Database Mining of Seismic Data Using Wavelets



How the Wavelet Features Combined with a Classifier (Pattern Table) Will Be Used to Query the Coherency Database.




The Steps Used to Both Build the Query Database (Upper) and Execute Queries (Lower) in the Proposed RDBMS-Based Storage of Wavelet Features




The Lamont-Based System Architecture




Schematic of Using Wavelet Features and Rule-Based Patterns for Database Mining




The Wavelet Representation Puts the "Bits" were the Information is Located




We Propose the Use of a Polar Gabor Wavelet Pyramid


Click here for info on Gabor and 3-D Gabor Transforms and
here to learn more about the group ISPG. and
3-D region growing for Video Compression




The Use of a Genetic Algorithm to Evolve Rule-Based Patterns




Instead of a Genetic Algorithm, a Tree Classifier can be Used.




Performance for the Tree is Measured with a Confusion Matrix




The Decision Tree Cuts the Feature Hyperspace One Dimension per Tree Node




Unsupervised Clustering is a Way of Identifying New Patterns


In the example below from speech recognition, an "acoustic distance" based on spectral features called cepstra is used to separate transmissions from air traffic controllers (red) and pilots (white).