الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 220 |  |  |
| | | from | 220 | | |
Abstract
This thesis has been concerned with the investigation of the seismic signals and earth explosions. The main goal was to seek for the appropriate set of features that is convenient to express the main characteristics of the seismic earthquakes and explosions data and to develop an automated system that can discriminate between micro earthquakes and micro explosions; also between earthquakes and underground nuclear explosions. The seismic records were composed of two data sets; local at Suez Gulf area, Egypt, recorded by Egyptian National Seismic Network (ENSN), regional at Lop Nor, China, it was obtained from the Incorporated Research Institutions in Seismology Data Management Center (IRIS DMC). Four different features extraction techniques were used. These are: spectral analysis, autoregressive (AR) modeling, wavelet transform (WT) and wavelet packet (WP). Reduction of feature dimensionality was made using principle component analysis (PCA), and average energy coefficient. The resultant derived features sets were used as input to the classifier to characterize the difference between micro earthquakes, micro explosions and between earthquakes and underground nuclear explosions. Using specially designed neural networks, a multilayer feed forward neural network trained using backpropagation technique; it was possible to discriminate between micro earthquakes and micro explosion at Suez Gulf, Egypt and between earthquakes and underground nuclear explosions at Lop Nor, China. For the Suez Gulf, Egypt data, it has been found that the highest percentage for correct classification between micro earthquakes and micro explosions is obtained using the set of features extracted from the autoregressive LPC technique in addition to the source parameters reached 97.88%. While, for the data of Lop Nor, China, the highest percentage for correct classification obtained is 96.6% using the features extracted from the Daubechies Wavelet DAU(1:8) and PCA in addition to the source parameters. An attempt was made to combine the individual scores from multiple matchers in order to compensate their individual weakness and to preserve their strength: a data fusion approach at the matching score level was adopted. Three normalization methods were used for fusion at the matching score level. These are; MinMax (MM), ZScore (ZS) and Tanh (TH) normalization method. The correct classification rate for micro earthquakes and micro explosion at Suez Gulf, Egypt, improved and for earthquakes and underground nuclear explosion at Lop Nor, China, reached 100%. The developed system is promising and shows that the adopted signal processing techniques are quite satisfactory discrimination performance.