Self-organizing map based artificial olfactory system online correction sample generation method

Abstract

The invention belongs to the technical field of smell analysis, and discloses a self-organizing map based artificial olfactory system online correction sample generation method. The method comprises an initial training stage and an online updating stage. In the initial training stage, a self-organizing map neural network multi-layer structure is constructed according to the number of sample categories, self-organizing map neural network neuron weight initialization is performed according to training samples, the self-organizing map neural network neuron weights are enabled to act as an initial training sample set; and in the online updating stage, and adjustment is performed on local region neuron weights by using a test sample according to a classification result of a follow-up classifier. The current self-organizing map neural network neuron weights are enabled to act as an online training sample set to perform online correction on a mode recognition method. Results show that the method can improve the long-term drift resisting capacity of an artificial olfactory system under an online working condition, can automatically generate a correction sample in the online working process and provides guarantee for the artificial olfactory system to realize automatic online correction.

Description

technical field [0001] The invention belongs to the technical field of odor analysis, in particular to a self-organizing map-based method for generating online calibration samples for an artificial olfactory system. Background technique [0002] The artificial olfactory system is a novel method of odor analysis, which has the advantages of rapid detection, non-invasive, easy operation, and low cost. The artificial olfactory system is mainly divided into two parts: a gas sensor array and a pattern recognition method. The "gas sensor array" uses low-cost gas sensors with cross sensitivity to obtain odor maps; the "pattern recognition" uses artificial intelligence, machine learning, etc. Methods Qualitative and quantitative analysis of the odor. The "long-term drift" of the gas sensor is an unavoidable problem in the artificial olfactory system. With the prolongation of the use time, the odor spectrum of the "gas sensor array" will change slowly and irregularly; making the rec...

AI Technical Summary

Problems solved by technology

[0003] To sum up, the problems existing in the existing technology are: 1) For equipment that needs to work online for a long time, routine calibration cannot be performed, so the system recognition accuracy will drop significantly with the working time; 2) Conventional calibration methods need to prepare specific calibration samples And the equipment cannot work normally during the calibration period, wasting a lot of financial, manpower and material resources

Images