recognition Based on Multi-Level Scales
  12. Spatiotemporal Fusion Augmentation of Short-term Frame-level Localized Features Based on 1D ConvLSTM
  13. BiLSTM-based Representation Learning on Long-term Sample-level Feature Scales
  14. Transformer-encoder-based Learning of Short and Long-term Representations at Global Feature Scales
  15. Training and optimization
  16. ...and 12 more sections

Figures (10)

  • Figure 1: Flowchart of mobile recording device recognition problem.
  • Figure 2: Overall framework of the proposed cross-scale and multi-level representation learning based mobile recording device recognition method.
  • Figure 3: Process of MFCC-based timing tandem feature extraction.
  • Figure 4: Schematic of the dynamic range of the timing tandem feature.
  • Figure 5: Schematic diagram of the module for processing short-time frame-level features.
  • ...and 5 more figures