Neural network model of human intoxication functional state determining in some problems of transport safety solution

Akhmetvaleev, A. M.; Katasev, Alexey S. · 2018 · DOAJ

DOI: 10.20537/2076-7633-2018-10-3-285-293

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical issue of transport safety by proposing a method to detect driver intoxication during pre-trip medical examinations. The authors identify that current diagnostic practices, such as breathalyzer tests, are prone to errors due to methodological inaccuracies and subjective human judgment. To mitigate these risks, particularly for public and municipal transport operators, the study introduces an objective, automated diagnostic tool based on pupillometry and neural network analysis. The methodology relies on pupillometry, which measures the pupil’s reaction to a short light pulse. The system captures video data of the pupil and constructs a papillogram, a time series representing pupil diameter changes. From this curve, twelve specific parameters are extracted, including initial and minimum diameters, constriction amplitude, reaction latency, and various time intervals for constriction and expansion. These parameters serve as inputs for a neural network model designed to classify the driver’s state into two categories: normal or intoxicated. The model is a single-layer perceptron architecture consisting of twelve input neurons, twenty-five hidden neurons, and one output neuron. The structure adheres to the Arnold-Kolmogorov-Hecht-Nilsson theorem to determine the optimal number of hidden neurons. The network is trained on specially prepared datasets where samples are grouped by functional state. The study reports that the neural network model achieves 100% sensitivity and 90% specificity in detecting intoxication. These performance metrics were optimized using ROC analysis to determine the optimal cut-off point for classification, prioritizing the minimization of Type I errors (false negatives). This high sensitivity ensures that all intoxicated individuals are correctly identified, which is crucial for safety-critical applications. The proposed diagnostic scheme involves video registration of the pupillary reaction, parameter calculation, neural network analysis, and final classification, providing medical staff with an objective assessment to support their decision to allow or restrict a driver from operating a vehicle. The significance of this work lies in its potential to enhance the reliability and efficiency of pre-trip medical checks by removing subjective human factors and reducing the need for expensive laboratory equipment or specialized ophthalmological expertise. The authors suggest that this approach can be integrated into broader transport safety systems, including roadside inspections by traffic police and security screening in public transport hubs like airports and train stations. By providing an objective, video-recorded diagnostic record, the model also offers legal evidence for medical decisions, thereby improving overall transport security and reducing accidents caused by impaired drivers.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: computational model