Recursive Noise Power Estimation in Seismic Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional seismic data processing techniques are inefficient in estimating noise power spectrum, especially in multi-measurement marine seismic data, due to spatial and temporal variations, low signal-to-noise ratio conditions, and non-stationary noise environments, which affects the robustness of noise reduction and signal quality.
Innovation Solution
A method that involves receiving multi-measurement seismic data from a multi-dimensional seismic sensor array, partitioning it into overlapping time-space windows, computing the frequency-domain spectrum, estimating signal presence probability, and recursively updating the spectral noise power based on the frequency spectrum and prior probabilities to iteratively improve noise power estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional techniques estimate noise power spectrum by averaging data power spectrum assuming noise is time-space stationary, then the processing is simple, but the estimation accuracy deteriorates under non-stationary noise environments and low signal-to-noise ratio conditions
Solution Approach 1:
The patent segments the seismic data into multiple time-space windows and processes each window separately to capture local non-stationary noise characteristics. This allows the noise estimation to adapt to temporal and spatial variations while maintaining computational feasibility through localized processing.
Solution Approach 2:
The patent implements a dynamic noise estimation approach that updates noise power spectrum estimates iteratively across different time-space windows rather than assuming stationarity. This dynamic adaptation enables accurate noise characterization in non-stationary environments while preserving signal components.
2Ease of operation
If noise power spectrum is estimated separately in each component of multi-measurement data, then the processing is straightforward, but the estimation efficiency deteriorates
Solution Approach 1:
The patent merges the processing of multiple measurement components (pressure and particle velocity) by jointly estimating the noise power spectrum across all components simultaneously. This combined approach leverages correlations between components to improve estimation efficiency while maintaining operational simplicity through unified processing.
3Measurement precision
If iterative recursive estimation is used to update spectral noise power based on signal presence probability, then the noise power estimation accuracy improves, but the computational complexity increases
Solution Approach 1:
The patent implements a feedback mechanism where the estimated signal presence probability from previous iterations is used to guide the current noise power estimation, which in turn refines the signal presence probability for the next iteration. This iterative feedback loop progressively improves estimation accuracy while the structured approach prevents exponential complexity growth.
Solution Approach 2:
The patent performs preliminary estimation of signal presence probability before conducting the full noise power spectrum estimation. This preliminary action provides initial guidance that simplifies subsequent iterative refinement, reducing the overall computational burden while maintaining high accuracy.
Data Source
AI summary
Various implementations described herein are directed to methods for processing seismic data, including estimating a spectral noise power of multi-measurement seismic data received from a multi-dimensional seismic sensor array having multiple seismic sensors. The methods may include receiving a shot record of multi-measurement seismic data in time-domain, partitioning the shot record into overlapping time-space windows, and computing a frequency-domain spectrum for each time-space window. The methods may include computing a signal presence probability for each time-space window using the frequency-domain spectrum and prior probabilities of signal presence and absence for each time-space window. The methods may include iteratively updating a collective spectral noise power by recursively estimating the spectral noise power of a current time-space window based on the frequency spectrum for the current time-space window, the signal presence probability computed for the current time-space window, and a previously estimated spectral noise power of a previous time-space window.