Fuzzy Modular Neural Model for Blinking Coding Detection and Classification for Linguistic Expression Recognition

Abstract

EEG signal analysis provides a new alternative to implement brain computer interfaces.Among the possible signals that can be used for brain computer interfaces are signals generated during blinking. This chapter presents a novel fuzzy modular neural model for linguistic expression recognition using blinking coding detection and classification. The linguistic expressions are coded into blinking sequences, and the proposed system analyzes these sequences to detect possible existence of events, expression codes. The blinking signals are first preprocessed to eliminate possible offset and to limit their bandwidth. Then, a new processing step obtains statistical information of the signals and makes them invariant to future users. If a code expressions is detected, it is processed to generate a feature vector with statistical and frequency features. The feature vector is classified for a set of specialized modular neural networks, and finally an output analysis scheme is used to reduce an improper decision. The fuzzy detection systems was tested with signals corresponding to blinking codes, involuntary blinking, and noise and achieved 100% of correct detection, and the final classification of expression with the modular network was 94.26%. Regarding these results, the propose system is considered suitable for applications with seriously impaired persons to establish a basic communication with other person through blinking code generation.