NEURAL FUZZY MODEL FOR DETERMINING THE FUNCTIONAL STATE OF HUMAN FATIGUE
Authors:
1.
KNITU-KAI
2.
Kazanskiy nacional'nyy issledovatel'skiy tehnicheskiy universitet imeni A.N. Tupoleva
3.
KNITU
4.
Kazan National Research Technical University named after A.N. Tupolev
Type:
Article
Pages:
from 130
to 135
Status:
In work
Language:
Russian
Keywords:
NEURAL FUZZY MODEL, HUMAN FUNCTIONAL STATE, FATIGUE, FUZZY NEURAL NETWORK, OBJECTS STATE ASSESSMENT
Abstract (English):
The article is devoted to the construction and testing of a neuro-fuzzy model for determining the state of human fatigue. Classical approaches to assessing the state of human fatigue are often based on the use of subjective methods influenced by individual characteristics of a person and situational factors. This actualizes the need to develop more objective and accurate methods that can take into account various aspects of the functional state of fatigue. Currently, fuzzy and neuro-fuzzy models are gaining popularity in the tasks of determining human fatigue. They allow taking into account the complexity and uncertainty inherent in the processes associated with the functional state of a person. The key advantage of such models is their ability to work with inaccurate, incomplete and noisy data that often accompany processes associated with a person. To build a neuro-fuzzy model for determining the functional state of human fatigue, it was necessary to solve the following problems: preparing the initial data for analysis, training a fuzzy neural network and assessing its adequacy. The study used an experimental data set obtained by the pupillometry method in people in one of two functional states: a normal state and a state of fatigue. The data set contained files presented in two formats: as text (TXT) and as an image (PNG). The "Norm" folder contained 236 files, and the "Deviation" folder contained 216. The TXT text files were sequences of normalized pupil size values every 0.04 seconds. TXT files were selected as the initial data to form neuro-fuzzy models. For ease of use, all the initial data from these files were loaded into one Excel table. Based on the table data, the following pupil characteristics were calculated: minimum and final diameters, constriction amplitude, constriction and expansion speed, and constriction time. The total data volume for analysis contained 452 records. The author's software package was selected as the modeling environment, allowing training a fuzzy neural network. The following parameter values were used for its training: chromosome length in the genetic algorithm - 10, number of chromosomes - 150, probability of daughter chromosome mutation - 10%, number of idle work epochs - 50, number of idle network training cycles - 2. The classification accuracy of the constructed model on the training data sample was 95.95%, on the test sample - 92.59%. The training time of the model was 23 minutes 35 seconds. The analysis of the obtained results indicates the adequacy of the constructed model and the possibility of its effective practical use.
The article is devoted to the construction and testing of a neuro-fuzzy model for determining the state of human fatigue. Classical approaches to assessing the state of human fatigue are often based on the use of subjective methods influenced by individual characteristics of a person and situational factors. This actualizes the need to develop more objective and accurate methods that can take into account various aspects of the functional state of fatigue. Currently, fuzzy and neuro-fuzzy models are gaining popularity in the tasks of determining human fatigue. They allow taking into account the complexity and uncertainty inherent in the processes associated with the functional state of a person. The key advantage of such models is their ability to work with inaccurate, incomplete and noisy data that often accompany processes associated with a person. To build a neuro-fuzzy model for determining the functional state of human fatigue, it was necessary to solve the following problems: preparing the initial data for analysis, training a fuzzy neural network and assessing its adequacy. The study used an experimental data set obtained by the pupillometry method in people in one of two functional states: a normal state and a state of fatigue. The data set contained files presented in two formats: as text (TXT) and as an image (PNG). The "Norm" folder contained 236 files, and the "Deviation" folder contained 216. The TXT text files were sequences of normalized pupil size values every 0.04 seconds. TXT files were selected as the initial data to form neuro-fuzzy models. For ease of use, all the initial data from these files were loaded into one Excel table. Based on the table data, the following pupil characteristics were calculated: minimum and final diameters, constriction amplitude, constriction and expansion speed, and constriction time. The total data volume for analysis contained 452 records. The author's software package was selected as the modeling environment, allowing training a fuzzy neural network. The following parameter values were used for its training: chromosome length in the genetic algorithm - 10, number of chromosomes - 150, probability of daughter chromosome mutation - 10%, number of idle work epochs - 50, number of idle network training cycles - 2. The classification accuracy of the constructed model on the training data sample was 95.95%, on the test sample - 92.59%. The training time of the model was 23 minutes 35 seconds. The analysis of the obtained results indicates the adequacy of the constructed model and the possibility of its effective practical use.
Keywords:
NEURAL FUZZY MODEL, HUMAN FUNCTIONAL STATE, FATIGUE, FUZZY NEURAL NETWORK, OBJECTS STATE ASSESSMENT
NEURAL FUZZY MODEL, HUMAN FUNCTIONAL STATE, FATIGUE, FUZZY NEURAL NETWORK, OBJECTS STATE ASSESSMENT
