Seismic signal understanding: a knowledge-based recognition system

The authors address the issue of automating routine signal analysis in the seismological domain and propose an approach that combines artificial intelligence and signal processing techniques. Distinctive features of the knowledge involved in the expert activity are investigated and used to design a knowledge-based system to support seismological interpretation. The architecture of the system, which is based on the blackboard scheme, is discussed. The implementation of a prototype (SNA2) is presented, and details are given on its hybrid problem-solving activity. Emphasis is given to the initial, selective inspection of data records, a critical aspect on the interpretive process; accurate parameter estimates are seen as subsequent, straightforward applications of well-known procedures. Several solutions are proposed to modeling the expert´s focus of attention, simple but effective tools are adopted to extract relevant signal features, and a method is proposed for approximate location of events. Results of the application of the system confirm the effectiveness of the approach