CONTROLLING PUBLIC ADDRESS SYSTEMS BASED ON FUZZY INFERENCE AND NEURAL-NETWORK

A public address system has been constructed that automatically adjust s output speech level and equalization in accordance with ambient nois e levels. The system employs a neural network to distinguish between m ale/female voices, and fuzzy inference for decision making regarding o utput speech levels and equalization. The results of auditory experime nts aimed at the improvement of speech intelligibility show gain adjus tment along with the use of equalization to be an advantage, and also that appropriate equalizer patterns for male/female are different. The fuzzy inference subsystem uses these results as rules and membership functions.