Detection of blood alcohol concentration in a subject

Abstract

System (1, 50, 100) for detecting the alcohol concentration in the blood of subject comprising: a sensory system (2) configured to generate the at least one physiological signal of a subject indicative of blood volume changes in the microvascular bed of the tissue of the subject; and electronic processing resources (5). The electronic processing resources (5) are configured to: receive the at least one physiological signal from the sensory system; process the received at least one physiological signal to determine a level of alcohol concentration in the blood of the subject; and provide assistance to the subject based on the determined level of alcohol concentration in the blood of the subject. The electronic processing resources (5) are configured to process the received at least one physiological signal to determine a level of alcohol concentration in the blood of the subject by classifying the at least one physiological signal into one or more classes indicative of the condition of the subject.

Description

[0001] DETECTION OF BLOOD ALCOHOL CONCENTRATION IN A SUBJECT

[0002] Cross-Reference to Related Applications

[0003] This Patent Application claims priority from European Patent Application No. 24219190.6 filed on December 11, 2024 and Italian Patent Application No. 102025000032692 filed on December 10, 2025, the entire disclosure of which is incorporated herein by reference.

[0004] Technical Field of the Invention

[0005] The present invention generally relates to monitoring the condition of a subject, in particular of a driver of a vehicle; more in particular, the present invention relates to a system for detecting the alcohol concentration in the blood of a subject, in particular the driver of a vehicle, and a related computer program product or software.

[0006] Background of the Invention

[0007] As is known, Driving Under the Influence (DUI) significantly threatens human life and public safety; furthermore, with an increased number of vehicles on the roads, manual checking for cases of drunk driving using breathalyzers becomes ineffective.

[0008] In further detail, alcohol is a component found in alcoholic beverages and is one of the various central nervous system depressants, acting as a psychoactive substance. If one engages in drunk driving, it can lead to issues such as impaired visual function, decreased driving abilities, and cognitive impairments; accordingly, this poses a significant problem for safely operating a vehicle.

[0009] According to National Highway Traffic Safety Administration, part of the U.S. Department of Transportation, about 32% of all traffic crash fatalities in the United States involve drunk drivers, with Blood Alcohol Concentration (BAC) of .08 g / dL or higher. In 2022, 13,524 people were killed in these preventable crashes. On average, from 2013- 2022, about 11,000 people died every year in drunk-driving crashes, in US only. Extensive studies already characterized the effect of different BACs on drivers, as reported in Table 1 below (https: / / www.nhtsa.gov / riskv-driving / drunk-driving).

[0010] To address these issues, each country enacts laws and conducts alcohol enforcement using alcohol measurement devices. The most common breathalyser, which is used for alcohol measurement, poses hygiene concerns due to the need for a mouthpiece during testing; furthermore, invasive measurement methods involving blood sampling are highly accurate but time-consuming for concentration determination, making immediate detection of alcohol consumption difficult.

[0011] Known solutions provide systems and methods for measuring the alcohol in the blood of a subject by detecting air circulating within the vehicle; however, this method has the drawback of low accuracy as it can also detect substances containing alcohol, such as mouth freshener or vehicle fragrance.

[0012] Further known solutions have also been implemented.

[0013] For example, electrochemical alcohol measurement utilizes changes in electromotive force due to oxidation reduction reactions within an electrolyte; however, this method requires periodic calibration, is susceptible to external factors like temperature and humidity, and can experience device fatigue with repeated use.

[0014] Further solutions provide gas chromatography which involves injecting the target bodily fluid into a separation column; in this case, substances are separated by heat and enter individual detectors. The concentration of alcohol is determined based on the amplitude of detector responses. However, this method is costly, experimentally complex, not suitable for continuous measurements, and requires operator expertise.

[0015] Further solutions provide semiconductor-based measurement sensors which measure changes in conductivity due to adsorption and catalytic reaction. Nevertheless, these sensors exhibit nonspecific responses to alcohol, reduced sensitivity over time, and a limited sensor lifespan as measurement time progresses.

[0016] Other known solutions provide enzyme-based alcohol concentration measurement methods which utilize dehydrogenase enzymes to convert ethanol into acetaldehyde and reduced NADH (Reduced Nicotinamide Adenine Dinucleotide). The reduced NADH, which is proportional to the ethanol concentration, is measured using optical methods. However, this method has a limited range of alcohol concentration and limited accuracy. PCF (Photonic Crystal Fiber) alcohol sensing exhibits high measurement sensitivity due to significant changes in the optical refractive index but is influenced by temperature, affecting both refractive index and measurement sensitivity.

[0017] Further known solutions provide for OECT (Organic Electrochemical Transistor)- based enzyme reaction alcohol sensors which can be produced quickly and inexpensively but require continuous replacement as they are disposable devices; in particular, patch- based alcohol concentration measurement methods detect alcohol metabolites in sweat to sense alcohol concentration. However, the accuracy of alcohol concentration is low due to the influence of sweat, and the periodic replacement of the patch is necessary due to limited usage time.

[0018] According to “Leveraging driver vehicle and environment interaction: Machine learning using driver monitoring cameras to detect drunk driving" by Koch K. et al., an automatic approaches based on machine learning using driver monitoring cameras to detect drunk driving is proposed; specifically, the approach extracts information on gaze behaviour and head movements from driver monitoring cameras and then predicts whether drivers exceed two critical thresholds for BAC (Blood Alcohol Concentration). The abovementioned approach is based on indirect measurement of the effect of BAC and cannot identify the root cause of alterations to gaze / head movements: the same patterns of gaze / head movements can be produced by different root causes such as fatigue, drowsiness and BAC.

[0019] In general, any solution for determining BAC and its effects in a correct way must consider, in addition to its accuracy, also its compliance and acceptability from the user’s viewpoint; thus, the following concepts should be considered:

[0020] - compliance / adherence, which referred to the extent to which people follow a test’ s instructions and complete it as intended;

[0021] - usability, which refers to effectiveness, efficiency, and satisfaction in the specified context of use;

[0022] - acceptability, which refers to how appropriate / agreeable the test feels to users (comfort, privacy, stigma);

[0023] - tolerability, which refers to the level of discomfort user is willing to endure to complete the test; and

[0024] - feasibility, which refers to practicality of performing the test in the intended setting (home, roadside, clinic).

[0025] In general, high-compliance methods reduce invalid / insufficient rates, operator interventions, and total time-to-valid results, improving overall program performance.

[0026] Object and Summary of the Invention

[0027] The Applicant notes that the currently available techniques and methods are passible of improvements.

[0028] In particular, the Applicant notes that, although direct blood alcohol measurement via gas chromatographic is the standard, said method is impractical in use cases other than laboratory testing. For these reasons alternative methodologies were proposed to determine BAC from other bodily fluids, including saliva, urine, and sweat, and from the breath samples; however, the Applicant notes that the abovementioned further known methods are also impractical in the case of an automatic approach meant for being embedded in use cases such as automotive, where a system shall be capable of observing a person, of estimating the BAC, and of taking the proper course of action based on the BAC level.

[0029] An object of the present invention is thus providing systems and computer program products or software that allow to overcome at least in part the disadvantages of the known prior art.

[0030] According to the present invention, a system for evaluating the status, in particular detecting the alcohol concentration in the blood of a subject, in particular the driver of a vehicle, and a related computer program product or software are provided, as claimed in the appended set of claims.

[0031] Brief Description of the Drawings

[0032] Figure 1 schematically shows a system for detecting the alcohol concentration in the blood of subject according to the present invention.

[0033] Figure 2 schematically shows the computing chain implemented by a system for detecting the alcohol concentration in the blood of subject according to the present invention.

[0034] Figure 3 shows the Fourier transformation of the behavior of the amplitude of the PPG signal over frequency and the extraction of peak parameters thereof.

[0035] Figure 4 shows the Fourier transformation of the behavior of the amplitude of the PPG signal over frequency in sober and alcohol phases.

[0036] Figure 5 schematically shows the operation of a classification and BAC estimation module of a system for detecting the alcohol concentration in the blood of subject according to the present invention.

[0037] Figure 6 schematically shows an example of BAC tables based on RA and RF ratios.

[0038] Figures 7 and 8 schematically show further embodiments of a system for detecting the alcohol concentration in the blood of subject according to the present invention. Description of Preferred Embodiments of the Invention

[0039] The present invention will now be described in detail with reference to the accompanying drawings in order to allow a skilled person to implement it and use it. Various modifications to the described embodiments will be readily apparent to those of skill in the art and the general principles described may be applied to other embodiments and applications without however departing from the protective scope of the present invention as defined in the appended claims. Therefore, the present invention should not be regarded as limited to the embodiments described and illustrated herein but should be allowed the broadest protection scope consistent with the features described and claimed herein.

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning commonly understood by one of ordinary skill in the art to which the 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 as such should not be considered limiting.

[0041] In particular, the block diagrams included in the attached figures and described below are not to be understood as a representation of the structural features, i.e. constructional limitations, but must be understood as a representation of functional features, i.e. intrinsic properties of the devices defined by the effects obtained, that is to say functional restrictions, which can be implemented in different ways, so as to protect the functionalities thereof (operational capability).

[0042] In order to facilitate the understanding of the embodiments described herein, reference will be made to some specific embodiments and a specific language will be used to describe the same. The terminology used herein is used for the purpose of describing particular embodiments only and is not intended to limit the scope of the present invention.

[0043] To summarise, the present invention provides for a system detecting the alcohol concentration in the blood of a subject, in particular the driver of a vehicle, and a related computer program product or software. In particular, the abovementioned system is automatic and cost-effective and allows to estimate the Blood Alcohol Contents (BAC) level, based on the analysis of blood volume changes in the microvascular bed of tissue, by using contact and contact-less sensing technologies. Furthermore, as also disclosed in further detail below, the present invention can be extended to all domains where the person must be sober to accomplish his / her task due to mandatory normative (e.g. professional truck / bus drivers, airplane pilot, train driver and the like).

[0044] In particular, to provide an accurate estimation of the BAC, the present invention provides systems and computer program products for observing the physiological modifications alcohol produces to the cardiorespiratory system of individuals, as also observed in “ Advanced Spectroscopic Characterization of Impact of Alcoholic Intake on Variation in Blood-Pulse Waveform'" by Shimizu Y. el al..

[0045] In further detail, the present invention provides a non-invasive method, implemented by the instructions comprised in the abovementioned computer program product, and system that proposes an automatic framework to estimate or classify the Blood Alcohol Concentration (BAC) or alcohol -related physiological state of an individual based on the analysis of blood volume changes in the microvascular bed of tissue using photoplethysmograph (PPG) signal, via contact-based or contactless optical sensing technologies.

[0046] As also disclosed in further detail in the following paragraphs, unlike existing breath or sweat-based methods, the present system and computer program product are configured to analyse alcohol related physiological changes directly from the cardiovascular and microcirculatory system as reflected in the PPG waveform.

[0047] Furthermore, the present invention introduces a unified computational framework capable of operating with multiple sensing configurations, providing flexibility and robustness to environmental or user conditions:

[0048] - contact PPG: optical sensors embedded in smartwatches, rings, earphones, or armbands positioned on any skin region providing sufficient perfusion (e.g. finger, wrist, ear lobe); and

[0049] - camera-based PPG (remote PPG) either (a) by placing a fingertip over a smartphone camera illuminated by e.g. an integrated LED, or (b) by observing the subjects’ facial or neck regions though e.g. an external RGB or infrared camera positioned in the environment.

[0050] In addition, the present invention implements a SENSE-COMPUTE- ACT mechanism aimed at providing real-time information about the physiological state of the driver or user, and, where applicable, an estimation of the associated BAC level:

[0051] - SENSE: acquire signals, in particular optical ones, representing blood volume pulse variations, using in particular contact or contactless sensors;

[0052] - COMPUTE: process the acquired signal to extract spectral, temporal, and nonlinear features related to the cardiorespiratory response to alcohol; and - ACT: classify the subject’s state and optionally trigger vehicle safety measures, alerts or remote notifications or the like.

[0053] Following, the present invention allows to obtain the following results:

[0054] - inform the driver about his / her current alcohol level;

[0055] - provide or not the consensus to turn on the vehicle and start driving according to the measured automatically detect or infer alcohol level; and

[0056] - influence and eventually activate emergency or preventive actions to bring the system or vehicle into a safe state.

[0057] As better disclosed in the following paragraphs, the present invention combines the following features for achieving the evaluation of the levels of alcohol in the blood of a subject:

[0058] - spectral analysis of the vital signs, namely of the PPG signal, to assess frequency and harmonic changes caused by alcohol;

[0059] - Heart rate Variability (HRV) analysis (i.e. time-domain, frequency-domain, and nonlinear measures) to evaluate alcohol -related autonomic modulation;

[0060] - machine learning algorithms trained on labeled physiological data to improve discrimination between sober and alcohol-influenced states; and

[0061] - eventually comparing the obtained results with a baseline reference window representing the subject’s known sober condition, allowing for personalized calibration and improved reliability.

[0062] Figure 1 schematically shows a system 1 for detecting the alcohol concentration in the blood of subject

[0063] - a sensory system 2, here comprising either a contact-based or werable sensing unit 3 and / or a contactless sensing unit 4 configured to acquire physiological data of a subject and generate corresponding output signals, in particular at least one physiological signal of a subject indicative of blood volume changes in the microvascular bed of the tissue of the subject;

[0064] - electronic computing resources 5, here an automotive embedded platform in the vehicle driven by the subject, configured to communicate, e.g. via wireless connection (e.g. Wi-Fi or Bluetooth), with the sensory system 2 to receive the outputted signals (thereby receiving the acquired physiological data) and to store, load and execute, when in use, a computer program product or software or algorithm 6 therein to output data relative to a level of alcohol concentration in the blood of the subject; and

[0065] - an interface to vehicle system 7 configured to communicate with the electronic computing resources 5 and to receive the output generated by the latter to provide feedback to external resources, e.g. further electronic devices external to the system 1 and / or to the subject.

[0066] In further detail, the sensory system 2 is configured to measure the variations in the blood volume in the microvascular bed of tissue of the subject to detect the concentration of alcohol in the blood of the same subject; according to an aspect of the present invention, the at least one physiological signal comprises at least one plethysmograph, PPG, signal indicative of the blood volume changes in the microvascular bed of the tissue of the subject. Thus, according to an aspect of the present invention, the sensory system 2 exploits the photoplethysmography technology for allowing the determination of the level of alcohol in the blood of the subject.

[0067] As also shown in Figure 1, the sensory system 2 comprises at least one among contact-based and contactless sensing units configured to acquire physiological data related to the blood volume changes in the microvascular bed of the tissue of the subject and output the related at least one physiological signal. Exemplarily and without it being limiting to the present invention, the sensory system 2 of Figure 1 comprises both contactbased and contactless sensing units 3, 4. According to an aspect of the present invention, the contact-based sensing units 3 comprise sensors integrated in smartwatches, rings, earphones, bands or other wearable devices; in these cases, the integrated sensors are e.g. located to be in contact with the finger, wrist, arm, ear lobe, or any skin region with a dense microvascular network. Furthermore, according to an aspect of the present invention, the contactless sensing units 4 comprise sensors such as imaging-based PPG (for example, RGB or IR cameras); in particular, such sensors are configured to observe and monitor portion of the body of the subject, such as the face, forehead, or neck areas from a fixed or mobile position (e.g. inside a vehicle) to extract remote PPG (rPPG) signals via optical reflectance variations.

[0068] According to a further aspect of the present invention, not discussed in further detail in the following, the sensory system 2 may have a hybrid mode, meaning that, e.g. finger- on-camera hybrid sensing unit(s) may be present; in this case, for example and without it being limiting to the present invention, a smartphone camera and LED flash are used to acquire the PPG signal by having the user lightly rest a fingertip on the camera lens. In particular, the LED flash illuminates the skin while the camera measures variations in reflected or transmitted light caused by blood pulsations. Moreover, automatic detection of proper contact may ensure signal quality. According to another aspect of the present invention, the contactless sensing unit 4 may also be the camera integrated in a nomadic device (e.g. smartphone, tablet) able to perform an rPPG analysis. Furthermore, in case of an hybrid mode, data fusion and redundancy may be implemented to enhance reliability under varying conditions (e.g., motion, lightning, skin tome or sensor orientation).

[0069] Furthermore, the sensing units 3, 4 of the sensory system 2 carry out sensing modes which may operate independently and / or simultaneously with each other.

[0070] According to an aspect of the present invention, the electronic processing resources 5 are of a local type, e.g. they are integrated and / or implemented as a local unit; according to a further and alternative aspect of the present invention, the electronic processing resources 5 are of a distributed type, e.g, on nodes in a network, and implementing paradigms such as cloud computing. Without it being limiting to the present invention, the electronic processing resources 5 are hereinafter of the distributed type.

[0071] Considering the above, the system 1 according to Figure 1 entails a rapid deployment via wearable (e.g. smartwatch, smart band, smart ring, here the contact-based sensing unit 3) and integration with telematics / EHS / provider systems is possible. According to a further aspect of the present invention, not disclosed in further detail in the following, the sensory system 2 comprises only the contact-based sensing unit 3.

[0072] Referring to Figure 2, a method for determining the level of blood alcohol concentration (BAC) is now disclosed in the following paragraphs according to a nonlimiting aspect of the present invention.

[0073] In particular, the electronic processing resources 5 are configured to:

[0074] - receive the at least one physiological signal from the sensory system 2;

[0075] - process the received at least one physiological signal to determine a level of alcohol concentration in the blood of the subject; and

[0076] - provide assistance to the subject based on the determined level of alcohol concentration in the blood of the subject.

[0077] In further detail, the electronic processing resources 5 are configured to process the received at least one physiological signal to determine a level of alcohol concentration in the blood of the subject by classifying the at least one physiological signal into one or more classes indicative of the condition of the subject.

[0078] According to an aspect of the present invention, in particular referring to Figure 2, in order to process the received at least one physiological signal to determine a level of alcohol concentration in the blood of the subject, the electronic processing resources 5 are configured to: - process the at least one physiological signal to remove any motion artifact from the same at least one physiological signal;

[0079] - determine a periodic and harmonic behavior of the at least one physiological signal based on the processed at least one physiological signal; and

[0080] - extract one or more features characterizing the periodic and harmonic behavior of the at least one physiological signal.

[0081] In particular, referring to Figure 2, the electronic processing resources 5 comprises a pre-processing and motion reduction module 10 configured to receive the at least one physiological signal from the sensory system 2 to process it and remove any motion artifact from the same at least one physiological signal, hereinafter referred to as PPGinput. In further detail, the pre-processing and motion reduction module 10 is configured to process the abovementioned at least one physiological signal PPGinput to determine a pre- processed at least one physiological signal PPGoutput by applying the formula reported in Equation (1):

[0082] PPGoutput = PPG^put ~ mean PPGinput, N) (1) wherein mean()' is a function that allows to calculate the arithmetic means of N samples (where N is a natural number) of the at least one physiological signal PPGinput. Furthermore, in this case, the parameter N is used to tune the filtering effects of the same pre-processing and motion reduction module 10; exemplarily and without it being limiting to the present invention, N is here equal to 32.

[0083] According to further aspects of the present invention, not disclosed in further detail in the following, further, alternative and / or additional pre-processing operations aimed at ensuring that the acquired at least one physiological signal is of sufficient quality and provided with minimal or no motion artifacts, without limiting the use of alternative or complementary filtering or stabilization techniques.

[0084] According to another aspect of the present invention, signal quality indicators (e.g., spectral peak-to-noise ratio, kurtosis, skewness or motion magnitude thresholds) may be computed to automatically exclude analysis windows with excessive artifacts, without limiting the statistical or heuristic criteria used for quality gating.

[0085] Furthermore, according to an aspect of the present invention, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources 5 are configured to classify the extracted one or more features into the one or more classes indicative of the condition of the subject. According to a further aspect of the present invention, in order to determine a periodic and harmonic behavior of the at least one physiological signal based on the processed at least one physiological signal, the electronic processing resources 5 are configured to:

[0086] - determine a Fourier transformation of the at least one physiological signal to identify periodic and harmonic components of the cardiac rhythm and its modulations;

[0087] - determine a power spectral density of the Fourier transformation of the at least one physiological signal; and

[0088] - determine one or more harmonic components of the power spectral density of the Fourier transformation of the at least one physiological signal based on the power spectral density of the Fourier transformation of the at least one physiological signal.

[0089] In further detail, as also shown in Figure 2, the electronic processing resources 5 comprise a spectral transformation module 11 configured to computer a Fast Fourier Transform (FFT) of the pre-processed at least one physiological signal (i.e. PPGoutput) to determine periodic and harmonic components corresponding in particular to the cardiac rhythm and its modulations. According to an aspect of the present invention, the abovementioned transformation is implemented using a pre-stored algorithm in the electronic processing resources 5 (in particular the spectral transformation module 11), in particular a FFT algorithm. According to another aspect of the present invention, the electronic processing resources 5 are configured to use alternative mathematical transforms or frequency-domain estimators, such as one of the following

[0090] - Discrete Fourier Transform (DFT), Short Time Fourier Transform (STFT), Discrete Cosine Transform (DCT), or Discrete Sine Transform (DST);

[0091] - Lomb-Scargle periodogram, multitaper spectral estimation, Welch’s method, autoregressive (AR) or ARMA modeling, and Cepstrum analysis;

[0092] - Continuous or Discrete Wavelet Transform (CWT / DWT) for time-frequency localization; and

[0093] - Hilbert-Huang Transform (HHT) or Empirical Mode Decomposition (EMD) based instantaneous frequency extraction.

[0094] It is furthermore noted that the input signal, here the pre-processed at least one physiological signal PPGinput, is sampled with sampling frequency fsfor a period of time T, thereby leading to M samples, wherein M is determined as the product between the sampling frequency fsand the period of time T, which are processed using the FFT algorithm. Exemplarily and without it being limiting to the present invention, the sampling frequency fsis comprised between 15 Hz and 200 Hz; furthermore, exemplarily and without it being limiting to the present invention, the period of time T is comprised between 120 s and 300 s.

[0095] Furthermore, the FFT or equivalent transformation performed by the electronic processing resources 5 yields the frequency spectrum of the PPG waveform, allowing the identification of the fundamental frequency component associated with the heart rate and its harmonic structure.

[0096] To increase robustness, according to an aspect of the present invention, the electronic processing resources 5 are configured to implement windowing functions such as e.g. Hamming, Hann, Blackman, or Kaiser may be applied to the input segment prior to transformation (z.e. to the pre-processed at least one physiological signal PPGinput before the transformation); further windowing and tapering functions, not disclosed in further detail in the following, may also be implemented.

[0097] Furthermore, the electronic processing resources 5 are further configured to employ further functions such as e.g. spectral averaging, overlap-ass segmentation, or adaptive spectral tracking can also be employed to improve frequency resolution or noise robustness, without limiting the use of other equivalent signal segmentation strategies.

[0098] Thus, the result of the abovementioned operation is a spectral profile that can be expressed in magnitude, power, or logarithmic scales, depending on subsequent processing needs.

[0099] As also disclosed in further detail in the following, according to an aspect of the present invention, the electronic processing resources 5 are further configured to retain phase information to support coherence or phase-synchronization analyses between harmonics and may extract phase-related descriptors, such as phase variance or interharmonic phase shifts, for use in subsequent feature computation, without limiting the type of representation domain (z.e. amplitude, power, phase or complex spectrum).

[0100] According to Figure 2, the electronic processing resources 5 further comprise a power spectral density computation module 12 configured to determine the power spectral density of the Fourier transformation of the at least one physiological signal; in particular, the power spectral density computation module 12 is configured to determine the power spectral density according to the following Equation (2): <2)

[0101] Wherein X is the result of the Fast Fourier transform of the preprocessed at least one physiological signal PPGoutput and psd is the power spectral density.

[0102] According to a further aspect of the present invention, not disclosed in further detail in the following, the power spectral density computation module 11 is configured to implement further alternative or equivalent method for determining the power spectral density, such as:

[0103] - Welch’s averaged periodogram with selectable overlap and window parameters;

[0104] - multitaper estimation for variance reduction;

[0105] - parametric estimation using AR / ARMA models or Prony’s method;

[0106] - non-linear or adaptive spectral estimation, such as time-varying PSD tracking;

[0107] - Wavelet energy density or Wigner- Ville distribution for time-frequency characterization; and

[0108] - power ratio calculations between spectral sub -bands or normalized harmonic powers

[0109] According to an aspect of the present invention, without it being limiting to the present invention, the power spectral density computation module 12 is configured to determine the power spectral density psd over absolute frequency (Hz), normalized frequency or log-frequency scales, without limiting the frequency representation domain; according to a further aspect of the present invention, the power spectral density computation module 12 is configured to estimate the power spectral density psd in an averaged, smoothed, or normalized manner by total energy to enhance comparability between different sensing sessions, devices, or users.

[0110] The resulting power spectral density psd allows the identification of dominant spectral peaks (here corresponding to heart rate and harmonics), the distribution of power among low-, mid- and high-frequency bands, and shifts in spectral balance potentially indicative of alcohol-induced physiological effects; additional derived measures from the determined power spectral density psd may include spectral entropy, band power ratios, frequency dispersion, and spectral flatness, without limiting the use of other spectral metrics derived from the power spectral density psd.

[0111] According to a further aspect of the present invention, the power spectral density computation module 12 is configured to estimate the power spectral density psd separately for each color channel in camera-based PPG (z.e., the contactless sensing unit 4), for each optical wavelength in multi -wavelength sensors or for each spatial region of interest (ROI) in facial or fingertip imaging; it is furthermore noted that further single or multi-channel spectral estimation techniques may be used. In particular, the power spectral density computation module 12 is configured to fuse values of channel-wise and ROI-wise power spectral density psd through averaging, weighted combination, or principal-component projection, to enhance signal stability and reduce motion sensitivity; it is noted that further mathematical fusion methods may be used.

[0112] In further detail, it is noted that the power spectral density psd determined by the spectral density computation module 12 quantifies how the signal’s power is distributed over frequency and allows the extraction of features related to the cardiorespiratory modulation caused by alcohol intake; it is furthermore noted that further mathematical formulations for the estimation of the same power spectral density psd may be used.

[0113] The spectral density computation module 12 is further configured to determine the one or more harmonic components of the power spectral density of the Fourier transformation of the at least one physiological signal based on the power spectral density of the Fourier transformation of the at least one physiological signal; in particular, the spectral density computation module 12 is configured to identify and process one or more harmonic components from the power spectral density psd of the FFT transformation of the pre-processed at least one physiological signal, corresponding to the fundamental cardiac frequency and its higher harmonics. In further detail, the spectral density computation module 12 is configured to extract one or more quantities related to the harmonic components of the power spectral density psd, esemplarily and without it being limiting to the present invention amplitude, frequency, phase and power band of the first harmonic and subsequent orders (i.e. 2nd, 3rdand the like), as well as any derived or combined features thereof according to further aspects of the present invention.

[0114] According to a non-limiting aspect of the present invention, the spectral density computation module 12 is configured to compute at least one of the following parameters:

[0115] - amplitude ratios RA, for example determined as the ratio between the harmonic amplitudes of the first and second harmonics A1 / A2, of the first and third harmonics Ai / Asor any further harmonic ratios;

[0116] - frequency ratios RF, for example determined as the ratio between the frequencies of the first and second harmonics fi / fz, of the first and third harmonics fi / fa or a normalized frequency spacing; - energy ratios, for example determined as the ratio between the energies of the first and second harmonics, of the first and third harmonics or any further harmonic ratios;

[0117] - harmonic phase difference, phase-locking value, or spectral coherence;

[0118] - total energy, relative harmonic energy, or spectral slope between harmonics;

[0119] - harmonic entropy spectral centroid, and spectral bandwidth as measured of energy distribution;

[0120] - dominant frequency drift or instantaneous frequency variability across consecutive analysis windows; and

[0121] - harmonic balance index, defined as the ration between the sum of even and odd harmonic powers.

[0122] It is noted that further equivalent mathematical formulations may also be used.

[0123] According to an aspect of the present invention, the spectral density computation module 12 is further configured to calculate time-frequency representations, such as Short-Time Fourier Transform (STFT), continuous wavelet transform (CWT), or empirical mode decomposition spectra, to track transient spectral variations associated with alcohol-induced physiological changes, without limiting the invention to a specific transform or windowing technique.

[0124] Furthermore, it is noted that spectral features can further be expressed in normalized or logarithmic form, computed on absolute or relative scales, and combined using linear, nonlinear, or statistical operations.

[0125] In addition, according to an aspect of the present invention, the computed quantities are stored in a dedicated memory storage (not shown) integrated in the electronic processing resources 5 or transmitted as a feature set for further analysis.

[0126] In order to extract one or more features characterizing the periodic and harmonic behavior of the at least one physiological signal, the electronic processing resources 5 are configured to determine at least one among spectral features, temporal and morphological features, heart rate variability features, multimodal and fused features and statistical and machine learning features; consequently, in order to classify the extracted one or more features into the one or more classes indicative of the condition of the subject, the electronic processing resources 5 are configured to process the extracted features to determine the condition of the subject. In addition, according to an aspect of the present invention, the electronic processing resources 5 are further configured to acquire and process phase information related to the spectral components of the at least one physiological signal. In particular, the electronic processing resources 5 are configured to determine phase-based quantities, e.g. phase stability, phase variability and interharmonic phase differences, starting from the complex spectrum of the at least one physiological signal.

[0127] Referring to Figure 3, the latter shows, on a smoothed FFT of the at least one physiological signal, here represented as a graph A, how some frequency domain peaks parameters (e.g. height and width) are calculated; in this case, according to the method disclosed hereinafter, based on the BAC values, it is possible to identify different classes of data combining different feature analysis.

[0128] Furthermore, Figure 4 shows a possible contribution due to alcohol ingestion on the FFT waveform of the at least one physiological signal; in particular, graph B refers to the sober behavior of the subject and graph C refers to the alcohol behavior of the same subject. Thus, Figure 4 shows the difference between the sober and alcohol state of the subject in different situations, indicated as a) and b).

[0129] According to an aspect of the present invention, the electronic processing resources 5 are configured to consider several orders of harmonics for the calculation of the power spectral density and the related amplitude / frequency values, without limiting the use of additional orders of harmonics as resulting from experimental activity.

[0130] On this regard, according to an aspect of the present invention, the electronic processing resources 5 comprise a feature extraction module 13 and a feature combination, normalization and decision-preparation module 14. The feature extraction module 13 is configured to extract one or more features characterizing the periodic and harmonic behavior of the at least one physiological signal; furthermore., the feature combination, normalization and decision-preparation module 14 is configured to combine, normalize and process the abovementioned features in a manner disclosed in further detail below, in particular with reference to Figure 5 and steps S10-S16.

[0131] In further detail, after performing the abovementioned spectral analysis, as also partially anticipated above, the electronic processing resources 5 are configured to additional temporal, morphological, and statistical features to provide a multidimensional characterization of the waveform of the at least one physiological signal.

[0132] In particular, the feature extraction module 13 is configured to extract spectral features from the output provided by the spectral density computation module 12; in further detail, exemplarily and without it being limiting to the present invention, at least one among the following spectral features is determined:

[0133] - absolute and relative power within configurable frequency bands (for example and without limiting the frequency range or number of sub-bands considered, 0 Hz-0.5 Hz, 0 Hz-7-2.1 Hz, 2 Hz-4 Hz and 4 Hz-7 Hz);

[0134] - ratios between band powers (e.g. cardiac-to-harmonic, cardiac-to-noise, harmonic-to-total or VLF-to-total power);

[0135] - spectral entropy, spectral centroid and spectral flatness as measures of spectral distribution and waveform complexity;

[0136] - frequency drift, spectral slope and local bandwidth metrics around dominant frequencies or harmonics; and

[0137] - frequency drift, spectral slope and relative energy distribution across consecutive windows, without limiting the use of alternative frequency-domain measures.

[0138] In particular, the feature extraction module 13 is configured to extract temporal and morphological features from the output provided by the spectral density computation module 12; in further detail, exemplarily and without it being limiting to the present invention, at least one among the following temporal and morphological features is determined:

[0139] - time-domain waveform descriptors such as mean, variance, skewness, pulse amplitude, rise time, fall time and inter-beat interval (BBI) statistics;

[0140] - derived indices such as heart rate mean, BBI variability, pulse area, pulse width and interquartile range of amplitude or time between peaks;

[0141] - morphological indices from the pulse shape, without limiting the waveform decomposition technique; and

[0142] - moving-window or baseline-reference comparisons between a current segment and a previous “sober” segment to quantify relative physiological changes, without limiting the normalization or baseline modeling approach.

[0143] Furthermore, the feature extraction module 13 is configured to extract heart rate variability (HRV) features from the output provided by the spectral density computation module 12; in further detail, exemplarily and without it being limiting to the present invention, at least one among the following HRV features is determined:

[0144] - time-domain HRV measures, such as mean NN (mean value of normal-to-normal inter-beat intervals), SDNN (standard deviation of NN intervals), RMSSD (root mean square of successive differences between adjacent NN intervals), pNN50 (percentage of successive NN interval differences grater than 50 milliseconds), CVNN (coeffitient of variation of NN intervals);

[0145] - frequency-domain HRV measures, such as LF power, HF power, LF / HF ratio, normalized LF and HF power, total power;

[0146] - non-linear HRV measures: sample entropy, approximate entropy, Poincare SD1 / SD2 ratio and Poincare area;

[0147] - phase-related spectral descriptors, comprising e.g. phase stability, phase variability, or phase difference between harmonic components; and

[0148] - dynamic indices of autonomic balance, sympathetic / parasympathetic modulation, or stress level estimation, without limiting the physiological model or algorithmic derivation.

[0149] Furthermore, the feature extraction module 13 is configured to extract multimodal and fused features from the output provided by the spectral density computation module 12; in further detail, exemplarily and without it being limiting to the present invention, at least one among the following multimodal and fused features is determined:

[0150] - combination of features from multiple PPG channels, wavelengths or regions of interest (e.g. face and fingertip), with weighted or principal -component fusion to fusion strategy; and

[0151] - integration of PPG-based features with auxiliary data such as respiration rate, stress score, or accelerometer-derived motion indices.

[0152] Furthermore, the feature extraction module 13 is configured to extract statistical and machine learning (ML) features from the output provided by the spectral density computation module 12; in further detail, exemplarily and without it being limiting to the present invention, at least one among the following statistical and ML features is determined:

[0153] - feature normalization using z-score, median / IQR scaling or adaptive baseline correction;

[0154] - weighted feature aggregation, feature selection and dimensionality reduction via Principal Component Analysis (PCA), Independent Component Analysis (ICA) or autoencoders, without limiting the data-processing method; and

[0155] - non-linear combinations or mathematical transformations (ratios, logarithms, polynomial or exponential combinations) between extracted features to amplify alcohol related contrasts, without limiting the mathematical operation domain.

[0156] According to a further aspect of the invention, the abovementioned phase-based quantities are treated as additional features within the overall extracted feature set extracted by the feature extraction module 13; in particular, the electronic processing resources 5 are configured to exploit the abovementioned phase-based features may be exploited either as descriptors of alcohol-related physiological changes or as auxiliary indicators of signal quality for detecting or excluding motion affected analysis windows. The electronic processing resources 5 are configured to exploit said phase-based quantities either as additional features reflecting alcohol -related physiological changes or as auxiliary indicators of signal quality for detecting or excluding motion-affected analysis windows. Furthermore, in order to classify the extracted one or more features into the one or more classes indicative of the condition of the subject, the electronic processing resources 5 are configured to process the phase-based features together with the other extracted features to determine the condition of the subject, as also disclosed in the following.

[0157] Furthermore, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources are configured to:

[0158] - determine if the at least one physiological signal is indicative either of a sober or alcohol-influenced state; and / or

[0159] - determine if the at least one physiological signal is indicative of one or more values comprised in a predetermined range.

[0160] On this regard, the electronic processing resources 5 comprise a classification and BAC estimation module 15 configured to receive the determined features and classify the at least one physiological signal into the one or more classes indicative of the condition of the subject; in particular, as also disclosed in detail in the following, the classification and BAC estimation module 15 is configured to interpret the features to determine whether the physiological state of the subject corresponds to a sober or alcohol-influenced condition, or to an estimated BAC range.

[0161] According to an aspect of the present invention, the classification and BAC estimation module 15 is configured to classify the abovementioned features by relying on deterministic, statistical or data-driven methods; according to further aspects of the present invention, further classification methods may be used. Furthermore, according to an aspect of the present invention, the classification and BAC estimation module 15 is configured to process said phase-based features together with the remaining extracted features to determine the condition of the subject.

[0162] According to an embodiment of the present invention, the classification and BAC estimation module 15 is configured to implement at least one of the following classification methods: - rule-based thresholding methods, where selected features (e.g. harmonic ratios RA, RF; spectral entropy; HRV indices) are compared to fixed or adaptive thresholds defining sober, moderate or high-alcohol states;

[0163] - baseline references comparison methods, where each subject’s current features are contrasted with an individually calibrated baseline representing their sober condition, without limiting the normalization or statistical method;

[0164] - weighted scoring methods, in which multiple features are linearly or nonlinearly combined using pre-determined weights or coefficients derived from empirical data or optimization procedures;

[0165] - machine learning classification methods, including but not limited to logistic regression, Support Vector Machines (SVM), random forests, gradient boosting or neural networks trained on labeled datasets to discriminate between alcohol-influence and sober physiological patterns;

[0166] - blind or adaptive scoring methods, where a global or personalized score is computed by combining normalized feature values according to directionality and importance weights, without limiting the weighting or normalization formula; and

[0167] - quality gating and decision confidence weighting methods, using artifact, motion or signal quality indices to discard unreliable analysis windows or modulate detection thresholds.

[0168] In view of the above, the classification and B AC estimation module 15 is configured to output at least one among the following:

[0169] - a binary state indicator (alcohol detected / non detected), i.e. a class;

[0170] - a probabilistic confidence score; and / or

[0171] - an estimated BAC value expressed e.g. in g / dL or equivalent units, without limiting the output representation or scaling.

[0172] According to an aspect of the present invention, the decision outputted by the classification and BAC estimation module 15 may trigger adaptive alerts, vehicle control actions or communication with external system for driver safety management., thereby providing assistance to the user.

[0173] According to another aspect of the present invention, the classification and BAC estimation module 15 is configured to operate over consecutive temporal windows, applying sequential or voting-based decision rules to ensure temporal consistency and reduce false positives or false negatives; in fact, for example, an alcohol-influenced state may be confirmed only if a minimum number of consecutive analysis windows satisfy predefined detection criteria or exceed confidence thresholds, without limiting the number, duration, or overlap of such windows.

[0174] According to a further aspect of the present invention, the classification and BAC estimation module 15 is configured to dynamically update baseline or reference feature values (when used, as also disclosed in further detail in the following paragraphs with reference to Figure 5) during operation to account for gradual physiological or environmental changes, allowing real-time recalibration of thresholds or scoring functions.

[0175] Furthermore the electronic processing resources 5 comprise a BAC level module 16 configured to receive the output provided by the classification and BAC estimation module 15 and provide it to the user to provide assistance to the latter.

[0176] According to an aspect of the present invention, the electronic processing resources 5 are configured to map the extracted feature set, or any subset or transformation thereof, to an estimated alcohol-related physiological state or BAC value regardless of the mathematical, statistical, computational, or machine-learning formulation employed.

[0177] Reference is now made to Figure 5, wherein a method for classifying the received features by the classification and BAC estimation module 15 is shown and disclosed in further detail in the following.

[0178] In particular, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources 5, in particular the classification and BAC estimation module 15, are configured to:

[0179] - determine (block S10) whether the extracted features and a related analysis window are suitable for determining the condition of the subject;

[0180] - if the extracted features and a related analysis window are suitable for determining the condition of the subject, normalize and / or transform (block S12) the extracted features;

[0181] - evaluate (blocks S13) each normalized and / or transformed extracted feature to determine its contribution in determining the condition of the subject;

[0182] - aggregate (block S14) the determined evaluations to determine a composite score indicative of the condition of the subject;

[0183] - process (block S15) the aggregated score to generate a classification of the at least one physiological signal into the one or more classes indicative of the condition of the subject.

[0184] Furthermore, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources 5 are configured to compare (block Sil) current extracted features with previously extracted features referring to a sober condition of the subject before normalizing and / or transforming the extracted features.

[0185] In addition, according to an aspect of the present invention, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources 5 are configured to aggregate (block S16) decisions from consecutive analysis windows using temporal logic, in particular to improve the robustness over time of the system 1.

[0186] In particular, according to Figure 5, the classification and BAC estimation module 15 is configured to carry out sub-operations on the received features to determine the BAC level and, thus, provide assistance to the user; it is noted that the representation shown in Figure 5 is not to be considering as limiting to the present invention, as it is an exemplary embodiment and does not limit the invention to the specific ordering, internal logic, or implementation depicted; equivalent stages, additional steps, or reordered sequences may be employed.

[0187] In particular, referring to block S10, the classification and BAC estimation module 15 is configured to evaluate whether the analysis window and its associated features are suitable for reliable decision-making; in particular, the classification and BAC estimation module 15 is configured to perform quality and validity checks which include at least one among:

[0188] - motion and artifact indices derived from accelerometer / gyroscope data, spectral noise metrics, or PPG waveform morphology;

[0189] - HRV-related confidence criteria, such as valid-beat ratios, artifact percentages, or minimum usable window length;

[0190] - Statistical indicators such as peak-to-noise ratio, signal kurtosis, skewness, or distributional stability; and

[0191] - Inter-channel or inter-ROI consistency checks in camera-based PPG systems.

[0192] If quality criteria are not satisfied, the classification and BAC estimation module 15 is configured to reject, skip or mark with reduced confidence the current window; further rejection or weighting strategies may also be used.

[0193] Referring to block Sil, according to an embodiment of the present invention where personalized calibration is present, the classification and BAC estimation module 15 is configured to compare the current feature vector with a previously stored reference profile referring to a sober condition of the user. In particular, according to an aspect of the present invention, the baseline comprises at least one among:

[0194] - a single reference window;

[0195] - an average of multiple windows, and / or

[0196] - a dynamically updated baseline adapting to gradual physiological changes.

[0197] Furthermore, the classification and BAC estimation module 15 is configured to quantify a deviation using e.g. absolute differences, z-scores, Mahalanobis distance, correlation coefficients, or any statistical or machine-learning distance metric. In addition, baseline-referenced features may be forwarded to the next stage without limiting the normalization or comparison method implemented by the electronic processing resources 5.

[0198] Referring to block S12, the classification and BAC estimation module 15 is configured to standardize or transform the received features before actually perfoming an evaluation of the same for determining the BAC level; in particular and without it being limiting to the present invention, the classification and BAC estimation module 15 is configured to implement at least one among the following operations:

[0199] - z-score normalization, median / IQR scaling, or robust normalization;

[0200] - logarithmic, polynomial, or nonlinear transformations;

[0201] - dimensionality reduction or projection ( .g., PC A, ICA, autoencoder embeddings); and

[0202] - ratio-based, difference-based, or composite mathematical transformations.

[0203] It is noted that, according to further aspects of the present invention, the present system 1 does not provide limitation on which features are transformed or the choice of mathematical formulation.

[0204] Referring to block S13, the classification and BAC estimation module 15 is configured to evaluate each feature or selected subsets of features to determine its contribution to alcohol detection or BAC estimation; according to an aspect of the present invention, the classification and BAC estimation module 15 is configured to evaluate also the abovementioned phase-based features. In particular, according to a non-limiting aspect of the present invention, the classification and BAC estimation module 15 is configured to evaluate the abovementioned features or subsets thereof by implementing at least one of the following:

[0205] - carrying out a comparison with fixed, adaptive, or machine-learned thresholds;

[0206] - carrying out a computation of scoring functions reflecting physiological directionality or monotonic relationships;

[0207] - generating probabilistic or confidence values from per-feature statistical models;

[0208] - evaluating outputs from feature-specific machine-learning components; and / or

[0209] - evaluating derivation of intermediate metrics such as harmonic amplitude ratios (RA), frequency ratios (RF), spectral contrasts, HRV indices, or temporal variability measures.

[0210] Thus, the classification and BAC estimation module 15 is configured to output a set of intermediate feature-level metrics.

[0211] Referring to block S14, the classification and BAC estimation module 15 is configured to combine the intermediate feature-level metrics into a single alcohol- influence index or equivalent composite score; in particular, according to a non-limiting aspect of the present invention, the classification and BAC estimation module 15 is configured to implement the abovementioned aggregation by using linear or nonlinear methods, including:

[0212] - weighted summation, maximum / minimum operators, logical rules, or fuzzy aggregation;

[0213] - probabilistic fusion, Bayesian integration, or likelihood-ratio combination; and

[0214] - machine-learning aggregation models such as logistic regression, support vector machines, random forests, neural networks, or gradient-boosting ensembles.

[0215] It is noted that, without limiting the type of fusion strategy, weights or aggregation rules may be fixed, adaptive or learned from training data.

[0216] Referring to block S15, the classification and BAC estimation module 15 is configured to interpret the aggregated score to a classification of the abovementioned features and, thus, of the at least one physiological signal or generate a BAC estimate.

[0217] In particular, the classification and BAC estimation module 15 is configured to apply at least one among the following decision rules:

[0218] - deterministic thresholding for binary or multi-level alcohol-state classification;

[0219] - mapping of the global score to categorical BAC levels (e.g., “sober”, “low”, “moderate”, “high”);

[0220] - continuous regression models producing a BAC estimate in g / dL or equivalent units; and

[0221] - uncertainty estimation, probabilistic outputs, confidence measures, or soft decision logic.

[0222] It is noted that the number of categories, thresholds, or mapping functions that may be used by the classification and BAC estimation module 15 is not limited to the ones disclosed above. Furthermore, it is also noted that the operations at blocks S14 and S15 are also applied to the abovementioned phase-based features.

[0223] Referring to block S16, as also anticipated above, the classification and BAC estimation module 15 is configured to aggregate decisions from consecutive analysis windows using temporal logic to improve robustness over time. In particular, the classification and BAC estimation module 15 is configured to implement at least one of the following consolidation approaches:

[0224] - majority voting rules or N-of-M confirmation strategies;

[0225] - rolling averages, exponential decay weighting, or smoothing filters;

[0226] - sequential probability ratio testing or hysteresis-based decision stabilization; and

[0227] - adaptive rules requiring persistence of alcohol-related features across time.

[0228] Therefore, after the abovementioned steps disclosed with reference to blocks S10- S16, the classification and BAC estimation module 15 is configured to forward the obtained output to the BAC level module 16, the latter being configured to generate the final alcohol-state indication or numerical BAC estimate.

[0229] Referring to the amplitude and frequenct ratios RA, RF, reference is also made to Figure 6. In particular, it is noted that European countries have drink driving laws that are strictly enforced; however, they have different limits across the continent. Using BAC limits, where the amount of alcohol is present in the driver’s bloodstream, there are percentage limits which are used to determine if someone is over the limit; in particular, the limits go from zero alcohol, to low (0.03%-0.05%), moderate (0.05%), and higher (0.08%) and, typically, the large majority of European countries adopt the limit of 0.05%. In particular, it is also noted that 0.05% equates to 50 mg of alcohol in 100 ml of blood.

[0230] Thus, based on the BAC level of the driver, a calibratable classification table named BAC Table is generated and continuously updated during the driving mission; on this regard, Figure 6 shows a non-limiting example of such BAC table relating BAC to the abovementioned amplitude and frequency ratios, which thus provides an indication for the classification performed by the electronic processing resources 5.

[0231] It is noted that the threshold values, geometric regions, feature ratios and categorical boundaries depicted in Figure 6 are provided solely for explanatory purposes, i.e. for aiding the understanding of the present invention; thus, the abovementioned threshold values, geometric regions, feature ratios and categorical boundaries, as well as specific axes, ratios, thresholds, decision regions or the rectangular layout, depicted in Figure 6 are not be considered as limiting to the present invention.

[0232] Furthermore, in order to provide assistance to the subject, the electronic processing resources 5 are configured to, in particular based on the abovementioned BAC table, implement at least one among:

[0233] - activating the vehicle's interior control system (e.g. climate control, lighting, seat vibration, and music) aimed at regaining a proper consciousness level;

[0234] - interacting with the infotainment system to send audio messages; and / or

[0235] - interacting with a remote service center through a call.

[0236] Figures 7 and 8 show further embodiments of the system 1 shown in Figure 1; in particular, Figures 7 and 8 show systems 50, 100 whose elements which are common to the embodiments of Figures 7-8 and 1 are referred to with the same reference numbers and will not discussed in further detail below.

[0237] In particular, referring to Figure 7, the system 50 comprises only the contactless sensing unit 4, here e.g. a camera of a modern nomadic devices (e.g. smartphone, tablet), and is configured to be in communication with the electronic processing resources 5 via e.g. Wi-Fi or Bluetooth; in this case, the type of sensing unit 4 and the integration with telematics / EHS / provider systems allows to have an improved monitoring of the subject.

[0238] According to an aspect of the present invention, the system 50 shown in Figure 7 offers several advantages, included but not limited to:- rapid deployment, without requiring dedicated hardware installed in the vehicle;

[0239] - high user accessibility, as smartphones and tablets are widely available and familiar with users;

[0240] - reduced installation and maintenance costs, since no fixed sensors need to be integrated into the vehicle;

[0241] - flexibility and portability, allowing monitoring both inside and outside the vehicle when required; and

[0242] - facilitated integration with telematics, EHS or fleet-management provider platforms, enabling cloud-based supervision, compliance monitoring, or remote support without limiting the communication method used.

[0243] Furthermore, referring to Figure 8, the sensory system 2 of system 100 is integrated into the electronic processing resources 5 and only comprise the contactless sensing unit 4, here an rPPG camera; in particular, according to an aspect of the present invention, the contactless sensing unit 4 is fully integrated into the electronic processing resources 5, thus transferring the physiological data either directly from the sensory system 2 to the portion of the electronic processing resources 5 determining the B AC level or through the vehicle communication network (not shown).

[0244] The system 100 according to Figure 8 also offers advantages, including but not limited to:

[0245] - continuous and automatic monitoring without requiring any action from the user;

[0246] - passive sensing, improving compliance and allowing long-term or real-time supervision during driving;

[0247] - optimal camera positioning and lighting control, improving signal stability and reducing environmental variability compared to handheld devices;

[0248] - high reliability and robustness, as the sensing components are fixed and calibrated in the vehicle; and

[0249] - seamless integration with the vehicle’s safety systems, enabling immediate responses such as alerts, access control or emergency procedures.

[0250] From the disclosure above, the present invention has several advantages.

[0251] In particular, the present system 1 employs PPG signals, either from wearable (contact) or cameras (contactless), to eliminate consumables, allow passive capture and support remote workflows. Furthermore, in case of aftermarket applications, the PPG approach adopted by the present system 1 runs on standard devices (e.g. smartwatch, smart band, smart ring, smartphone camera), thereby avoiding the need to introduce further devices or consumables; this also allows to make it easy to deploy at scale with minimal upkeep.

[0252] Furthermore, the present invention avoids the need to have, like breath testing systems, dedicated instruments and disposable mouthpieces, which add cost, require stocking and sanitation, and increase operational overhead.

[0253] In addition, because the abovementioned PPG-based approach requires no forced exhalation, the present system 1 results to be more accessible for people with limited pulmonary capacity e.g., people with COPD or asthma).

[0254] Finally, it is clear that modifications and variations may be made to the object of the present patent application described and illustrated herein without departing from the protective scope of the present invention as defined in the appended claims.

Claims

CLAIMS1. System (1, 50, 100) for detecting the alcohol concentration in the blood of subject comprising:- a sensory system (2) configured to generate the at least one physiological signal of a subject indicative of blood volume changes in the microvascular bed of the tissue of the subject; and- electronic processing resources (5) configured to be in communication with the sensory system (2), the electronic processing resources (5) are configured to:- receive the at least one physiological signal from the sensory system;- process the received at least one physiological signal to determine a level of alcohol concentration in the blood of the subject; and- provide assistance to the subject based on the determined level of alcohol concentration in the blood of the subject, wherein the electronic processing resources (5) are configured to process the received at least one physiological signal to determine a level of alcohol concentration in the blood of the subject by classifying the at least one physiological signal into one or more classes indicative of the condition of the subject.

2. The system (1, 50, 100) according to claim 1, wherein the sensory system (2) comprises at least one among contact and contactless sensors (3, 4) configured to acquire physiological data related to the blood volume changes in the microvascular bed of the tissue of the subject.

3. The system (1, 50, 100) according to claims 1 or 2, wherein the at least one physiological signal comprises at least one plethysmograph, PPG, signal indicative of the blood volume changes in the microvascular bed of the tissue of the subject,4. The system (1, 50, 100) according to any one of the preceding claims, wherein, in order to process the received at least one physiological signal to determine a level of alcohol concentration in the blood of the subject, the electronic processing resources (5) are configured to:- process the at least one physiological signal to remove any motion artifact fromthe same at least one physiological signal;- determine a periodic and harmonic behavior of the at least one physiological signal based on the processed at least one physiological signal; and- extract one or more features characterizing the periodic and harmonic behavior of the at least one physiological signal, and wherein, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources (5) are configured to classify the extracted one or more features into the one or more classes indicative of the condition of the subject.

5. The system (1, 50, 100) according to claim 4, wherein, in order to determine a periodic and harmonic behavior of the at least one physiological signal based on the processed at least one physiological signal, the electronic processing resources (5) are configured to:- determine a Fourier transformation of the at least one physiological signal to identify periodic and harmonic components of the cardiac rhythm and its modulations;- determine a power spectral density of the Fourier transformation of the at least one physiological signal; and- determine one or more harmonic components of the power spectral density of the Fourier transformation of the at least one physiological signal based on the power spectral density of the Fourier transformation of the at least one physiological signal.

6. The system (1, 50, 100) according to any one of claims 4-5, wherein, in order to extract one or more features characterizing the periodic and harmonic behavior of the at least one physiological signal, the electronic processing resources are configured to determine at least one among spectral features, temporal and morphological features, heart rate variability features, multimodal and fused features and statistical and machine learning features, wherein the electronic processing resources (5) are further configured to determined phase-based quantities based on the at least one physiological signal, and wherein, in order to classify the extracted one or more features into the one or more classes indicative of the condition of the subject, the electronic processing resources (5) are configured to process the extracted features and the determined phase-based quantities to determine the condition of the subject.

7. The system (1, 50, 100) according to any one of the previous claims, wherein, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources (5) are configured to:- determine if the at least one physiological signal is indicative either of a sober or alcohol-influenced state; and / or- determine if the at least one physiological signal is indicative of one or more values comprised in a predetermined range.

8. The system (1, 50, 100) according to any one of the preceding claims, wherein, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources (5) are configured to:- determine (S10) whether the extracted features and a related analysis window are suitable for determining the condition of the subject;- if the extracted features and a related analysis window are suitable for determining the condition of the subject, normalize and / or transform (S 12) the extracted features;- evaluate (S 13) each normalized and / or transformed extracted feature to determine its contribution in determining the condition of the subject;- aggregate (S14) the determined evaluations to determine a composite score indicative of the condition of the subject;- process (S15) the aggregated score to generate a classification of the at least one physiological signal into the one or more classes indicative of the condition of the subject.

9. The system (1, 50, 100) according to claim 8, wherein, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources (5) are configured to compare (Sil) current extracted features with previously extracted features referring to a sober condition of the subject before normalizing and / or transforming the extracted features, and wherein, in order to classify the at least one physiological signal into the one or more classes indicative of the condition of the subject, the electronic processing resources (5) are further configured to aggregate (S16) decisions from consecutive analysis windows using temporal logic to improve robustness over time.

10. The system (1, 50, 100) according to any one of the preceding claims, wherein, in order to provide assistance to the subject based on the determined level of alcohol concentration in the blood of the subject, the electronic processing resources (5) are configured to implement at least one among:- activating a vehicle's interior control system (e.g. climate control, lighting, seat vibration, and music) aimed at regaining a proper consciousness level;- interacting with an infotainment system of the vehicle to send audio messages; and / or - interacting with a remote service center through a call.

11. Computer program product loadable in and executable by a system (1, 50, 100) for detecting the alcohol concentration in the blood of subject as according to any one of the previous claims and configured to cause, when executed, the system (1, 50, 100) to operate as according to any one of the preceding claims.

Images