Block-adaptive filtering and its application to seismic-event detection
Block digital filtering involves the calculation of a block or finite set of filter output samples from a block of input samples. The motivation for block processing arises from computational advantages of the technique. Block filters take good advantage of parallel processing architectures, which are becoming more and more attractive with the advent of very large scale integrated (VLSI) circuits. This thesis extends the block technique to Wiener and adaptive filters, both of which are statistical filters. The key ingredient to this extension turns out to be the definition of a new performance index, block mean square error (BMSE), which combines the well known sum square error (SSE) and mean square error (MSE). A block adaptive filtering procedure is derived in which the filter coefficients are adjusted once per each output block in accordance with a generalized block least mean-square (BLMS) algorithm. Convergence properties of the BLMS algorithm are studied, including conditions for guaranteed convergence, convergence speed, and convergence accuracy. Simulation examples are given for clarity. Convergence properties of the BLMS and LMS algorithms are analyzed and compared. They are shown to be analogous, and under the proper circumstances, equivalent. The block adaptive filter was applied to the problem of detecting small seismic events in microseismic background noise. The predictor outperformed the world-wide standardized seismograph network (WWSSN) seismometers in improving signal-to-noise ratio (SNR).
- Research Organization:
- California Univ., Livermore (USA). Lawrence Livermore National Lab.
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 6149531
- Report Number(s):
- UCRL-53147; ON: DE81028361
- Resource Relation:
- Other Information: Thesis
- Country of Publication:
- United States
- Language:
- English
Similar Records
MVSE (minimizes variance squared error) adaptive filtering subject to a constraint on MSE (mean squared error)
Efficient block processing of long duration biotelemetric brain data for health care monitoring