STATISTIC CHARACTERIZATION USING TIME DOMAIN IN EARLY DETECTION ELECTRONIC NOSE INSTRUMENT OF DIABETES MELLITUS
1)Sari Ayu Wulandari, 2) Sutikno Madnasri*, 3)Susilo, 4) Ratih Pramitasari, 5) Retno Evita Dewi
1,4,5) Department of Electrical Engineering, Dian Nuswantoro University, Semarang, Indonesia
2,3) Advanced Composite Laboratory, Physics Department, Faculty of Mathematics and Natural Sciences, Universitas Negeri Semarang
Abstract
Aroma based electronic nose recognition instrument develops for medical needs. This instrument is expected to be able to detect diseases in the body, through urine, blood, breath, sketch, sweat and human feces. This is a new challenge for instrument enose, especially for recognizing type 2 diabetes mellitus (DM). Some previous studies applied statistical methods using the time domain, such as using maximal, minimal, standard deviation, variance and means. Each sensor uses only 1 statistical feature. Enose has not used more than 1 statistical feature. This article discusses the effect of the number of features included in the process of recognition of the results of the introduction. Enose uses 4 polymer gas sensors, then calculates the statistical characteristics of the changes in the gas sensor resistance. This results in a feature matrix. The next process is calculating feature extraction using the PCA (Principle Component Analysis) method and clustering using the FCM (Fuzzy C Means) method of the feature matrix. Enose uses characteristic variations ranging from 2 traits, 3 traits, 4 traits and 5 traits. The results of the introduction show that, enose using 3 characteristics has a high accuracy of 83.3%.
Keywords: characterization, statistical characteristics, enose
Topic: Theoretical and Analysis in Materials