Real-time sleep prediction based on statistical analysis of a reduced set of biometric data

Abstract

Software storable on and executable by an electronic computing resource (10), the software being designed, when executed, to configure the electronic computing resource (10) to detect and / or predict in real time a behavioral state and / or transition among one or more of a subject's wakefulness (W), drowsiness (D), and sleep (S) phases. The software, when executed, is designed to configure the electronic computing resource (10) to: receive (20) a subject's biometric signal, process (21-29) the received biometric signal to classify the received biometric signal into one of different classes associated with the behavioral state and / or transition among one or more of the subject's wakefulness (W), drowsiness (D), and sleep (S) phases, and detect and / or predict (30) the subject's behavioral state and / or transition among the wakefulness (W), drowsiness (D), and sleep (S) phases based on the classified biometric signal. The software, when executed, is designed to configure the electronic computing resource (10) to: calculate (24) at least a first one composite quantity for and based on the received biometric signal; calculate (29) at least one threshold value for the received biometric signal based on the at least first one composite quantity calculated therefor; and classify (29) the biometric signal into one of different classes associated with behavioral states and / or transitions among one or more of wakefulness (W), drowsiness (D) and sleep (S) phases based on the calculated threshold value.

Description

[Technical Field]

[0001] This application claims priority to European Patent Application No. 22197077.5, filed September 22, 2022, and Italian Patent Application No. 102023000019470, filed September 21, 2023, the contents of which are incorporated herein by reference in their entirety.

[0002] The present invention relates generally to real-time detection and / or prediction of a subject's behavioral state and / or transitions among one or more of the Awake (W), Drowsiness (D) and Sleep (S) phases, and specifically to sleep onset in a subject through statistical analysis applied to a reduced set of biometric data.

[0003] Thus, the present invention aims to warn a subject before loss of cognitive ability and to provide a means for interacting with the subject, thereby significantly reducing the likelihood of a traumatic event, such as a traffic accident. [Background technology]

[0004] As is well known, the non-invasive identification of different behavioral phases of an individual (Awake (W), Drowsiness (D), Sleep (S)) is a challenge that concerns several major areas of modern life, such as public health and safety in transportation and work environments, and has significant impacts from a socio-economic perspective, as well as significant impetus for research and development.

[0005] The wake-sleep transition is difficult to define because it is a process, not an on-off phenomenon or the moment when such a transition occurs. Specifically, falling asleep can be defined as the transition phase between wakefulness and sleep, i.e., the time interval between the moment when a subject is prone to sleep and the moment when the subject actually falls asleep.

[0006] Although operationally, defining wakefulness and sleep is relatively straightforward, defining the stages of sleep onset is more challenging because they involve continuously alternating wakefulness and sleep fragments, thereby determining the absence of a single "moment" at which sleep onset occurs, but rather the existence of a period during which numerous variables (i.e., neurophysiological and behavioral) fluctuate until sleep onset (Ogilvie 2001). The difficulty in defining the wakefulness-sleep transition also stems from the standard system of sleep phase classification (i.e., Rechtschaffen & Kales) adopted in known studies of sleep onset, which uses the same parameters (i.e., between 30-second epochs) to analyze heterogeneous durations, such as sleep onset lasting a few minutes and sleep lasting several hours.

[0007] Ultimately, therefore, the process of falling asleep (also defined as sleep onset (SO)) can be considered multidimensional, involving, for example, subjective, behavioral, and physiological aspects.

[0008] In order to systematically assess the stages of drowsiness and facilitate the development of an automatic early drowsiness detection system, the applicant has realized that an accurate measurement scale for drowsiness levels is needed, and in this regard, several known methods have been proposed, some of which are described below.

[0009] First, the applicant states that one of the most widely used scales in the literature is the Karolinska Sleepiness Scale (KSS), which measures subjective sleepiness levels at specific times during the day. Specifically, the KSS measures sleepiness through a driver's verbal description on a nine-point scale ranging from level (1) "very alert" to level (9) "very sleepy, making great efforts to remain alert." Subjective sleepiness assessment scales such as the KSS are quick, simple, and low-cost methods. Furthermore, the applicant states that it is important to emphasize that because they are self-assessments, they can be influenced by both the surrounding environment and the subject's emotional state. Furthermore, because individuals very often tend to overestimate or underestimate their sleepiness, the results obtained from these methods often do not overlap with objective measurements.

[0010] To detect drowsiness, various known evaluation criteria were implemented: Image-based assessment criteria: Some signs of drowsiness are visible, specifically the recording of a driver's facial expressions and movements, especially head movements, which can be recorded with a camera or visual sensor. Such systems are generally non-intrusive, non-invasive, and cost-effective because they require only a camera to collect the necessary data and can provide details of a subject's drowsiness at a very advanced stage when the subject's cognitive state no longer allows them to perform the activity. However, the system's capabilities are significantly impacted when facial data is difficult to track due to impairments, and it cannot cover the large population of people who sleep with their eyes open, especially those with obstructive sleep apnea syndrome (OSAS).

[0011] Biologically Based Metrics: Many biological signals have been used to detect driver drowsiness, such as signals from brain activity, heart rate, respiratory rate, pulse rate, and body temperature. These biological signals, also known as physiological metrics, have proven to be more accurate and reliable in detecting drowsiness, particularly due to their ability to capture early biological changes that may be present in case of drowsiness, thus alerting the driver before physical signs of drowsiness appear. Several efforts are aimed at developing cost-effective, as minimally intrusive, and preferably contactless, sensors that can provide accurate measurements of the required biometric parameters.

[0012] Vehicle-based assessment: This method relies on tracking and analyzing driving patterns, which form distinctive driving patterns. Therefore, the driving patterns of a drowsy driver can be distinguished from those of an alert driver. However, being an indirect method of drowsiness detection, such a solution is neither accurate nor fast enough to restore the driver's level of consciousness.

[0013] -Hybrid-based metrics: Hybrid drowsiness detection systems utilize a combination of image-, biological-, and vehicle-based metrics to extract drowsiness features, with the goal of producing a more robust, accurate, and reliable drowsiness detection system.

[0014] A further known system for detecting and predicting transitions among wakefulness, drowsiness and sleep phases using intact photoplethysmography (PPG) signals is disclosed in WO 2020 / 043855. Summary of the Invention [Problem to be solved by the invention]

[0015] Applicant states that in modern wearable devices, PPG signals are collected more frequently and processed internally by dedicated hardware circuitry that optimizes cost-performance characteristics, and the same is true for non-contact sensors (e.g., RF, RADAR) where only processed biometric data is available.

[0016] The objective of the present invention is to provide a new approach through statistical analysis applied to a reduced set of biometric data to predict real-time sleep onset on a wide range of modern edge devices, both contact and contactless. [Means for solving the problem]

[0017] In accordance with the present invention, software and electronic processing resources are provided for real-time detection and / or prediction of a subject's behavioral state and / or transitions among one or more of the wakefulness (W), drowsiness (D) and sleep (S) phases, as claimed in the accompanying claims. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is a diagrammatic representation of a system comprising electronic processing resources according to the present invention; [Figure 2] 1 is a diagrammatic representation of software implementing 4KSS level classification according to the present invention. [Figure 3] 1 shows a software block diagram according to the present invention. [Figure 4] 1 is a diagrammatic representation of the mapping of architecture in an integrated and distributed manner according to the present invention; [Figure 5] 1 shows a block diagram of an integrated architecture in accordance with the present invention; [Figure 6] 1 shows a block diagram of a distributed architecture in accordance with the present invention; [Figure 7] 1 shows a distributed architecture applied to a first use case according to the present invention. [Figure 8]1 shows a distributed architecture applied to a second use case according to the present invention. [Figure 9] 1 is a diagrammatic representation of the mapping of architecture in a distributed and remote manner in accordance with the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0019] The present invention will now be described in detail with reference to the accompanying drawings so that those skilled in the art can implement and use it. Various modifications of the described embodiments will be readily apparent to those skilled in the art, and the general principles described may, however, be applied to other embodiments and applications without departing from the scope of protection of the present invention as defined in the appended claims. Therefore, the present invention should not be considered limited to the embodiments described and shown herein, but should be accorded the widest scope of protection consistent with the features described and claimed herein.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. In case of conflict, the present specification, including the definitions provided, will control. Furthermore, the examples are provided for illustrative purposes only and therefore should not be considered limiting.

[0021] In particular, the block diagrams contained in the accompanying drawings and described below should not be understood as representations of structural features, i.e., structural limitations, but as representations of functional features, i.e., inherent characteristics of a device defined by functional limitations that can be implemented in various ways to preserve its functionality (operational capabilities). To facilitate an understanding of the embodiments described herein, reference will be made to certain specific embodiments and specific language will be used to describe them. The terminology used herein is used for the purpose of describing specific embodiments only and is not intended to limit the scope of the present invention.

[0022] In particular, as will be explained below, the present invention aims to extend the possibility of accurately predicting sleep onset to a wide variety of systems where a complete PPG signal is not accessible or is not available in satisfactory quality, thereby satisfying a wider range of application areas.

[0023] Furthermore, the present invention is based on a simplified model of the autonomic nervous system (ANS), which controls the behavior of the cardio-respiratory system (CRS). In particular, CRS behavior is monitored in real time by an algorithm for sleep prediction. This algorithm processes a small number of physiological parameters measured through contact-based sensing devices (e.g., custom or consumer smartwatches) or through contactless sensing devices (e.g., RF sensors, such as RADAR or camera-based sensors) to obtain several statistical parameters. Based on the variation of such statistical parameters over time, the present invention enables classification of a person's drowsiness state and identifies the onset of sleep.

[0024] Thus, as will be explained below, the present invention introduces a methodology for processing physiological data from contact or non-contact detection solutions and obtaining a multilevel classification of a person's drowsiness; according to the present invention, reference is made, without limitation, to a four-level scale derived from the KSS, in particular the four-step KSS. In any case, the present invention can be applied to any newly defined drowsiness assessment scale. In this way, the present invention makes it possible to both provide a more accurate measure in the drowsiness domain, which by definition addresses non-binary processing, and to accurately identify whether a subject is alert (i.e., the subject is able to perform the task perfectly) or sleepy (i.e., the subject is no longer able to perform the task in a controlled manner).

[0025] Figure 1 shows a block diagram of System 1. System 1 includes: a sensing unit 2 configured for communication with a sensory system, the sensing unit 2 comprising either a wearable sensor 3 (e.g., a smartwatch) and / or a contactless sensor 4 (e.g., RADAR, camera) for receiving a biometric signal of the subject; an electronic computational resource 10 comprising a processing unit 5 configured to communicate with the sensing unit 2 to receive the biometric signal and to store, load and execute software or algorithms 6 therein that, in use, output data relating to behavioral states and / or transitions between behavioral states of a subject based on the biometric signal; a feedback unit comprising either a local feedback unit 7 and / or a remote feedback unit 8, configured to communicate with the electronic computational resource 10 and to receive an output generated by the latter in order to provide feedback to an external resource, e.g. a further electronic device external to the system 1; Equipped with.

[0026] In particular, the software 6 is designed such that, when executed, the electronic processing resource 10 is configured to detect and / or predict in real time the behavioral state and / or transition of one or more of the subject's wakefulness (W), drowsiness (D) and sleep (S) phases. More particularly, and referring to Figure 3, the software 6, when executed, causes the electronic computing resource 10 to: receiving (block 20) ​​a subject's biometric signal from the sensory system, here via a sensing unit 2; - processing the received biometric signal (blocks 21-29) to classify the received biometric signal into one of different classes associated with behavioral states and / or transitions among one or more of the subject's wakefulness (W), drowsiness (D) and sleep (S) phases; Detecting and / or predicting the subject's behavioral state and / or transitions among wakefulness (W), drowsiness (D) and sleep (S) phases based on the classified biometric signals (Block 30). It is designed to be configured as follows.

[0027] In particular, when the software is executed, it causes the electronic computing resource 10 to: - calculating (block 24) at least a first resultant quantity for and based on the received biometric signal; - calculating (block 29) at least one threshold value Th for the received biometric signal based on the at least one first combined quantity calculated therefor, - configured to classify (block 29) the biometric signal into one of different classes associated with behavioral states and / or transitions among one or more of the wakefulness (W), drowsiness (D) and sleep (S) phases based on a threshold Th calculated therefor.

[0028] Here, sensing unit 2: providing a physical interface between the sensors 3, 4 and the electronic computational resources 10, extracting physiological data or variables from the received biometric signals, - transferring physiological data to an electronic processing resource 10, which is configured to load and execute software 6 so as to output the data to be presented to the end user through a feedback unit as a local and / or remote feedback unit; It should be noted that the

[0029] As will become apparent in the following sections, with reference to Figure 2, software 6 is configured to receive physiological data, process it, and determine in real time the subject's drowsiness state according to a reduced KSS scale based on four levels or classes, herein labeled KS1, KS2, KS3, and KS4 for alertness (KS1), drowsiness (KS2, KS3), and sleep (KS4), respectively. Applicant states that the compression of the KSS scale from the known ten levels to the currently used four levels better reflects the subject's behavioral state (e.g., KS1 is associated with alertness and KS4 with sleep), while focusing more on the gray areas describing drowsiness states or regions (represented here by classes KS2 and KS3) that are most relevant for in-cabin monitoring applications, and in particular for Drowsiness Monitoring Systems (DMS).

[0030] As anticipated above, each biometric signal comprises either a subset of the physiological data or variables derived from the aforementioned physiological data. It should be noted that the current software 6 is configured to process different physiological data depending on the capabilities of the adapted sensing solution (either contact or contactless), and according to one embodiment of the invention, the physiological data and / or variables derived from the aforementioned physiological data comprise values ​​indicative of at least one of cardiac, respiratory or ocular outputs, advantageously heart rate, heart rate variability (HRV), respiration rate (RR), respiration amplitude, eye blink or gaze.

[0031] In other words, the current electronic computing resources 10 are configured to process physiological data, subsets thereof, and combinations thereof.

[0032] Also as shown in FIG. 3, the current electronic computing resource 10 is configured to execute the software 6 in a continuous loop, in particular according to the following steps:

[0033] Referring first to blocks 20-21, according to a pre-processing phase, the physiological data extracted from the biometric signal is validated by performing range checks on the input data, utilizing Quality of Service (QOS) metrics, if possible, provided directly by the sensory system, and in particular, if the quality of the physiological data is deemed insufficient to effectively quantify the drowsiness level, the physiological data is discarded and the end user is notified of the inability to provide service and the next iteration of the loop is initiated. Otherwise, if the signal quality is deemed sufficient, the following steps are executed:

[0034] Continuing with reference to blocks 22-26, according to the composite variable calculation step, the software is designed to, when executed, configure the electronic computing resource 10 to calculate at least one first composite quantity for the received biometric signal and based thereon. In particular, the composite quantity or variable used to classify drowsiness is calculated by means of a first-in-first-out (FIFO) queue of size "n" implemented through a circular buffer. Given a sequence of n physiological measurements X = {x1, x2, ..., xn}, according to one embodiment of the present invention, two statistical parameters are evaluated: the corrected sample standard deviation for X and the mean of X. The two statistical parameters are combined into a composite quantity XV as the ratio of the corrected sample standard deviation of X to the mean of X. Thus, according to one embodiment of the present invention, the composite quantity XV is calculated by a statistical diffusion index, i.e., the standard deviation σ 2 is.

[0035] For HRV physiological data, the resultant volume HRVV is calculated as follows:

[0036]

number

[0037] For RR physiological data, the resultant quantity RRV is calculated as follows:

[0038]

number

[0039] The calculated composite quantity XV is stored in a FIFO queue of size "w".

[0040] In other words, the electronic computing resource 10 has two FIFO queues when the software 6 is running: a first queue for X, storing the last n physiological measurements (i.e., n physiological data), each measurement being collected currently, for example, every second; and a second queue for the composite quantities XV, storing the last w composite variables (i.e. the w composite quantities XV) from which statistical parameters for X are obtained, here the corrected standard deviation and the ratio of the mean, The system is configured to implement and utilize the following:

[0041] Referring now to blocks 27-28, according to a calibration step, if predetermined permissive conditions are met (here, upon initialization of software 6 or upon user request), electronic computing resource 10 is programmed to define a baseline state of the subject from which biometric signals are collected. Specifically, the threshold or baseline Th is defined as a function of the contents of the second cue, and thus threshold Th is a single value that references statistical characteristics of the mean and corrected standard deviation of predetermined physiological data for the subject. Applicant thus states that threshold Th is a snapshot of the subject's cardiopulmonary state at a predetermined time. Given the predetermined threshold Th, subsequent processing steps enable electronic computing resource 10 to analyze changes in the subject's cardiopulmonary state behavior with respect to the predetermined threshold Th; further, according to one embodiment of the present invention, since baseline cardiopulmonary activity may change over time, calibration may be repeated as deemed necessary by either electronic computing resource 10 or the user.

[0042] Continuing with reference to block 29, according to a processing step, the subject's drowsiness level, defined as a drowsiness index (DOD), is calculated. In particular, to classify the biometric signal into one of different classes relating to behavioral states and / or transitions among one or more of the wakefulness (W), drowsiness (D) and sleep (S) phases based on the calculated threshold therefor, the software when executed causes the electronic computing resource 10 to: - calculating (block 29) a drowsiness index DOD based on the at least one resultant quantity and the calculated threshold value, - classifying the received biometric signal into one of different classes based on the calculated drowsiness index DOD (block 29), It is designed to be configured to

[0043] It should be noted that in an observation window of size w, which is the size of the second queue, the composite variable XV is lower than the threshold Th determined in the calibration step many times. Since the composite variable XV captures the corrected standard deviation and mean of the given physiological data, the current processing step constitutes a situation in which cardiopulmonary activity is transitioning from a state of high variability (i.e., threshold Th) to a state of low variability.

[0044] Finally, referring to block 30, according to a feedback step, a class according to the KSS scale is determined based on the value of DOD, i.e. an rKSS level is determined. In particular, in order to classify the received biometric signal into one of the different classes based on the calculated drowsiness index, the software, when executed, causes the electronic computing resource 10 to: determining that the received biometric signal is classified into a first class KS1, indicating a wakefulness (W) phase, i.e. an rKSS level equal to 1, if the drowsiness index is less than or equal to a predetermined value w divided by 2; determining that the received biometric signal is classified into a second class KS2, indicating a drowsiness (D) phase, i.e., an rKSS level equal to 2, if the drowsiness index is greater than a predetermined value w divided by 2 and less than or equal to a predetermined value w divided by 2 and added to a first parameter p; determining that the received biometric signal is classified into a third class KS3, indicating a drowsiness (D) phase, i.e., an rKSS level equal to 3, if the drowsiness index is greater than a value obtained by dividing a predetermined value w by 2 and adding the result to a first parameter p and less than or equal to a value obtained by dividing a predetermined value w by 2 and adding the result to a second parameter (r); determining that the received biometric signal is classified into a fourth class KS4, indicating a sleep (S) phase, i.e., an rKSS level equal to 4, if the drowsiness index is greater than a predetermined value w divided by 2 and added to a second parameter (r); It is configured as follows.

[0045] The first and second parameters p and r are constant values ​​of the sensitivity and specificity functions of the electronic computing resource 10 .

[0046] In particular, in the development of current software for sleep prediction, in the context of classification and subsequent feedback, it is important to note the following: -True Positive (TP) is the state when the subject falls asleep under test conditions and the electronic computational resource 10 predicts the event.

[0047] True Negative (TN) is the state when the subject under test conditions does not fall asleep and the electronic computational resource 10 predicts the event.

[0048] False Negative (FN) is the state when the subject under test conditions falls asleep and the electronic computing resource 10 does not predict an event.

[0049] False Positive (FP) is a condition when the subject under test conditions does not fall asleep and the electronic computing resource 10 predicts that the subject will fall asleep.

[0050] In summary, it should be noted that three numbers can be used: - Sensitivity or true positive rate is

[0051]

number

[0052] , defined as the ratio of positive predicted conditions (TP) to the number of actual positive conditions (TP+FN).

[0053] -Specificity or true negative rate is

[0054]

number

[0055] , defined as the ratio of negative predicted conditions (TN) to the number of actual negative conditions (FP+TN).

[0056] -Accuracy is

[0057]

number

[0058] , i.e., the ratio of correctly predicted conditions to actual conditions.

[0059] Therefore, according to the present invention, the electronic computing resource 10 is programmed to identify the variability of the physiological data over time, defined by the statistical parameters, the corrected standard deviation of X and the mean of X, with respect to an established baseline condition, i.e., a threshold Th. According to the predefined reduced KSS, a decrease in variability means an increase in the feeling of fatigue.

[0060] When defining the rKSS and therefore the class associated with the biometric signal, the drowsiness index DOD is set to zero and a new loop is initialized.

[0061] The present invention can be implemented in different architectural configurations, thereby expanding the possibilities for using the present invention in different contexts and applications. Architectural examples of the present invention are shown in Figures 4-9 and are briefly described below.

[0062] From an architecture point of view, and as depicted in Figures 4-6, both integrated and distributed solutions can be implemented. With particular reference to Figures 4a and 5 depicting an integrated solution, edge device 40 is configured to fully integrate the various sensors, i.e., the sensory systems of system 1, and the processing unit 5 of electronic computational resource 10, and furthermore, in this case, the feedback output by electronic computational resource 10 is generated locally. On the other hand, with reference to Figures 4b and 6 depicting a distributed solution, sensors of first device 41 are configured to communicate locally, for example through Wi-Fi, Bluetooth, RF signals or a wired connection, and transmit data to processing unit 5 stored in second device 42, which processing unit 5 is also configured to generate feedback.

[0063] Referring to the integrated solution shown in Figures 4a and 5, Applicant: -collecting biometric signals and therefore associated physiological data; - Executes software 6 that outputs feedback to the end user; - alerting the end user, e.g., the driver, by generating a relevant notification, e.g., in the form of a haptic signal and / or a visual signal that generates a visual notification to be displayed on the screen of the smartwatch; We conducted experimental validation of the integrated architecture using a modern smartwatch as an edge device40.

[0064] With reference to the distributed solutions shown in Figures 4b and 6, and as shown in Figures 7-9, it should be noted that the present invention can be applied to both short-range and long-range communications.

[0065] In particular, when considering short-range communication, the end user and system 1 are in close proximity, for example in an in-cabin automotive application or a monitoring room in a medical center, whereas long-range communication can enable new services for the end user as well as new Internet-of-Things (IoT) applications.

[0066] With reference to Figure 6, the Applicant has carried out an experimental validation of the distributed architecture in a fleet of heavy trucks with the participation of professional drivers, and with reference to Figure 7, devices 41 and 42 are respectively a smartwatch and a smartphone connected to each other through a wireless connection. Device 41 is configured to collect biometric signals and therefore physiological data and transmit them to device 42 for processing according to the steps described in the previous paragraph. In particular, device 42 is configured to execute software 6 in real time and to communicate with a center remote to both devices 41 and 42 in order to provide feedback to the end user. In that case, interaction with the driver can be in two ways: 1) through a smartphone, for example through a dedicated application installed on it, and / or 2) Through an infotelematic system that connects to a remote center, It will be held.

[0067] Referring to Figure 8, applicant has conducted further experimental validation of the distributed architecture on a static vehicle simulator, and in particular in the case of Figure 8, two different sensors, here device 41 and specifically a wearable device such as a smart watch, are wirelessly connected to an embedded platform, here device 42 and specifically a RADAR sensor.

[0068] A further architectural implementation of the present invention is shown in FIG. 9, where remote communication is envisaged and established between different devices where sensing, processing and feedback generation are performed independently.

[0069] The present invention has several advantages.

[0070] In particular, the present solution allows accurate prediction of sleep onset in a wide variety of systems where the complete PPG signal is not accessible or not available in satisfactory quality, and in particular, according to the invention, using only a small number of selected biometric parameters collected from or derived from sensors.

[0071] Furthermore, the present solution implements statistical methods to develop a reliable automatic method for predicting sleep onset, thereby compensating for the relatively low quality of the collected / derived biometric signals.

[0072] Furthermore, the present invention allows users to use lower quality signals and less specific acquisition parameters (e.g., sampling frequency, resolution, noise reduction, artifacts, etc.), thereby expanding the utility of the present invention across a wider range of applications and use cases. [Explanation of symbols]

[0073] 1 System 2 Sensing Unit 3. Wearable sensors 4. Non-contact sensors 5 Processing Unit 6 Algorithms 7 Local Feedback Unit 8 Remote Feedback Unit 10. Electronic Computing Resources 40 Edge Devices

Claims

1. software storable on and executable by an electronic computing resource (10), the software being designed, when executed, to configure the electronic computing resource (10) to predict in real time a behavioral state and / or transition of one or more of a subject's wakefulness (W), drowsiness (D) and sleep (S) phases; When the software is executed, it causes the electronic computing resource (10) to: - receiving (20) a subject's biometric signal from a sensory system (3, 4) through a sensing unit (2) in communication with said electronic computational resource (10); - processing (21-29) the received biometric signal to classify it into one of different classes related to the behavioral state and / or transition of one or more of the subject's wakefulness (W), drowsiness (D) and sleep (S) phases; - detecting and / or predicting (30) the subject's behavioral state and / or transition among wakefulness (W), drowsiness (D) and sleep (S) phases based on the classified biometric signals; It is designed to be configured so that the sensory system comprises a wearable sensor (3) and / or a contactless sensor (4) configured to output a respective biometric signal; each biometric signal alternatively comprises a subset of physiological data or a variable derived from said physiological data, said physiological data and / or said variable derived from said physiological data comprising a value indicative of at least one of cardiac, respiratory or ocular output, advantageously heart rate, heart rate variability (HRV), respiration rate (RR), respiration amplitude, eye blink or gaze; The software, when executed, causes the electronic computing resource (10) to: - calculating (24) at least a first resultant quantity for and based on said received biometric signal, - calculating (29) at least one threshold value for the received biometric signal based on the at least one first combined quantity calculated therefor, - classifying (29) said biometric signal into one of different classes related to said one or more behavioral states and / or transitions among wakefulness (W), drowsiness (D) and sleep (S) phases, based on thresholds calculated therefor, It is designed to be configured so that The software, when executed, causes the electronic computing resource (10) to classify (29) the biometric signal into one of different classes associated with the one or more behavioral states and / or transitions among wakefulness (W), drowsiness (D) and sleep (S) phases based on a threshold calculated therefor: - calculating (29) a drowsiness index DOD based on said at least one combined quantity and said calculated threshold, - classifying (29) the received biometric signal into one of different classes based on the calculated drowsiness index DOD, It is designed to be configured so that In order to classify the received biometric signal into one of different classes based on the calculated drowsiness index, the software, when executed, causes the electronic computing resource (10) to: - determining that the received biometric signal is classified into a first class (KS1) indicative of a wakefulness (W) phase if said drowsiness index (DOD) is less than or equal to a predetermined value (w) divided by 2; - determining that the received biometric signal is classified into a second class (KS2) indicative of a drowsiness (D) phase if the drowsiness index (DOD) is greater than the predetermined value (w) divided by 2 and less than or equal to the predetermined value (w) divided by 2 and added to a first parameter (p); - determining that the received biometric signal is classified into a third class (KS3) indicative of a drowsiness (D) phase if the drowsiness index (DOD) is greater than the predetermined value (w) divided by 2 and added to the first parameter (p) and is less than or equal to the predetermined value (w) divided by 2 and added to a second parameter (r); - determining that the received biometric signal is classified into a fourth class (KS4) indicative of a sleep (S) phase if the drowsiness index (DOD) is greater than the predetermined value (w) divided by 2 and added to the second parameter (r); It is designed to be configured so that The first parameter and the second parameter (p, r) are constant values ​​of a function of the sensitivity and specificity of the electronic computing resource (10); Software characterized by:

2. 2. The software of claim 1, wherein the software is designed to, when executed, configure the electronic computing resource to calculate at least a first one composite quantity for the received biometric signal and based on a sample thereof, for calculating at least a first one composite quantity for the received biometric signal and based on a sample thereof.

3. The software of claim 2 , wherein the composite quantity is a statistical diffusion index.

4. 1. An electronic computing resource (10) configured for real-time prediction of a subject's behavioral state and / or transitions among one or more of the wakefulness (W), drowsiness (D) and sleep (S) phases, the electronic computing resource (10) being configured to store, load and execute the software of claim 1 in order to perform the operations of claim 1.

Images