Feature Extraction and Computational Complexity in Seismological Pattern Recognition.
Abstract
Although seismic recognition problem has been extensively studied by researchers for many years, there does not seem yet to have a satisfactory solution primarily because of the difficulty in determining an effective set of features. In this paper the authors are concerned with the following promising approaches for feature extraction in seismic recognition: (1) seismic discrimination criteria including complexity, spectral ratio and third moment of frequency, (2) orthogonal expansions or multidimensional rotations, (3) information statistics, (4) nonparametric methods, and (5) digital filtering. The feature effectiveness and the computational complexity of various approaches are compared. Although sophisticated computer programming may reduce the computation time, feature selection procedures that are computationally simple are most desirable. The problem of selecting two features (variables) for two-dimensional display of the seismic data is considered. Finally some limiting factors that influence the seismic recognition results are examined. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- May 17, 1973
- Accession Number
- AD0762545
Entities
People
- Chia‐Hung Chen
Organizations
- University of Massachusetts Dartmouth