PURPOSE: To optimumly converge all the learning patterns at high speed by correcting a parameter concerning learning while using fuzzy inference.

CONSTITUTION: The presence/absence of abnormality in an object is judged by providing a monitor device to photograph the object to be monitored and a hierarchical neural network device equipped with a learning function to fetch the photographed image data of this monitor device and to input the luminance and/or hue of the image data. A neural network composed of an S layer to input the 100 numerical values of luminance, A layer to input 30 numerical values and R layer composed of two neurons is constituted for each learning pattern, for example, and parameters α and β are optimumly adjusted concerning the corrected amount of synapus load last time and corrected amount based on error at that time. In this case, the parameters α and β are optimumly adjusted concerning all the learning patterns by using fuzzy inference at a fuzzy inference equipment.