Control method for motor of domestic appliance e.g. vacuum cleaner|has fuzzy node trained to recognise classes of patterns recognised by principal nodes in output layer, but with thresholds set at higher levels than corresp threshold levels in principal nodes

Research output: Patent

Abstract

The neural network (10) includes an input layer (12) and an output layer (14) having a number of principal nodes (16), each of which is trained to recognise a different class of pattern. At least one fuzzy node (18) is trained to recognise all classes of pattern recognised by the principal nodes, but with activation thresholds set at levels higher than the corresp threshold levels in the principal nodes (16).The network determines whether an input (20) belongs to a known class of pattern, or whether classification as an unknown class is a more appropriate assignment. A further hidden layer (22) of nodes may be included, and a feedforward architecture may be used in which, the number of nodes in the hidden layer (22) is equal to the number of nodes in the input layer, plus a bias node (24).USE - In e.g processing signals from multi-element array of gas sensors that display broad and overlapping sensitivity to different classes of chemicals, e.g classification of odours into global classes.

Bibliographical metadata

Original languageEnglish
Patent numberEP811198-A; WO9626492-A; WO9626492-A1; AU9647278-A; EP811198-A1; JP11500843-W; EP811198-B1; DE69607460-E; US6192351-B1
Publication statusPublished - 1996