An expert system has been created to realize a microcomputer-based speech analysis and speech recognition system which uses a low-cost analog filter bank for extracting the acoustic features of speech. The expert system is used to create high-level speech knowledge and offers knowledge-based speech processing. Two speaker-dependent recognition methods were tested with each method resulting in average recognition scores of 95% on a vocabulary size of ten words and branching factor of ten. The system is in an early stage of development and has the potential for improved recognition accuracy and applications in continuous speech recognition.