New method for generation of artificial earthquake record by new model in compression and Artificial Neural Networks

In this paper, a new method for generation of artificial earthquake record from the target spectra is proposed. This method uses new model in compression named MFCC analysis and MLFF Artificial Neural Networks and wavelet analysis. This procedure uses the learning capabilities of neural network to expand the knowledge of the inverse recording from response spectra to earthquake accelerogram. In the first step, wavelet analysis is used to decompose earthquake accelerograms to several levels in which each level covers a special range of frequencies, and then for every level a neural network is trained to learn to relate the response spectra to wavelet coefficients. Mel-Frequency Cepstral Coefficient (MFCC) compress signals for better training of neural networks. Finally, the generated accelerogram using inverse discrete wavelet transform is obtained. Some example is presented to demonstrate the effectiveness of the proposed method.