Aural Based Scene Understanding for Sonar Applications

Abstract

The objective of this research is to investigate approaches using alternative data representations to directly exploit the information content of raw or near-raw sonar data. In particular, this effort explored how one can encode and exploit streaming sonar data as it is received, and the associated performance tradeoffs when beamforming is omitted. The problem of seabed texture characterization is employed as a testbed for this investigation. The approach herein examines the information in individual and combined sonar echo returns as they pertain to forming auditory objects that combine to form a scene. In particular, we investigated the application of summary statistics as a means for representing underlying scene characteristics such as sand ripples, gravel beds, grass, rocks, sediment, and sea grass. Summary statistics were drawn from the sonar performance estimation community and from communities that study aural perception systems, and the features they represent range from nearly stationary to episodic whether spatially or temporally.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 28, 2023
Accession Number
AD1215976

Entities

People

  • Daniel C Brown

Organizations

  • Pennsylvania State University

Tags

DTIC Thesaurus Topics

  • Acoustics
  • Algorithms
  • Automated Target Recognition
  • Classification
  • Computer Vision
  • Data Sets
  • Detection
  • Detectors
  • Dimensionality Reduction
  • Feature Extraction
  • Frequency
  • Military Research
  • Parkinson'S Disease
  • Signal Processing
  • Stainless Steel
  • Standards
  • Target Recognition

Readers

  • Acoustical Oceanography.
  • Computer Vision.
  • Systems Analysis and Design