Neural Network Fuzzy Art System Design for Biomedical Signals Processing

Introduction. This paper proposes the design of an electroencephalogram (EEG) signal analysis system based on a Fuzzy Adaptive Resonance Theory Mapping (Fuzzy ARTMAP) neural network for clinical signal classification. Materials and Methods. To accelerate the signal clustering process, a parallelized Fuzzy ART architecture is introduced. The system parallelizes the evaluation of choice functions and vigilance condition verification, replacing iterative sequential search with a unified parallel process. Standard Winner-Takes-All (WTA) units are replaced with rank-order filter (ROF)-based WTA neural circuits to determine the maximum choice function values. The architecture is structured for deployment on hybrid analog/digital hardware platforms to ensure real-time operation. Results. Parallelizing category choice and resonance verification significantly reduces computational latency compared to traditional sequential Fuzzy ARTMAP iterations. The integration of rank-order filter hardware circuits facilitates rapid identification of winning categories while maintaining clustering stability during continuous EEG processing. Conclusions. The proposed parallel Fuzzy ARTMAP model eliminates sequential search bottlenecks in neural category selection, enabling high-speed processing of EEG signals. Its compatibility with low-cost analog/digital hardware provides a foundation for compact, energy-efficient, real-time biomedical monitoring devices.