Adaptive neuro-fuzzy inference system for generation of diffuser dot patterns in light guides

摘要

We present an adaptive neuro-fuzzy inference system (ANFIS) strategy to generate dot patterns in a liquid-crystal display light guide panel. The ANFIS model combines the learning capabilities of neural networks and the knowledge illustration of fuzzy logic systems using linguistic expressions. A hybrid learning algorithm, based on the least square method and the back propagation algorithm, is utilized to identify the parameters of ANFIS. Two inputs of ANFIS are the dot radius and the distance from dots to a light source, and one output is the illuminance over a light guide panel. During the process of generating diffuser dot patterns, ANFIS carries out efficient input selection, rule creation, networks training, and parameter estimation to create an appropriate model by the learning algorithm. The results show that the proposed model can achieve an even illuminance condition and effectively improve brightness in accordance with the light source position. Moreover, a comparative analysis suggests that the ANFIS-based approach outperforms the traditional model in terms of overall illuminance and color uniformity.