Acoustic Data Compression Using AR Model and Peak Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing acoustic data compression techniques, such as the MP3 format, are insufficient for high fidelity directional acoustic data, particularly in applications requiring compression ratios of 16:1 to 24:1, as they fail to preserve the essential features necessary for detecting or characterizing subsurface structures and underwater bodies effectively.
Innovation Solution
The method involves determining coefficients of a complex auto-regression (AR) model to fit a complex average spectrum of acoustic beams, filtering residuals, and quantizing the spectrum to achieve high fidelity compression, allowing for the transmission of compressed data over limited bandwidth channels while preserving critical features for reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If MP3 compression is used for acoustic data, then compression ratio of about 7:1 to 8:1 is achieved, but compression ratio of 16:1 to 24:1 is insufficient
Solution Approach 1:
The acoustic spectrum is divided into multiple frequency bands, with different compression strategies applied to each band. Critical frequency bands containing important features for object detection are preserved with higher fidelity, while less critical bands are compressed more aggressively, achieving overall compression ratio of 16:1 to 24:1 while maintaining detection capability.
Solution Approach 2:
Different quality levels are applied to different portions of the acoustic data based on their importance. Frequency regions containing signals from subsurface structures or underwater bodies are preserved with high fidelity, while other regions undergo more aggressive compression, allowing high compression ratios while preserving critical detection features.
2Reliability
If full data rate is transmitted, then all acoustic information is preserved, but bandwidth requirement exceeds limited communication channel capacity
Solution Approach 1:
The method extracts and transmits only the most critical features of the acoustic data at full fidelity, while representing less critical portions in a compressed form. This selective extraction allows the system to operate within limited bandwidth constraints while preserving the essential information needed for object detection and characterization.
Solution Approach 2:
The system dynamically adjusts compression parameters based on the content of the acoustic data. When important features are detected in certain frequency bands, the compression ratio for those bands is reduced to preserve fidelity. This adaptive parameter adjustment enables efficient bandwidth utilization while maintaining data quality where needed.
Data Source
AI summary
Techniques include determining coefficients of a complex auto regression (AR) model to fit a complex average spectrum at a base frequency resolution of a set of one or more measured acoustic beams during a time block. Residuals derived by filtering actual data through an inverse of the AR model are determined at frequencies below a first threshold frequency. A quantized spectrum of the residuals is determined at the base frequency resolution. Magnitude, phase, and frequency bin at the base frequency resolution are determined for each peak of a set of one or more narrowband peaks above a second threshold frequency for the set of one or more measured acoustic beams. A message is sent, which indicates without loss the coefficients of the AR model, the quantized spectrum of the residuals, and the frequency bin, magnitude and phase for each peak of the set of one or more narrowband peaks.