| Popis: |
Conventional speech recognition systems employ procedural languages such as FORTRAN to encode a large amount of heuristic knowledge, which makes the systems inflexible. SPREX (a SPeech Recognition EXpert) is implemented as a knowledge‐based system that is superior to procedural ones in the use of heuristics. This approach is expected to contribute to the establishment of a new methodology of speech recognition research. SPREX has a knowledge base for feature parameter trajectory reading and the knowledge is encoded in the form of production rules. It has the following three advantages. (1) It can cope with the coarticulation effects. (2) It can be constructed by stepwise development. (3) It is easy to debug. In addition to the recognition system, SPREX also has a powerful environment for knowledge base construction, which consists of five subsystems; a speech database, rule database, rule translator, rule learning module, and interface. The environment helps developers in many respects by employing knowle... |
