Application of wavelet transforms for data compression, noise reduction and interpretation of digital seismic signals

Mining induced seismic signals are often analyzed in the frequency domain by making use of the fast Fourier transform. This is usually done to determine the corner frequency of the displacement spectrum which is related to the physical dimension of the source; this property provides the rock mechanics engineer with information about the size of the failing region which generated the seismic signal. Wavelet transform methods have not yet been applied to mining induced seismic events and have the potential to compactly describe the seismograms and moreover provide additional information about the seismic source.; Although Fourier analysis has served seismologists for many years, transforming a seismic signal with this procedure is inherently flawed, since it is based on an attempt to decompose the finite seismic signal into a collection of infinite sines and cosines or both. Wavelet transforms are based on basis functions that have compact support (finite in an interval of time) and are therefore attractive for seismic data interpretation. The wavelet transform is a relatively new signal processing technique, which has found application in several fields, mainly digital signal compression and data storage algorithms for images, etc. Wavelet transforms of a signal have the property that many of the wavelet coefficients are close to zero and can be discarded with little or no information loss in the reconstructed signal. Thus, the method is useful as a signal compression algorithm. Another interesting property of the wavelet transform is related to the flexibility of the analyzing window. This allows a signal to be viewed and analyzed simultaneously in the time domain and in the frequency domain.; In the present thesis, the benefits of the implementation of wavelets as a data compression tool for seismic signals are shown. For instance compression ratios of more than 20:1 with little loss of information are achieved. The capabilities of wavelets in the recognition of the different features from the seismic signal, such as noise and superimposed events, are also explored. In the type of environment where the seismic signal is recorded, a seismogram could include noise, coda-waves (reflection or refraction waves) and other type of anomalies. In this study the possibilities of identifying the different components of the signal are shown and methods to improve the overall seismic source location are suggested.