Method and system for real-time monitoring of vehicle alcohol concentration and linkage with vehicle control

Abstract

This invention relates to the field of vehicle control technology, specifically disclosing a method and system for real-time monitoring of in-vehicle alcohol concentration and linkage with vehicle control. The invention acquires multispectral images of the driver's face and hand pressure data through in-vehicle sensors, generating a unique identifier through multimodal feature fusion and deep belief network encoding. Simultaneously, multi-channel gas sensors collect concentrations of ethanol, carbon dioxide, and interfering substances. Through intrinsic mode decomposition and confidence-weighted reconstruction, anti-interference odor and voiceprint features are formed. After associating the identifier with the odor and voiceprint, an alcohol risk index is calculated based on parameters such as the acetaldehyde-ethanol concentration ratio and basal metabolic rate. Then, road complexity is assessed based on vehicle speed, geographical location, and high-precision map data. Finally, a comprehensive risk coefficient is generated by fusing alcohol risk, vehicle speed, and road complexity, thereby implementing graded vehicle control, including warning prompts, power limiting, and automatic deceleration and parking, to achieve safety intervention.

Description

Technical Field

[0001] This invention relates to the field of vehicle control technology, and in particular to a method and system for real-time monitoring of vehicle alcohol concentration and linkage with vehicle control. Background Technology

[0002] Driving under the influence of alcohol is one of the main causes of road traffic accidents. In the current technology, the main measure to prevent drunk driving is breathalyzer testing before getting into the car. However, breathalyzer testing is a passive intervention, usually a one-time static test before the vehicle is started. It cannot be continuously monitored during driving, and therefore cannot address the driving risks caused by drivers drinking alcohol while driving (such as "leftover alcohol" or ingesting alcohol-containing items while driving). Therefore, a method that links real-time onboard alcohol concentration monitoring with vehicle control is needed to solve the above problems. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for real-time monitoring of vehicle alcohol concentration and linkage with vehicle control, so as to solve the technical problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: Methods for linking real-time monitoring of vehicle alcohol concentration with vehicle control include: Acquire data from multiple sources of sensors inside the vehicle, including biometric data and gaseous data. Facial and hand feature data are obtained based on the biometric data, and driver identification is obtained based on the facial and hand feature data. Based on the gas characteristic data, obtain ethanol concentration parameters and volatile organic compound parameters, and based on the volatile organic compound parameters, obtain interfering gas concentrations, and based on the ethanol concentration parameters and the interfering gas concentrations, obtain odor voiceprint characteristics; The driver's identity identifier and the odor and voiceprint features are associated to obtain the identifier odor and voiceprint features, and the alcohol risk index is obtained based on the identifier odor and voiceprint features; Acquire vehicle operating status data, and obtain vehicle speed and geographical location information based on the vehicle operating status data; The driving safety level is obtained based on the vehicle speed, the alcohol risk index, and the geographical location information, and the vehicle is controlled in conjunction with the driving safety level.

[0005] The beneficial effects of this application are as follows: This invention acquires multispectral images of the driver's face and hand pressure data through vehicle-mounted sensors, generates a unique identity identifier through multimodal feature fusion and deep belief network encoding, and simultaneously collects the concentrations of ethanol, carbon dioxide and interfering substances using multi-channel gas sensors. Through intrinsic mode decomposition and confidence-weighted reconstruction, anti-interference odor and voiceprint features are formed. After associating the identity identifier with the odor and voiceprint, the alcohol risk index is calculated by combining parameters such as the acetaldehyde-ethanol concentration ratio and basal metabolic rate. Then, the road complexity is assessed based on vehicle speed, geographical location and high-precision map data. Finally, the alcohol risk, vehicle speed and road complexity are integrated to generate a comprehensive risk coefficient, and graded vehicle control is performed accordingly, including warning prompts, power limiting and automatic deceleration and parking, to achieve safety intervention. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of a method flow according to an embodiment of this application.

[0007] Figure 2 This is a schematic diagram of the system structure according to an embodiment of this application.

[0008] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0009] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0010] like Figures 1-2 As shown, this application provides a method for real-time monitoring of vehicle alcohol concentration and linkage with vehicle control, including: S1. Acquire in-vehicle multi-source sensor data, wherein the in-vehicle multi-source sensor data includes biometric data and gas characteristic data; S2. Obtain facial feature data and hand feature data based on the biometric data, and obtain driver identification based on the facial feature data and hand feature data; S3. Obtain ethanol concentration parameters and volatile organic compound parameters based on the gas characteristic data, obtain interfering gas concentration based on the volatile organic compound parameters, and obtain odor voiceprint characteristics based on the ethanol concentration parameters and the interfering gas concentration. S4. Associate the driver identification and the odor and voiceprint features to obtain the identification odor and voiceprint features, and obtain the alcohol risk index based on the identification odor and voiceprint features; S5. Obtain vehicle operating status data, and obtain vehicle speed and geographical location information based on the vehicle operating status data; S6. Obtain the driving safety level based on the vehicle speed, the alcohol risk index, and the geographical location information, and perform linkage control on the vehicle based on the driving safety level.

[0011] As described in steps S1-S6 above, alcohol intake during driving directly affects a driver's reaction speed, judgment, and operational coordination. Furthermore, alcohol metabolism varies among individuals, and unexpected situations such as drinking while driving or "leftovers" may occur. Insufficient intervention in addressing the risks of drunk driving can lead to driving hazards. This invention first acquires data from multiple in-vehicle sensors. Biometric data comes from in-vehicle facial recognition sensors, hand grip pressure sensors, and other devices. The facial recognition sensor captures multispectral images of the driver's face, the hand grip pressure sensor collects pressure distribution data when the driver grips the steering wheel, and gas characteristic data is collected by in-vehicle gas sensors, including concentration parameters of gases such as ethanol, volatile organic compounds, and carbon dioxide. Comprehensive acquisition of the driver's biometrics and in-vehicle gas environment data provides the initial basis for subsequent identity verification and alcohol concentration analysis. The acquired biometric data is used to extract facial and hand features, and further to obtain the driver's identity information. In the specific implementation, pupil contraction frequency and facial microvascular pulsation rhythm feature vectors are first extracted from facial multispectral image sequences. These two features are then combined to obtain facial key point topological feature vectors. This is because pupil contraction frequency and facial microvascular pulsation rhythm change with the human body's physiological state and have individual specificity, which can assist in identity recognition. At the same time, spatiotemporal analysis is performed based on the hand grip pressure distribution matrix to obtain knuckle positioning coordinates and palm print texture feature vectors. Palm print texture, as an important component of biometrics, has stability and uniqueness, while knuckle positioning coordinates can reflect the dynamic characteristics of hand grip. Subsequently, these multimodal features are fused based on an attention mechanism. The attention mechanism can highlight the importance weight of different features in identity recognition, reduce redundant information interference, and improve the effectiveness of feature fusion. The resulting biometric fusion feature vector can comprehensively represent the driver's biometrics. Finally, the biometric fusion feature vector is input into a pre-trained deep belief network for identity feature encoding. Through the collaborative training of multiple layers of neurons, the deep belief network can efficiently extract identity information from the features and generate a unique driver identification code, laying the foundation for the subsequent association between alcohol risk and individual drivers. The acquired gas characteristic data were used to obtain the time series concentrations of acetone and benzene series compounds. Analysis of these two concentration time series helps determine the concentrations of interfering gases, as volatile organic compounds such as acetone and benzene series compounds may cross-interfere with ethanol gas detection, affecting the accuracy of alcohol concentration detection. Subsequently, the time series concentrations of ethanol and carbon dioxide were obtained from the gas characteristic data. A multi-time-frequency alignment algorithm was used to obtain the intrinsic mode function component set, which includes respiratory rhythm characteristic components, environmental interference noise components, and basal metabolic components. Multi-time-frequency alignment ensures consistency of different time series data in both time and frequency dimensions. Next, the signal-to-noise ratio (SNR) for each of the three components was calculated. Combined with the interfering gas concentration, the reliability weight coefficients for each component were determined. The SNR reflects the ratio of effective signal to noise in each component, while the interfering gas concentration affects the purity of the signal. The weight coefficients determined by combining these two factors accurately quantify the reliability of each component. Then, based on the weighting coefficients, the intrinsic mode function component set is reconstructed to remove invalid information such as environmental interference noise, highlight the effective features of ethanol concentration and carbon dioxide waveform, and obtain the ethanol concentration feature sequence and carbon dioxide waveform feature sequence. Finally, the odor soundprint feature is formed by splicing the features. After drinking, the driver will expel ethanol through breathing, and the ethanol concentration is correlated with the blood alcohol content. At the same time, the breathing rhythm and carbon dioxide emission will also change with alcohol intake. However, the interfering gases in the environment may cause the alcohol concentration detection to be distorted. Therefore, it is necessary to use multi-step feature extraction and reconstruction to remove interference and enhance the effective features, which can effectively improve the accuracy of ethanol concentration detection. The system associates driver identification with odor and voiceprint features. This association involves binding driver identification information with odor and voiceprint features within the same time window to ensure accurate alcohol concentration detection results. This results in an identified odor and voiceprint feature set. Subsequently, based on this feature, the peak ethanol concentration, peak acetaldehyde concentration, and carbon dioxide waveform consistency are obtained. The peak ethanol concentration directly reflects the highest concentration of alcohol in the driver's breath. Acetaldehyde, as an intermediate product of alcohol metabolism, has a concentration that, when compared to ethanol, reflects the driver's alcohol metabolism rate. Since alcohol is a central nervous system depressant, it significantly affects the respiratory center, leading to changes in breathing patterns. Typical manifestations include irregular breathing rhythm and waveform distortion. Therefore, by monitoring the similarity between the real-time carbon dioxide concentration waveform of the driver's breathing and the pre-stored standard carbon dioxide breathing waveform of the driver in a conscious and normal state, and using the similarity as the carbon dioxide waveform consistency, the driver's blood alcohol concentration can be quantified as exceeding the limit. By combining the basal metabolic rate with the acetaldehyde-ethanol concentration ratio, the carbon dioxide waveform consistency, and the peak ethanol concentration, an alcohol risk index is obtained. This transforms qualitative alcohol detection into a quantitative risk indicator, providing a comprehensive core basis for subsequent driving safety level determination. The system acquires vehicle operating status data and extracts vehicle speed and geographical location information. This data comes from devices such as wheel speed sensors, engine speed sensors, transmission output shaft speed sensors, inertial measurement units (IMUs), and GPS receivers. First, based on the wheel speed sensor data, engine speed data, and transmission output shaft speed data, a Kalman filter algorithm is used for data fusion. The Kalman filter algorithm effectively suppresses sensor noise, improves the accuracy of data fusion, and yields accurate vehicle speed. The rate of change of speed over a preset time period is then calculated as vehicle acceleration. Vehicle speed and acceleration directly reflect the vehicle's driving dynamics and are important indicators for assessing driving safety. Simultaneously, the inertial measurement unit acquires the vehicle's angular velocity and measures acceleration. Combined with the raw latitude and longitude coordinates obtained from the GPS receiver, geographical location information is calculated. This provides environmental and motion status data for driving safety level assessment. By combining alcohol risk index, vehicle speed, and geographic location information, driving safety level assessment and vehicle-linked control are achieved. First, road attribute data, including road type (e.g., urban roads, highways), speed limits, and traffic flow density characteristics, is retrieved from a pre-set high-precision map database based on geographic location information. This data reflects the road's traffic conditions and complexity. Then, a road complexity index is calculated based on road type, speed limit, and traffic flow density characteristics. This index quantifies the impact of the road environment on driving safety. Finally, vehicle speed, alcohol risk index, and road complexity index are weighted and calculated to obtain a comprehensive driving risk coefficient. The calculation formula is as follows: ; in, This indicates the overall driving risk factor. This indicates the vehicle speed weight value. Indicates vehicle speed. This indicates the weighting value of the alcohol risk index. This indicates the alcohol risk index. The road complexity index is a comprehensive factor that takes into account three key factors: the driver's alcohol status, vehicle dynamics, and road environment. It fully reflects the safety risks of the current driving scenario and controls the vehicle's speed and stops accordingly to avoid dangerous accidents.

[0012] In one embodiment, step S2, which involves obtaining the driver's identity based on the facial feature data and the hand feature data, includes: S21. Obtain the topological feature vector of the driver's facial key points based on the facial feature data; S22. Obtain the driver's hand grip pressure distribution matrix based on the hand feature data, and perform spatiotemporal domain analysis on the hand grip pressure distribution matrix to obtain the knuckle positioning coordinates and palm print texture feature vectors. S23. Based on the attention mechanism, multimodal feature fusion is performed on the facial key point topological feature vector, the knuckle positioning coordinates and the palm print texture feature vector to obtain a biological fusion feature vector; S24. Input the biometric fusion feature vector into the pre-trained deep belief network for identity feature encoding to generate a driver identity code, and use the driver identity code as the driver's identity identifier.

[0013] As described in steps S21-S24 above, the present invention acquires facial feature data from biometric data, and then identifies key areas of the face such as the corners of the eyes, the wings of the nose, and the corners of the mouth in the facial feature data through a key point detection algorithm. It calculates the topological relationships such as the relative positions and angles between each key point to form a facial key point topological feature vector. The facial key point topological feature vector can lay the foundation for improving the accuracy of subsequent identity recognition. The acquired biometric data, specifically hand features, is collected through an array of pressure sensors built into the steering wheel, which captures the pressure distribution data of the driver's hands gripping the wheel. This data forms a hand grip pressure distribution matrix, where the rows and columns correspond to the array coordinates of the pressure sensors, and the matrix elements represent the pressure values ​​at the corresponding locations. Subsequently, this pressure distribution matrix undergoes spatiotemporal analysis: in the time domain, a sliding window is used to extract the temporal changes in pressure values, identifying pressure fluctuations caused by finger joint movements, and combining this with a skeletal keypoint regression algorithm to determine the finger joint positioning coordinates. In the spatial domain, Gaussian filtering is used to remove pressure noise, and a texture extraction algorithm is employed to extract palm print texture features from the denoised pressure distribution matrix, forming a palm print texture feature vector. Thus, the driver's hand grip pressure distribution and finger joint positioning coordinates reflect the dynamic gripping habits of the hands, while the palm print texture represents the static biometric features of the hands. Both are individual-specific, and since the driver's hands are constantly in contact with the steering wheel during driving, they can continuously provide feature data, ensuring the continuity of identity recognition. Multimodal feature fusion is performed on facial key point topological feature vectors, knuckle positioning coordinates, and palm print texture feature vectors based on an attention mechanism. Specifically, the three feature vectors are converted into feature vectors of the same dimension by unifying their dimensions through fully connected layers with activation functions. Then, the three feature vectors of the same dimension are weighted and summed based on the attention mechanism, with each weight equal to 1. Finally, a biometric fusion feature vector is obtained. This ensures that the contribution of biometric features of different modalities to identity recognition varies, while also ensuring that the fused feature retains key information with high discriminative power. Biometric fusion feature vectors are input into a pre-trained deep belief network for identity feature encoding. This deep belief network consists of one visible layer, two hidden layers, and one output layer. The network pre-training stage employs an unsupervised contrastive divergence algorithm. In the identity encoding stage, the biometric fusion feature vector is input into the pre-trained network. After feature abstraction and compression through two hidden layers, the final output feature vector is the driver's identification code. The deep belief network possesses powerful feature learning and abstraction capabilities, extracting the most distinctive core features from the high-dimensional biometric fusion feature vector and compressing them into a low-dimensional identification code. Simultaneously, through collaborative analysis of facial and hand features, it covers multi-dimensional features including dynamic and static aspects, deep physiological aspects, and surface geometric aspects. The dynamic weight allocation of the attention mechanism strengthens the role of key features. The pre-trained deep belief network improves the accuracy and generalization ability of feature encoding, ultimately achieving continuous and accurate driver identification during driving.

[0014] In one embodiment, step S3, which involves obtaining the concentration of interfering gases based on the volatile organic compound parameters and obtaining odor voiceprint features based on the ethanol concentration parameter and the concentration of interfering gases, includes: S31. Obtain the acetone concentration time series and benzene series concentration time series according to the volatile organic compound parameters, and obtain the interfering gas concentration according to the acetone concentration time series and the benzene series concentration time series; S32. Obtain the ethanol concentration time series and carbon dioxide concentration time series based on the gas characteristic data, and obtain a multi-time-frequency aligned intrinsic mode function component set based on the ethanol concentration time series and the carbon dioxide concentration time series, wherein the intrinsic mode function component set includes respiratory rhythm characteristic components, environmental interference noise components and basal metabolic components. S33. Obtain the first signal-to-noise ratio corresponding to the respiratory rhythm feature component, and obtain the respiratory confidence weight coefficient based on the first signal-to-noise ratio and the concentration of the interfering gas. S34. Obtain the second signal-to-noise ratio corresponding to the environmental interference noise component, and obtain the environmental credibility weighting coefficient based on the second signal-to-noise ratio and the concentration of the interfering gas. S35. Obtain the third signal-to-noise ratio corresponding to the basic metabolic component, and obtain the basic confidence weight coefficient based on the third signal-to-noise ratio and the concentration of interfering gas. S36. Based on the respiratory confidence weight coefficient, the environmental confidence weight coefficient, and the basic confidence weight coefficient, the intrinsic mode function component set is reconstructed to obtain the ethanol concentration feature sequence and the carbon dioxide waveform feature sequence. S37. The ethanol concentration feature sequence and the carbon dioxide waveform feature sequence are spliced ​​together to obtain the odor voiceprint feature.

[0015] As described in steps S31-S37 above, the volatile organic compound parameters in the gas characteristic data to be obtained in this invention are analyzed by a gas chromatography sensor mounted in the vehicle. The acetone concentration time series and benzene series concentration time series are extracted separately. A weighted summation algorithm is used to fuse the acetone concentration time series and benzene series concentration time series to obtain the concentration of interfering gases. Specifically, the detected value of the acetone concentration time series at a certain time t is taken. The time-series detection values ​​of benzene series compound concentrations The concentration of interfering gas at time t after fusion is The formula for calculating the weighted sum is: ; in, This represents the concentration of the interfering gas at time t. The measured value of acetone concentration at time t represents the time series value. The weighted value represents the time-series detection value of acetone concentration at time t. The detected value of benzene series concentration at time t represents the time series value. By quantifying the concentration of interfering gaseous substances, it can provide data basis for subsequent elimination of the influence of acetone and benzene series on ethanol detection results. Ethanol concentration time series data were collected using an ethanol sensor and carbon dioxide concentration time series data were collected using a carbon dioxide sensor. Then, the Empirical Mode Decomposition (EMD) algorithm was used to perform multi-component decomposition on these two time series data to obtain a set of intrinsic mode functions (IMFs) aligned to multiple time frequencies. Among them, the IMFs were identified based on the multi-time frequency aligned IMFs and three representative components corresponding to respiratory rhythm features, environmental noise interference, and basal metabolism were selected. The multi-component decomposition can separate effective and ineffective components, which provides the possibility for subsequent enhancement of effective features and elimination of ineffective features. The first signal-to-noise ratio (SNR) is calculated using a signal-to-noise ratio (SNR) calculation algorithm (SNR = signal energy / noise energy, where signal energy is the energy of the periodic fluctuation portion of the component and noise energy is the energy of the random fluctuation portion). The concentration of interfering gas is then obtained. The first SNR reflects the purity of the effective signal in the respiratory rhythm feature component. The higher the SNR, the stronger the effective signal. However, the higher the concentration of interfering gas, the greater the possibility of the respiratory rhythm feature component being interfered with, and the lower the reliability. Therefore, it is necessary to combine the two weighted values ​​to determine the reliability weight of the respiratory rhythm feature component, providing a basis for the component weight allocation during subsequent feature reconstruction. Secondly, the second signal-to-noise ratio (SNR) corresponding to the environmental interference noise component is obtained. Based on the second SNR and the concentration of the interfering gas, the environmental reliability weight coefficient is obtained, and its corresponding second SNR is calculated (at this point, the lower the SNR, the stronger the invalid signal in the noise component, and the lower the reliability). Then, combined with the weighted calculation based on the interfering gas concentration, the environmental reliability weight coefficient is calculated. The environmental interference noise component is an invalid component in the gas data, and its reliability weight coefficient needs to be determined based on the noise intensity and the degree of environmental interference in order to reduce its weight in subsequent feature reconstruction and reduce interference with effective features. The corresponding third signal-to-noise ratio (SNR) is calculated (the basal metabolic component signal is stable, and the SNR is usually high, so the reliability is relatively stable). This is then weighted by the concentration of interfering gases to calculate the basic reliability weighting coefficient. The basal metabolic component reflects the physiological basal gas characteristics of a driver when they have not consumed alcohol and serves as a benchmark for determining alcohol intake. Its reliability needs to be assessed in conjunction with the degree of environmental interference to ensure the reliability of the benchmark reference. The obtained respiratory confidence weight coefficients, environmental confidence weight coefficients, and baseline confidence weight coefficients are used to reconstruct the features of the intrinsic mode function component set. Specifically, each component is multiplied by its corresponding confidence weight coefficient. The respiratory rhythm feature component is multiplied by a higher respiratory confidence weight coefficient to enhance effective features, the environmental interference noise component is multiplied by a lower environmental confidence weight coefficient to weaken ineffective features, and the basal metabolic component is multiplied by a moderate baseline confidence weight coefficient to retain the baseline features. The three weighted components are then superimposed in the time domain to obtain the reconstructed ethanol concentration feature sequence and carbon dioxide waveform feature sequence. This feature reconstruction with weighted allocation highlights effective features reflecting alcohol intake (such as ethanol concentration fluctuations related to respiratory rhythm) and suppresses irrelevant effects from environmental interference and basal metabolism, enabling the final gas features to more accurately correspond to the driver's alcohol status. The obtained ethanol concentration feature sequence and carbon dioxide waveform feature sequence are concatenated using a channel concatenation method. After aligning the two sequences along the time dimension, they are merged into a multi-dimensional feature vector, which is the odor and voiceprint feature. The ethanol concentration feature sequence reflects the variation in ethanol concentration during the driver's respiration, while the carbon dioxide waveform feature sequence reflects the changes in respiratory rhythm related to alcohol metabolism. The combined odor and voiceprint feature can comprehensively characterize the driver's alcohol intake status from both concentration and rhythm dimensions. Alcohol intake not only leads to an increase in ethanol concentration but also affects respiratory rhythm. A single ethanol concentration indicator is insufficient to fully reflect the impact of alcohol on the driver. The odor and voiceprint feature integrates concentration and rhythm information, providing a more comprehensive and accurate depiction of alcohol intake status. All values ​​in the above weighted calculation process need to be normalized.

[0016] In one embodiment, step S4, which involves obtaining the alcohol risk index based on the identified odor and voiceprint features, includes: S41. Obtain the peak ethanol concentration, peak acetaldehyde concentration, and carbon dioxide waveform consistency based on the odor and voiceprint characteristics. S42. Obtain the acetaldehyde-ethanol concentration ratio based on the peak ethanol concentration and the peak acetaldehyde concentration; S43. Obtain the driver's basal metabolic rate, and obtain the alcohol risk index based on the basal metabolic rate, the acetaldehyde-ethanol concentration ratio, the carbon dioxide waveform consistency, and the peak ethanol concentration.

[0017] As described in steps S41-S43 above, the risk of alcohol consumption for drivers depends not only on the concentration of ethanol in exhaled breath, but also on the efficiency of alcohol metabolism in the body (reflected by the acetaldehyde-ethanol concentration ratio and basal metabolic rate), the authenticity of the respiratory signal (verified by the carbon dioxide waveform consistency), and the interference of the detection environment (corrected by in-vehicle noise parameters). Furthermore, the odor and voiceprint characteristics are identified after the gases produced by other people in the vehicle are expelled. Based on this, the present invention obtains the peak ethanol concentration, peak acetaldehyde concentration, and carbon dioxide waveform consistency through a feature extraction algorithm. The calculation of carbon dioxide waveform consistency involves matching the carbon dioxide waveform feature sequence in the odor voiceprint characteristics with the preset normal breathing carbon dioxide waveform feature sequence. The Dynamic Time Warping (DTW) algorithm is used to calculate the distance between the two, and the normalized distance value is used as the consistency between the two (the smaller the distance value, the higher the consistency). The peak values ​​of ethanol concentration and acetaldehyde concentration reflect the highest content. The maximum value is used to reflect the current serious state to avoid misjudgment and safety accidents caused by using the minimum value. The carbon dioxide waveform consistency is used to verify whether the collected gas characteristics come from the driver's actual breathing and to avoid interference from residual alcohol gas in the environment. The acetaldehyde-ethanol concentration ratio is obtained based on the peak ethanol concentration and the peak acetaldehyde concentration. The acetaldehyde-ethanol concentration ratio reflects the driver's ability to metabolize alcohol. The higher the ratio, the faster the conversion of ethanol to acetaldehyde, but the slower the decomposition of acetaldehyde. The more acetaldehyde accumulates in the body, the higher the risk of alcohol poisoning symptoms such as confusion and slow reaction. Conversely, the lower the ratio, the stronger the metabolic capacity and the lower the risk. The driver's basal metabolic rate is obtained, and the basal metabolic rate, the acetaldehyde-ethanol concentration ratio, the carbon dioxide waveform consistency, and the peak ethanol concentration are weighted after normalization to obtain the alcohol risk index. By synergistically integrating multi-dimensional parameters and introducing the basal metabolic rate, the individual difference adaptation can be further enhanced. Compared with risk assessment that relies solely on gas characteristics, the alcohol risk index is more in line with the driver's actual physiological state and alcohol risk level, improving the comprehensiveness and accuracy of the assessment.

[0018] In one embodiment, step S5, which involves obtaining vehicle speed and geographic location information based on the vehicle operating status data, includes: S51. Obtain the rate of change of speed within a preset time based on the vehicle speed, and use the rate of change of speed as the vehicle acceleration. S52. The vehicle's angular velocity and acceleration are obtained by the vehicle's inertial measurement unit, and the average acceleration is calculated based on the vehicle's acceleration and the measured acceleration. S53. Obtain the original latitude and longitude coordinates based on the global positioning system receiver, and obtain the geographical location information based on the vehicle's angular velocity, the average acceleration, and the original latitude and longitude coordinates.

[0019] As described in steps S51-S53 above, angular velocity reflects the vehicle's steering and tilting states, while average acceleration supplements the vehicle's linear acceleration and deceleration information. The combination of the two can provide motion state support for the calculation of geographic location information when GPS signals are interrupted or insufficient in accuracy, ensuring the continuity of location data.

[0020] In one embodiment, step S6, which involves obtaining a driving safety level based on the vehicle speed, the alcohol risk index, and the geographical location information, and then controlling the vehicle in conjunction with the driving safety level, includes: S61. Obtain road attribute data based on the geographical location information and the preset high-precision map database, wherein the road attribute data includes road type, speed limit standard, and traffic flow density characteristics; S62. Obtain the road complexity index based on the road type, the speed limit standard, and the traffic flow density characteristics; S63. Obtain a comprehensive driving hazard coefficient by combining the vehicle speed, alcohol risk index, and road complexity index; S64. Determine the relationship between the comprehensive driving risk coefficient and the preset safety range value; When the comprehensive driving risk coefficient is within the preset safety range and equal to the minimum value of the preset safety range, the current vehicle safety level is determined to be a warning level, and an audio command is generated. The vehicle is then controlled with audio reminders based on the warning command. When the comprehensive driving hazard coefficient is within a preset safety range and is greater than the minimum value of the preset safety range and less than the maximum value of the preset safety range, the current vehicle's safety level is determined to be a restricted level, and a power limiting command is generated. Power limiting linkage control is then performed on the vehicle according to the power limiting command. When the comprehensive driving risk coefficient is within the preset safety range and equal to the maximum value of the preset safety range, the current vehicle's safety level is determined to be the intervention level, and a parking command is generated. Based on the parking command, the vehicle is controlled to stop at a speed-linked speed.

[0021] As described in steps S61-S64 above, this invention obtains road attribute data based on the geographical location information and a preset high-precision map database. Road attribute data directly determines the environmental risk basis for vehicle operation. Different road attributes have different requirements for speed limits and driving operations, making them key environmental parameters for assessing comprehensive risk. Secondly, road type, speed limit standards, and traffic flow density characteristics are used to calculate the road complexity index using a weighted summation algorithm. Specific weights and scoring rules are determined through experiments and traffic engineering theory: the road type weight is set to 0.4 (3 points for urban arterial roads, 4 points for expressways, and 2 points for rural roads; higher scores indicate higher complexity); the speed limit standard weight is set to 0.3 (4 points for speed limits ≤40km / h, 2 points for 40-80km / h, and 1 point for >80km / h); and the traffic flow density characteristic weight is set to 0.3 (4 points for high density, 2 points for medium density, and 1 point for low density). The road complexity index transforms qualitative road attributes into quantitative risk indicators, achieving comparability of environmental risks for different roads. The comprehensive driving hazard coefficient is calculated by weighting vehicle speed, alcohol risk index and road complexity index. The comprehensive driving hazard coefficient integrates vehicle dynamic risk, driver alcohol risk and road environment risk. The system determines the relationship between the overall driving hazard coefficient and the preset safety range value, then classifies the safety level based on the relationship and executes linked control. When the overall driving hazard coefficient is equal to 0.2, it is determined to be a warning level. The system generates an audio command through the vehicle's audio system to provide an audio reminder and linked control. At this time, the risk is low, and there is no need to intervene in the vehicle's driving; the system only reinforces the driver's safety awareness through reminders. When the overall driving hazard coefficient is greater than 0.2 and less than 0.8, it is determined to be a restriction level. The system generates a power limiting command to the vehicle's ECU (Electronic Control Unit). By reducing the fuel injection frequency of the engine injectors and limiting the throttle opening, the system ensures that the vehicle's maximum speed does not exceed 80% of the current road speed limit, avoiding the superposition of speeding and high risk. When the overall driving hazard coefficient is equal to 0.8, it is determined to be an intervention level. The system generates a parking command. By gradually increasing the braking pressure and activating the hazard warning lights, the system smoothly pulls the vehicle to the emergency lane, completely eliminating the high-risk driving hazard. Different safety levels correspond to different levels of control measures.

[0022] This application also provides a vehicle-mounted alcohol concentration real-time monitoring and vehicle control linkage system, including: The first acquisition module is used to acquire in-vehicle multi-source sensor data, wherein the in-vehicle multi-source sensor data includes biometric data and gas characteristic data; The second acquisition module is used to acquire facial feature data and hand feature data based on the biometric data, and to acquire driver identification based on the facial feature data and the hand feature data; The third acquisition module is used to acquire ethanol concentration parameters and volatile organic compound parameters based on the gas characteristic data, acquire interfering gas concentration based on the volatile organic compound parameters, and acquire odor voiceprint features based on the ethanol concentration parameters and the interfering gas concentration. The association module is used to associate the driver's identity identifier with the odor and voiceprint features to obtain the identifier odor and voiceprint features, and to obtain the alcohol risk index based on the identifier odor and voiceprint features; The fourth acquisition module is used to acquire vehicle operating status data and acquire vehicle speed and geographical location information based on the vehicle operating status data; The fifth acquisition module is used to acquire the driving safety level based on the vehicle speed, the alcohol risk index and the geographical location information, and to perform linkage control on the vehicle based on the driving safety level.

[0023] In one embodiment, the second acquisition module includes: The first acquisition unit is used to acquire the topological feature vector of the driver's facial key points based on the facial feature data. The second acquisition unit is used to acquire the driver's hand grip pressure distribution matrix based on the hand feature data, and to perform spatiotemporal domain analysis on the hand grip pressure distribution matrix to obtain the knuckle positioning coordinates and palm print texture feature vectors. The fusion unit is used to perform multimodal feature fusion on the facial key point topological feature vector, the knuckle positioning coordinates and the palm print texture feature vector based on an attention mechanism to obtain a biological fusion feature vector; The encoding unit is used to input the biometric fusion feature vector into a pre-trained deep belief network for identity feature encoding, generate a driver identification code, and use the driver identification code as the driver's identity identifier.

[0024] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0025] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0026] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0027] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0028] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for real-time monitoring of vehicle alcohol concentration and linkage with vehicle control, characterized in that, include: Acquire data from multiple sources of sensors inside the vehicle, including biometric data and gaseous data. Facial and hand feature data are obtained based on the biometric data, and driver identification is obtained based on the facial and hand feature data. Based on the gas characteristic data, obtain ethanol concentration parameters and volatile organic compound parameters, and based on the volatile organic compound parameters, obtain interfering gas concentrations, and based on the ethanol concentration parameters and the interfering gas concentrations, obtain odor voiceprint characteristics; The driver's identity identifier and the odor and voiceprint features are associated to obtain the identifier odor and voiceprint features, and the alcohol risk index is obtained based on the identifier odor and voiceprint features; Acquire vehicle operating status data, and obtain vehicle speed and geographical location information based on the vehicle operating status data; The driving safety level is obtained based on the vehicle speed, the alcohol risk index, and the geographical location information, and the vehicle is controlled in conjunction with the driving safety level. The steps of obtaining the concentration of interfering gases based on the volatile organic compound parameters, and obtaining odor voiceprint features based on the ethanol concentration parameter and the concentration of interfering gases, include: The acetone concentration time series and benzene series concentration time series are obtained based on the volatile organic compound parameters, and the interfering gas concentration is obtained based on the acetone concentration time series and the benzene series concentration time series. Ethanol concentration time series and carbon dioxide concentration time series are obtained based on the gas characteristic data, and a multi-time-frequency aligned intrinsic mode function component set is obtained based on the ethanol concentration time series and the carbon dioxide concentration time series, wherein the intrinsic mode function component set includes respiratory rhythm characteristic components, environmental interference noise components and basal metabolic components; Obtain the first signal-to-noise ratio corresponding to the respiratory rhythm feature components, and obtain the respiratory confidence weighting coefficient based on the first signal-to-noise ratio and the concentration of the interfering gas; Obtain the second signal-to-noise ratio corresponding to the environmental interference noise component, and obtain the environmental confidence weighting coefficient based on the second signal-to-noise ratio and the concentration of the interfering gas; Obtain the third signal-to-noise ratio corresponding to the basic metabolic component, and obtain the basic confidence weighting coefficient based on the third signal-to-noise ratio and the concentration of interfering gas; Based on the respiratory confidence weight coefficient, the environmental confidence weight coefficient, and the basic confidence weight coefficient, the intrinsic mode function component set is reconstructed to obtain the ethanol concentration feature sequence and the carbon dioxide waveform feature sequence. The ethanol concentration feature sequence and the carbon dioxide waveform feature sequence are concatenated to obtain the odor voiceprint feature; The step of obtaining the alcohol risk index based on the identified odor and voiceprint characteristics includes: The peak values ​​of ethanol concentration, peak values ​​of acetaldehyde concentration, and the waveform consistency of carbon dioxide were obtained based on the odor and voiceprint characteristics. The acetaldehyde-ethanol concentration ratio is obtained based on the peak ethanol concentration and the peak acetaldehyde concentration. The driver's basal metabolic rate is obtained, and an alcohol risk index is obtained based on the basal metabolic rate, the acetaldehyde-ethanol concentration ratio, the carbon dioxide waveform consistency, and the peak ethanol concentration.

2. The method for real-time monitoring of vehicle alcohol concentration and linkage with vehicle control according to claim 1, characterized in that, The step of obtaining the driver's identity based on the facial feature data and the hand feature data includes: The topological feature vector of the driver's facial key points is obtained based on the facial feature data. The driver's hand grip pressure distribution matrix is ​​obtained based on the hand feature data, and the spatiotemporal domain analysis is performed on the hand grip pressure distribution matrix to obtain the knuckle positioning coordinates and palm print texture feature vectors. Based on the attention mechanism, multimodal feature fusion is performed on the topological feature vector of the facial key points, the positioning coordinates of the knuckles, and the palm print texture feature vector to obtain the biological fusion feature vector; The biometric fusion feature vector is input into a pre-trained deep belief network for identity feature encoding to generate a driver identification code, which is then used as the driver's identity identifier.

3. The method for real-time monitoring of vehicle alcohol concentration and linkage with vehicle control according to claim 1, characterized in that, The step of obtaining vehicle speed and geographical location information based on the vehicle operating status data includes: The rate of change of speed within a preset time is obtained based on the vehicle speed, and the rate of change of speed is used as the vehicle acceleration. The vehicle's angular velocity and acceleration are acquired using the vehicle's inertial measurement unit, and the average acceleration is calculated based on the vehicle's acceleration and the measured acceleration. The system obtains the original latitude and longitude coordinates based on the GPS receiver, and obtains the geographical location information based on the vehicle's angular velocity, the average acceleration, and the original latitude and longitude coordinates.

4. The method for real-time monitoring of vehicle alcohol concentration and linkage with vehicle control according to claim 1, characterized in that, The step of obtaining a driving safety level based on the vehicle speed, the alcohol risk index, and the geographical location information, and then performing coordinated control of the vehicle based on the driving safety level, includes: Road attribute data is obtained based on the geographical location information and a preset high-precision map database. The road attribute data includes road type, speed limit standard, and traffic flow density characteristics. The road complexity index is obtained based on the road type, the speed limit standard, and the traffic flow density characteristics; The comprehensive driving hazard coefficient is obtained by combining the vehicle speed, alcohol risk index, and road complexity index. Determine the relationship between the overall driving hazard coefficient and the preset safety range value; When the comprehensive driving risk coefficient is within the preset safety range and equal to the minimum value of the preset safety range, the current vehicle safety level is determined to be a warning level, and an audio command is generated. Based on the audio command, the vehicle is controlled with audio reminders. When the comprehensive driving hazard coefficient is within a preset safety range and is greater than the minimum value of the preset safety range and less than the maximum value of the preset safety range, the current vehicle's safety level is determined to be a restricted level, and a power limiting command is generated. Power limiting linkage control is then performed on the vehicle according to the power limiting command. When the overall driving hazard coefficient is within a preset safe range and equal to the maximum value of the preset safe range, the current vehicle's safety level is determined to be an intervention level, and a parking command is generated. Based on the parking command, the vehicle is controlled to stop at a speed-linked speed.

5. A vehicle-mounted alcohol concentration real-time monitoring and vehicle control linkage system, used to execute the vehicle-mounted alcohol concentration real-time monitoring and vehicle control linkage method as described in any one of claims 1-4, characterized in that, include: The first acquisition module is used to acquire in-vehicle multi-source sensor data, wherein the in-vehicle multi-source sensor data includes biometric data and gas characteristic data; The second acquisition module is used to acquire facial feature data and hand feature data based on the biometric data, and to acquire driver identification based on the facial feature data and the hand feature data; The third acquisition module is used to acquire ethanol concentration parameters and volatile organic compound parameters based on the gas characteristic data, acquire interfering gas concentration based on the volatile organic compound parameters, and acquire odor voiceprint features based on the ethanol concentration parameters and the interfering gas concentration. The association module is used to associate the driver's identity identifier with the odor and voiceprint features to obtain the identifier odor and voiceprint features, and to obtain the alcohol risk index based on the identifier odor and voiceprint features; The fourth acquisition module is used to acquire vehicle operating status data and acquire vehicle speed and geographical location information based on the vehicle operating status data; The fifth acquisition module is used to acquire the driving safety level based on the vehicle speed, the alcohol risk index and the geographical location information, and to perform linkage control on the vehicle based on the driving safety level.

6. The vehicle-mounted alcohol concentration real-time monitoring and vehicle control linkage system according to claim 5, characterized in that, The second acquisition module includes: The first acquisition unit is used to acquire the topological feature vector of the driver's facial key points based on the facial feature data. The second acquisition unit is used to acquire the driver's hand grip pressure distribution matrix based on the hand feature data, and to perform spatiotemporal domain analysis on the hand grip pressure distribution matrix to obtain the knuckle positioning coordinates and palm print texture feature vectors. The fusion unit is used to perform multimodal feature fusion on the facial key point topological feature vector, the knuckle positioning coordinates and the palm print texture feature vector based on an attention mechanism to obtain a biological fusion feature vector; The encoding unit is used to input the biometric fusion feature vector into a pre-trained deep belief network for identity feature encoding, generate a driver identification code, and use the driver identification code as the driver's identity identifier.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images