As reported by the American National Highway Traffic Safety Administration (NHTSA), sleep while driving is one of the most influential factors in fatal vehicle crashes, along with excessive vehicle speed and alcohol consumption. Physiologically speaking, driving for more than two hours in a nocturnal environment produces a driving impairment like a blood alcohol concentration of 0.05%. In this work, we present an innovative and patented sleep prediction method based on the analysis of the Autonomic Nervous System (ANS) (and its subsystems) that monitors the actions happening during the transition from awake to sleep onset. The prediction method processes the Heart Rate (HR) and the Heart Rate Variability (HRV) as collected by a wearable device on the subject wrist. Using a sliding window approach that operates on 20 seconds of samples (acquired at 1 Hz), the trend of the variance of HR and HRV is used to classify the subject condition according to a reduced Karolinska Sleepiness Scale (rKSS) that comprises five stages: Calibration, Awake, Low Drowsiness level, Medium Drowsiness level, High Drowsiness level. The prediction method has been validated experimentally using a set of recordings acquired in a realistic environment (AVL dynamic car simulator, in Graz (AT)). During the experiments, 15 subjects performed several rounds of the Maintenance Wakefulness Test (MWT). Each subject was equipped with a wearable device and apolysomnography medical equipment to gather both the data processed by the proposed approach, and the data set that constituted the ground truth under the supervision of a sleep expert medical doctor. A further experimental section has been conducted, involving the Italian truck company Chrono Express. 15 different drivers have utilized the built-up system for more than 13000 km. The proposed method is sensor agnostic, as it has been proven through preliminary activities with contactless Radio Frequency (RF) sensors. The output produced by the proposed method and the sleep scoring performed by the sleep expert medical doctor during the first experimental section were compared. The first sleep onset event showed an accuracy of 93.3%, a sensibility of 95% and a sensitivity of 100%. Instead, regarding the following sleep onset events an accuracy of 86.66%, a specificity of 66.67%, and a sensitivity of 95.24% were calculated.

Real-Time Sleep Prediction Algorithm Using Commercial off the Shelf Wearable Devices / Pugliese, Luigi; Violante, Massimo; Groppo, Riccardo. - (2023), pp. 634-638. (Intervento presentato al convegno IEEE Smart World Congress 2023 tenutosi a Portsmouth (UK) nel 28-31 August 2023) [10.1109/SWC57546.2023.10448558].

Real-Time Sleep Prediction Algorithm Using Commercial off the Shelf Wearable Devices

Luigi Pugliese;Massimo Violante;
2023

Abstract

As reported by the American National Highway Traffic Safety Administration (NHTSA), sleep while driving is one of the most influential factors in fatal vehicle crashes, along with excessive vehicle speed and alcohol consumption. Physiologically speaking, driving for more than two hours in a nocturnal environment produces a driving impairment like a blood alcohol concentration of 0.05%. In this work, we present an innovative and patented sleep prediction method based on the analysis of the Autonomic Nervous System (ANS) (and its subsystems) that monitors the actions happening during the transition from awake to sleep onset. The prediction method processes the Heart Rate (HR) and the Heart Rate Variability (HRV) as collected by a wearable device on the subject wrist. Using a sliding window approach that operates on 20 seconds of samples (acquired at 1 Hz), the trend of the variance of HR and HRV is used to classify the subject condition according to a reduced Karolinska Sleepiness Scale (rKSS) that comprises five stages: Calibration, Awake, Low Drowsiness level, Medium Drowsiness level, High Drowsiness level. The prediction method has been validated experimentally using a set of recordings acquired in a realistic environment (AVL dynamic car simulator, in Graz (AT)). During the experiments, 15 subjects performed several rounds of the Maintenance Wakefulness Test (MWT). Each subject was equipped with a wearable device and apolysomnography medical equipment to gather both the data processed by the proposed approach, and the data set that constituted the ground truth under the supervision of a sleep expert medical doctor. A further experimental section has been conducted, involving the Italian truck company Chrono Express. 15 different drivers have utilized the built-up system for more than 13000 km. The proposed method is sensor agnostic, as it has been proven through preliminary activities with contactless Radio Frequency (RF) sensors. The output produced by the proposed method and the sleep scoring performed by the sleep expert medical doctor during the first experimental section were compared. The first sleep onset event showed an accuracy of 93.3%, a sensibility of 95% and a sensitivity of 100%. Instead, regarding the following sleep onset events an accuracy of 86.66%, a specificity of 66.67%, and a sensitivity of 95.24% were calculated.
2023
979-8-3503-1980-4
File in questo prodotto:
File Dimensione Formato  
DrowsinessDetection.pdf

accesso aperto

Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 310.85 kB
Formato Adobe PDF
310.85 kB Adobe PDF Visualizza/Apri
Real-time_Sleep_Prediction_Algorithm_using_Commercial_Off_the_Shelf_Wearable_Devices.pdf

non disponibili

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Non Pubblico - Accesso privato/ristretto
Dimensione 1.02 MB
Formato Adobe PDF
1.02 MB Adobe PDF   Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2981367