A Novel Score-CAM based Denoiser for Spectrographic Signature Extraction without Ground Truth

  • 2024-10-30 03:02:40
  • Noel Elias
  • 0

Abstract

Sonar based audio classification techniques are a growing area of research inthe field of underwater acoustics. Usually, underwater noise picked up bypassive sonar transducers contains all types of signals that travel through theocean and is transformed into spectrographic images. As a result, thecorresponding spectrograms intended to display the temporal-frequency data of acertain object often include the tonal regions of abundant extraneous noisethat can effectively interfere with a 'contact'. So, a majority ofspectrographic samples extracted from underwater audio signals are renderedunusable due to their clutter and lack the required indistinguishabilitybetween different objects. With limited clean true data for supervisedtraining, creating classification models for these audio signals is severelybottlenecked. This paper derives several new techniques to combat this problem bydeveloping a novel Score-CAM based denoiser to extract an object's signaturefrom noisy spectrographic data without being given any ground truth data. Inparticular, this paper proposes a novel generative adversarial networkarchitecture for learning and producing spectrographic training data in similardistributions to low-feature spectrogram inputs. In addition, this paper also ageneralizable class activation mapping based denoiser for differentdistributions of acoustic data, even real-world data distributions. Utilizingthese novel architectures and proposed denoising techniques, these experimentsdemonstrate state-of-the-art noise reduction accuracy and improvedclassification accuracy than current audio classification standards. As such,this approach has applications not only to audio data but for countless datadistributions used all around the world for machine learning.

 

Quick Read (beta)

loading the full paper ...