1-D syntactic pattern recognition for the detection of bright spots

In a seismogram, the wavelets of a bright spot have some specific structural pattern; so syntactic pattern recognition approach is proposed for the detection of bright spots. Testing traces are selected from the input seismogram, and tree classification techniques are used in the extraction of bright spot wavelet patterns. The system for one-dimensional (1-D) syntactic pattern recognition includes the likelihood-ratio test, optimal amplitude-dependent encoding, probabili- (Figure Presented) ty of detecting the signal involving in the global and local detection, and threshold setting. The relation between error probability and the minimum Levenshtein-distance classification is proposed. The system is used to detect the candidate bright spot, trace by trace, in a simulated seismogram as well as real seismograms of Mississippi Canyon and High Island.