Non-inductive physical and psychological health monitoring and early warning system and method before coal miner goes down to well

Abstract

The invention discloses a non-inductive physical and psychological health monitoring and early warning system and method before a coal miner goes down a well, and belongs to the technical field of coal mine safety production and personnel health monitoring. A physiological monitoring unit comprises a millimeter-wave radar, an infrared camera and an imaging type photoplethysmography device, the millimeter-wave radar is installed in the center of the top of a monitoring channel, and the infrared camera is installed on the imaging type photoplethysmography device; the infrared cameras are symmetrically installed on the two sides of the channel, the imaging type photoplethysmography devices are integrated in the infrared cameras and used for synchronously collecting the core body temperature and the oxyhemoglobin saturation, and the integrated imaging type photoplethysmography devices focus on the forehead and cheek areas through the RGB channel. Reflected light intensity fluctuation signals caused by blood vessel blood volume changes are collected. Through cooperation of the three devices, non-contact synchronous collection of multiple parameters can be completed without deliberate cooperation of workers, comprehensive and continuous basic data are provided for subsequent physical and psychological health analysis, the wellhead passing efficiency is improved, and the cross infection risk of contact type devices is avoided.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of coal mine safety production and personnel health monitoring, and particularly relates to a coal worker pre-downhole non-inductive physical and mental health monitoring and early warning system and method. BACKGROUND

[0002] Coal mine safety production has always been a problem that needs to be paid attention to, and the health status of coal workers is an important part of coal mine safety production. Monitoring the physiological health status of coal workers can effectively reduce safety accidents caused by worker physical discomfort during production, timely screen out personnel who are not suitable for downhole work, and also record the physical health status of workers for a long time.

[0003] The existing non-invasive blood oxygen detection equipment uses photoplethysmography to obtain blood oxygen, and needs to contact the detection equipment to the human body surface. In the case of high personnel mobility, the use is more limited, and there is also the possibility of spreading infectious diseases. Remote photoplethysmography technology is used for non-contact blood oxygen measurement, but this method has the defects of poor precision and slow measurement. Therefore, a non-inductive, multi-dimensional, high-precision monitoring and early warning device and method are needed to improve the management level of coal mine safety production. SUMMARY

[0004] The purpose of the present application is to provide a coal worker pre-downhole non-inductive physical and mental health monitoring and early warning system and method to solve the problems raised in the background technology.

[0005] To achieve the above purpose, the present application provides the following technical scheme: a coal worker pre-downhole non-inductive physical and mental health monitoring and early warning system, comprising a data acquisition module, an edge computing module, a cloud platform module and an early warning module, the data acquisition module is used for synchronously acquiring physiological, psychological and environmental data, the edge computing module uses an edge computing server to pre-process and fusion analyze the collected data, the cloud platform module is used for storing historical data and establishing a personal health baseline, and comparing real-time data with the personal health baseline to generate a comprehensive risk level, the early warning module outputs a warning signal according to the risk level, the edge computing module is wired connected with the data acquisition module, the cloud platform module is wirelessly communicated with the edge computing module, and the early warning module is electrically connected with the edge computing module. The data acquisition module comprises a physiological monitoring unit, a psychological assessment unit and an environment monitoring unit. The physiological monitoring unit comprises a millimeter wave radar, an infrared camera and an imaging photoelectric plethysmograph device. The millimeter wave radar is installed at the top center of the monitoring channel. The infrared cameras are symmetrically installed on both sides of the channel. The imaging photoelectric plethysmograph device is integrated in the infrared camera and is used for synchronous acquisition of core body temperature and blood oxygen saturation. The integrated imaging photoelectric plethysmograph device focuses on the forehead and cheek areas through the RGB channel to collect the reflected light intensity fluctuation signals caused by the change of blood volume in blood vessels.

[0006] In the above implementation process, the millimeter wave radar is installed at the top center of the monitoring channel. With the help of frequency-modulated continuous wave technology, the millimeter wave radar transmits millimeter wave signals to the space in the channel. The physiological parameters such as heart rate and respiratory rate are captured in real time by using the frequency change of the echo caused by the fluctuation of the chest and the pulsation of the blood vessels. The infrared cameras are symmetrically distributed on both sides of the channel to collect the thermal radiation information of the workers' faces from different angles to construct the face temperature field. At the same time, the integrated imaging photoelectric plethysmograph device relies on the RGB channel to accurately focus on the forehead and cheek areas where the blood vessels are densely distributed and the light reflection signals are stable. When the blood flows in the blood vessels and the volume changes, the reflected light intensity in this area will fluctuate slightly. By capturing these fluctuation signals and combining the photoplethysmography algorithm, the light intensity change is converted into an electrical signal. After signal processing, the core body temperature and the blood oxygen saturation are calculated synchronously. The three devices work together to achieve non-contact and synchronous acquisition of multiple physiological parameters during the natural passing process without the workers' deliberate cooperation, which can provide comprehensive and continuous basic data for subsequent physical and mental health analysis.

[0007] In a specific embodiment, the psychological assessment unit comprises a microphone array and a behavior recognition camera. The microphone array is deployed at the entrance side of the channel to collect voice signals and extract voiceprint features. The behavior recognition camera is coaxially installed with the infrared camera to track the action frequency and the stationary time of the workers.

[0008] In the implementation process, the microphone array uses multi-channel audio acquisition and beamforming technology to accurately capture the voiceprint information of the worker when reporting voice in the channel. By extracting features such as fundamental frequency, speech rate, and sound pressure level fluctuation, and combining a pre-trained psychological state recognition model, such as an emotion and fatigue discrimination model based on long short-term memory network (LSTM), the emotion in the voice is analyzed, such as anxiety, which causes the speech rate to increase, the sound pressure level to fluctuate greatly, and the psychological state dimensions such as fatigue, which causes the fundamental frequency to decrease, and unclear pronunciation. The behavior recognition camera is based on computer vision algorithms to detect human posture key points in the pictures collected by the coaxially installed infrared camera. With the help of algorithm frameworks such as OpenPose, the worker's limb action trajectory is tracked, and the action frequency in a unit of time is counted, such as the arm swinging frequency when walking through the channel in a normal state and the proportion of stationary time. If the worker stops for a short time due to physical discomfort, and the time exceeds the pre-set threshold, it is marked as abnormal. The voice feature data is fused to build a multi-modal psychological state evaluation system, realize the comprehensive discrimination of psychological indicators such as worker fatigue, stress, and mental concentration, provide comprehensive psychological dimension data support for subsequent physical and mental health monitoring and early warning, make up for the single dimension of traditional physiological monitoring, and improve the accuracy and comprehensiveness of the evaluation of the worker's physical and mental state before going down the well.

[0009] In a specific embodiment, the environment monitoring unit includes a gas sensor, a temperature and humidity sensor, and a dust concentration sensor, all installed on the top edge of the channel, for real-time monitoring of underground environmental risk parameters.

[0010] In the implementation process, the gas sensor, temperature and humidity sensor, and dust concentration sensor are uniformly distributed on the top edge of the channel, which can form a comprehensive environmental monitoring network. The gas sensor is based on catalytic combustion or infrared absorption principle to capture the leaked gas in the underground in real time, accurately quantifies the concentration of methane and other combustible gases, and once the concentration exceeds the safety threshold, it can timely warn the risk of gas explosion. The temperature and humidity sensor uses high-precision sensing elements to continuously collect environmental temperature and relative humidity data. Abnormal temperature rise may indicate overheating of equipment, fire hazards, and excessive humidity may cause electrical equipment to malfunction due to moisture, mold growth in the working environment, and other problems. The dust concentration sensor uses light scattering, beta ray absorption, and other technologies to monitor the content of suspended dust in the underground. High-concentration dust not only harms the respiratory health of workers, but also may cause coal dust explosion. These sensors work together to transmit the collected environmental parameters to the data processing system in real time, and analyze the data in combination with the physical and mental health data of the workers. When the environmental risk parameters are abnormal, on the one hand, the environmental safety warning is triggered directly, and on the other hand, in combination with the physical and mental state of the workers, it is determined whether the environmental factors have caused discomfort to the workers, such as heat fatigue caused by high temperature and humidity, and shortness of breath caused by dust, to provide multi-dimensional data support for the regulation of the underground working environment and the protection of the health of the workers, and to reduce the correlation of environmental risks and physical and mental damage from the source, and to build an environmental defense line for coal mine safety production.

[0011] In a specific embodiment, the millimeter wave radar captures the chest fluctuation and blood vessel pulsation in real time through frequency modulation continuous wave technology, outputs the heart rate, respiratory rate, subcutaneous blood flow change and blood pressure fluctuation trend, and synchronously generates a respiratory waveform graph.

[0012] In the above implementation process, the millimeter wave radar relies on frequency modulation continuous wave technology. The millimeter wave signal emitted by the radar propagates through free space, is reflected after encountering the chest surface and blood vessel tissue of the worker, the echo signal is captured by the radar receiving antenna, and based on the principle that the frequency difference between the transmitted signal and the echo signal corresponds to the target motion speed, the periodic fluctuation of the chest due to respiration and the slight displacement of the blood vessels with the pulse are accurately analyzed. For heart rate monitoring, the frequency component of the blood vessel displacement synchronized with the heart beat in the echo is extracted, and through a digital signal processing algorithm such as fast Fourier transform, the characteristic frequency corresponding to the heart contraction and relaxation is separated from the mixed signal to calculate the heart rate value. Respiratory rate monitoring focuses on the slow frequency component of the chest fluctuation, and by tracking the chest movement cycle, the number of breaths per unit time is counted. Subcutaneous blood flow change monitoring uses the Doppler effect of millimeter waves on blood flow to analyze the amplitude and phase changes of the echo signal and to inverse the dynamic changes of the blood volume in the blood vessels. Blood pressure fluctuation trend acquisition is based on the correlation model of blood flow velocity, blood vessel wall displacement and blood pressure during blood vessel pulsation, combined with heart rate and blood flow change data, to calculate the relative fluctuation of blood pressure. At the same time, the radar signal processing module generates an intuitive respiratory waveform graph with time as the horizontal axis and displacement as the vertical axis based on the continuously collected respiratory-related displacement signals. These multi-dimensional physiological data lay the foundation for the subsequent fusion of blood oxygen, body temperature and other information collected by infrared cameras, photoplethysmography and other devices, and the construction of a comprehensive physiological state evaluation model for workers, realizing non-contact accurate monitoring of the heart function, respiratory function and circulatory system state of workers, efficiently and hygienically completing the physiological health screening of workers in the high flow scene at the coal mine shaft, and helping to identify cardiovascular disease risks, respiratory function abnormalities and other health risks that affect the safety of underground operations in advance.

[0013] In a specific embodiment, the cloud platform module is constructed based on an Ali Cloud ECS instance.

[0014] In the implementation process, when building the cloud platform module based on the Ali Cloud ECS instance, first, according to the data processing requirements of the coal mine wellhead monitoring scene, such as real-time reception, storage, and analysis of multi-sensor data, as well as health record management, model calculation, etc., the appropriate ECS instance specification is selected, including CPU core number, memory capacity, storage type, and bandwidth resource configuration, and the instance creation and initialization are completed through the Ali Cloud console, and the operating system such as Linux system and the necessary running environment such as database management system MySQL for storing worker health data, Python environment for running data analysis and machine learning model are deployed. Then, using the network services provided by Ali Cloud, such as private network VPC and security group, a secure and reliable network environment is built to ensure the security and stability of data transmission between the edge computing module and the cloud platform, and the communication connection is established through 5G or wired network to realize real-time uploading of multi-dimensional monitoring data such as physiological, psychological, and environmental data. On the cloud platform side, data processing and analysis applications are developed: on the one hand, the worker health data preprocessed by the edge computing module is received, and the personal health baseline model is combined, which is generated based on long-term historical data using machine learning algorithm, and is used to dynamically compare and judge whether the real-time data is abnormal, and to perform multi-dimensional fusion analysis to calculate health attenuation index, risk level, etc. On the other hand, relying on the storage services of Ali Cloud, such as object storage OSS, a worker health record database is established to realize long-term storage and traceability of data, support data query, statistics, and visualization display according to worker ID, time dimension, etc. At the same time, with the help of Ali Cloud's elastic scaling and load balancing services, the stable operation of the cloud platform during the period when the personnel concentrate in the wellhead is guaranteed, and the data interaction request of the edge computing module is efficiently responded, providing cloud computing power and data management support for the physical and mental health monitoring and early warning of coal mine workers before going down the well, realizing a complete closed loop from data collection to analysis and early warning.

[0015] In a specific embodiment, the early warning module includes an audible and visual alarm and a display screen. The audible and visual alarm emits different colored lights and prompt sounds according to the risk level. The display screen displays the worker's health status and abnormal indicators, and when there is a high risk, it locks the permission to go down the well in conjunction with the down-the-well permission management system.

[0016] In the above implementation process, the edge computing module transmits the analyzed risk level to the early warning module. The audible and visual alarm has built-in control logic. Upon receiving the risk level command, low risk triggers a green light and silent light prompt; medium risk activates a yellow light and intermittent sound prompt; and high risk activates a red light and rapid alarm sound. The display screen renders and displays the worker's health status in real time, such as normal physical condition, good mental condition, and abnormal indicators, such as an excessively fast heart rate (current value 120 beats / minute, baseline value 70-90 beats / minute). When a high risk is determined, the early warning module calls the underground access control system through the API interface to send a lock command. Upon receiving the command, the system immediately restricts the corresponding worker's access to the mine and records the early warning event, forming a complete safety closed loop of monitoring, early warning, and control, comprehensively ensuring the safety of underground operations in coal mines.

[0017] In one specific implementation plan, the method for monitoring and early warning of physical and mental health without conscious awareness is as follows: S1. When a coal miner naturally enters the pre-set monitoring channel at the mine entrance, the infrared camera at the channel entrance is blocked, automatically triggering the system to start. At this time, the data acquisition module and edge computing module are activated simultaneously and enter standby mode. The worker does not need to stop or touch any equipment. S2. After the system starts, the physiological monitoring unit, psychological assessment unit, and environmental monitoring unit of the data acquisition module work simultaneously to complete the synchronous acquisition of multi-dimensional data during the worker's passage, ensuring that the data timestamps are consistent. The multi-dimensional data includes physiological data, psychological data, and environmental data. S3. The collected raw data is transmitted to the edge computing module for targeted preprocessing to eliminate noise and interference; S4, the edge computing module performs multi-dimensional fusion analysis on the preprocessed data and uploads the fused real-time data to the cloud platform module via the 5G network. This data is then compared with the worker's personal health baseline. The cloud platform calculates the difference between the real-time data and the baseline using a deviation algorithm, and, combined with environmental risks, generates a health decline index and a comprehensive risk level. The risk level includes: Low risk: Health decline index <30, no environmental risk, and physiological / psychological indicators are all within ±10% of baseline; Medium risk: Health decline index 30-60, or a single indicator deviating from the baseline by 10% to 20%, or there is a slight environmental risk; High risk: Health decline index > 60, or single indicator deviates from baseline by > 20%, or environmental parameters exceed the standard; S5 and cloud platform will feed the comprehensive risk level back to edge computing module. Edge computing module will trigger the early warning module to take action. When the risk is low, the channel exit display screen will show that the health status is normal and you can go down the well without any audio or visual prompts. When the risk level is medium, the audible and visual alarm will emit a flashing yellow light and a low-frequency warning sound, and the abnormal indicators will be displayed on the monitoring room screen simultaneously. In cases of high risk, the audible and visual alarm will emit a solid red light and a high-frequency alarm sound. The system will automatically link with the wellhead access management system to temporarily lock the worker's access to the well. A detailed risk report will pop up in the monitoring room, prompting management personnel to intervene. S6. All raw data, processing results, risk levels, and early warning records are automatically stored in the cloud platform to update personal health records for long-term health trend analysis.

[0018] In S3, the specific data processing procedure is as follows: Physiological data: The photoplethysmography signal was averaged, detrended, and Butterworth filtered for the face region. The millimeter-wave radar data was filtered to remove limb swaying interference and extract pure physiological fluctuation components. Psychological data: Environmental noise is removed from the speech signal, and effective speech segments are extracted; behavioral data is used to remove fixed background interference within the channel and focus on the individual action characteristics of workers. Environmental data: Outlier removal is performed to retain continuous and stable environmental parameter sequences.

[0019] In S4, multi-dimensional fusion analysis includes: S4-1, Physiological state analysis: Based on subcutaneous temperature measured by millimeter-wave radar, epidermal temperature from infrared cameras, and blood flow rate change, core body temperature is calculated using a heat conduction model; Blood oxygen saturation calculation: Extract the AC and DC components of the photoplethysmography signal, substitute them into the linear regression model, and update the blood oxygen value every 10 seconds; Basic physiological indicators integration: summarizing heart rate, respiratory rate, and blood pressure trends to generate a physiological state vector; S4-2, Psychological state analysis uses a CNN plus LSTM deep learning model to fuse micro-expressions, speech features and behavioral patterns, and outputs fatigue index and emotional stability score. S4-3, Environmental risk assessment compares environmental parameters with safety thresholds in real time, serving as a reference factor for subsequent comprehensive assessment.

[0020] In S4, an individual's health baseline is established during the first test, including physiological and psychological benchmark values, and is dynamically updated every N tests thereafter, where N≥5, to ensure the baseline's adaptability.

[0021] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention deploys a multi-dimensional data acquisition module, covering physiological, psychological, and environmental monitoring units. It uses non-contact sensing technologies such as millimeter-wave radar, infrared cameras, and microphone arrays to achieve unobtrusive monitoring of workers' physical and mental health and environmental risks before they go down into the mine. Workers do not need to actively cooperate or come into contact with the equipment. This not only improves the efficiency of passage through the mine entrance but also avoids the risk of cross-infection from contact equipment, solving the problem of poor applicability of traditional contact monitoring in coal mine entrance scenarios.

[0022] 2. This invention constructs a multi-dimensional data fusion analysis system, which correlates physiological data such as heart rate and blood oxygen, psychological data such as fatigue and emotions, and environmental data such as gas and dust. By using edge computing for real-time fusion and cloud platform personalized baseline comparison, it breaks through the limitations of single-parameter monitoring in existing technologies, achieves more comprehensive and accurate health and risk assessment, and can improve the accuracy of early warning to a certain extent.

[0023] 3. This invention utilizes an edge computing module and a cloud platform collaborative architecture to perform real-time data preprocessing and preliminary analysis at the edge, while the cloud performs long-term data storage, model training, and in-depth mining. This ensures the real-time nature of monitoring and early warning, and also enables the creation of personalized health records based on historical data, providing workers with long-term health trend analysis and helping coal mining enterprises shift from a post-event handling to a pre-event prevention safety management model.

[0024] 4. This invention, through the multi-level linkage mechanism of the early warning module, directly links to the underground access control system when there is a high risk, forcibly restricting unsuitable workers from going down the mine. From a technical perspective, it strengthens the first line of defense for underground operation safety, effectively reduces the probability of safety production accidents caused by workers' physical and mental abnormalities or environmental risks, and improves the intelligence and precision of coal mine safety production management. Attached Figure Description

[0025] Figure 1 This is a system configuration diagram of the present invention; Figure 2 This is a schematic diagram illustrating the blood oxygen measurement principle of the present invention. Figure 3 This is a flowchart of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Example 1, Please see Figure 1 - Figure 3This invention provides a non-intrusive physical and mental health monitoring and early warning system for coal miners before they go underground. The system includes a data acquisition module, an edge computing module, a cloud platform module, and an early warning module. The data acquisition module synchronously acquires physiological, psychological, and environmental data. The edge computing module uses an edge computing server to preprocess and fuse the collected data. The cloud platform module stores historical data and establishes an individual health baseline, comparing real-time data with the baseline to generate a comprehensive risk level. The early warning module outputs an early warning signal based on the risk level. The edge computing module is wired to the data acquisition module, the cloud platform module is wirelessly connected to the edge computing module, and the early warning module is electrically connected to the edge computing module. The data acquisition module includes physiological monitoring... The monitoring system comprises a measurement unit, a psychological assessment unit, and an environmental monitoring unit. The physiological monitoring unit includes a millimeter-wave radar, an infrared camera, and an imaging photoplethysmography (IPC) device. The millimeter-wave radar is installed at the center of the top of the monitoring channel, while the infrared cameras are symmetrically installed on both sides. The IPC device, integrated within the infrared cameras, simultaneously collects core body temperature and blood oxygen saturation. The integrated IPC device focuses on the forehead and cheek areas via an RGB channel, acquiring the reflected light intensity fluctuation signal caused by changes in vascular blood volume. The millimeter-wave radar, using frequency-modulated continuous wave technology, captures chest rise and fall and vascular pulsation in real time, outputting heart rate, respiratory rate, subcutaneous blood flow changes, and blood pressure fluctuation trends, and simultaneously generates a respiratory waveform. The psychological assessment unit includes a microphone array and a behavior recognition camera. The microphone array is deployed at the channel entrance to collect voice signals and extract voiceprint features. The behavior recognition camera is coaxially mounted with the infrared cameras to track worker movement frequency and stillness time. The environmental monitoring unit includes a gas sensor, a temperature and humidity sensor, and a dust concentration sensor, all installed at the top edge of the channel for real-time monitoring of underground environmental risk parameters. The cloud platform module is built on Alibaba Cloud ECS instances. The early warning module includes an audible and visual alarm and a display screen. The audible and visual alarm emits different colored lights and prompts according to the risk level. The display screen shows the worker's health status and abnormal indicators. When there is a high risk, it will link with the well access management system to lock the well access.

[0028] Furthermore, after the system starts up, the physiological monitoring unit, psychological assessment unit, and environmental monitoring unit of the data acquisition module work simultaneously to complete the synchronous acquisition of multi-dimensional data during worker passage. All data timestamp errors are ≤0.05 seconds to ensure analysis consistency. Millimeter-wave radar is used to collect physiological data. It captures chest cavity fluctuations and vascular pulsation in real time through frequency-modulated continuous wave technology, outputting heart rate, respiratory rate, subcutaneous blood flow changes, and blood pressure fluctuation trends, and generating respiratory waveforms. An infrared camera is used to collect facial temperature field distribution and micro-expression features, such as eyelid closure frequency and frown intensity. An imaging photoplethysmography device is integrated into the infrared camera. It focuses on the forehead and cheek areas through the RGB channel to collect the reflected light intensity fluctuation signal caused by changes in vascular blood volume, which is used for subsequent blood oxygen calculation.

[0029] When collecting psychological data, the microphone array automatically triggers voice prompts such as "Please say 'Ready to go down the mine'" to collect voice signals and extract voiceprint features such as fundamental frequency, speech rate, and sound pressure level. Behavioral recognition cameras track the movement trajectory of workers during their movement, recording gait frequency, limb swing amplitude, and stillness time. Simultaneously, it can be linked with underground monitoring cameras to monitor workers' behavioral trajectories. Environmental data collection uses gas sensors, temperature and humidity sensors, and dust concentration sensors to collect environmental parameters in real time at a sampling frequency of 1Hz, ensuring that environmental risks in the monitored area are traceable.

[0030] The collected raw data is transmitted to the edge computing module via gigabit Ethernet for targeted preprocessing to eliminate noise and interference, specifically including: Physiological data: Face region signal averaging was performed on photoplethysmography (PPG) signals to reduce the impact of local skin differences; detrending processing eliminated baseline drift; and Butterworth filtering preserved the 0.5–4 Hz pulse signal frequency band. Millimeter-wave radar data was processed to remove limb movement interference and extract pure physiological fluctuation components. Psychological data: Speech signals underwent adaptive filtering to remove environmental noise, extracting effective speech segments and eliminating prompts. Behavioral data was processed to remove fixed background interference within the channel, focusing on individual worker movement characteristics. Environmental data: The 3σ criterion was used to eliminate outliers, such as instantaneous sensor errors, while retaining continuous and stable environmental parameter sequences.

[0031] At the edge of the multi-dimensional fusion analysis stage: the computing module performs fusion analysis on the preprocessed data to generate key indicators and state assessment results. Based on subcutaneous temperature measured by millimeter-wave radar, epidermal temperature from infrared cameras, and blood flow rate change, core body temperature is calculated using a thermal conduction model. Then, the AC and DC components of the photoplethysmography signal are extracted and substituted into a linear regression model to calculate blood oxygen saturation, updating the blood oxygen value every 10 seconds. Finally, heart rate, respiratory rate, and blood pressure trends are summarized to generate a physiological state vector. Psychological state analysis uses a CNN plus LSTM deep learning model. Micro-expressions (e.g., continuous blinking frequency > 5 times / second indicates fatigue), voice features (e.g., fundamental frequency standard deviation > 50Hz indicates emotional fluctuation), and behavioral patterns (e.g., stillness time > 3 seconds and slow movements indicate abnormal state) are fused to output a fatigue index of 0-100 and an emotional stability score of 0-100, with < 60 indicating abnormality. Environmental risk assessment compares environmental parameters with safety thresholds in real time. For example, a gas concentration > 0.5% LEL triggers an environmental risk warning, serving as a reference factor for comprehensive assessment. The merged real-time data is then uploaded to the cloud platform module via the 5G network through the edge computing server. It is compared with the worker's personal health baseline. The cloud platform calculates the difference between the real-time data and the baseline through the deviation algorithm, and generates a health decline index in combination with environmental risk, and classifies the risk level: low risk: health decline index <30, no environmental risk, and physiological / psychological indicators are all within ±10% of the baseline.

[0032] Medium risk: Health decline index 30-60, or a single indicator deviating from the baseline by 10% to 20%, or there is a slight environmental risk; High risk: Health decline index > 60, or a single indicator deviates from the baseline by > 20%, or environmental parameters exceed the standard.

[0033] The cloud platform can feed back the overall risk level to the edge computing module, which then triggers the early warning module to take action: Low risk: The channel exit display shows a normal health status, and it is safe to go down into the well, with no audible or visual prompts. Medium risk: The audible and visual alarm emits a flashing yellow light and a low-frequency warning sound, such as "Please rest for 10 minutes and check again." The monitoring room display screen simultaneously marks abnormal indicators, such as a slightly elevated heart rate. High risk: The audible and visual alarm emits a bright red light and a high-frequency alarm sound. If entering the well is prohibited, the system will automatically link with the wellhead access management system to temporarily lock the worker's access rights. The monitoring room will then display a detailed risk report, such as blood oxygen levels below the safety threshold, prompting management personnel to intervene.

[0034] All raw data, processing results, risk levels, and early warning records are automatically stored in the cloud platform and personal health records are updated for long-term health trend analysis, such as generating monthly health fluctuation reports to provide data support for health management and occupational disease prevention for coal miners.

[0035] Furthermore, to monitor the blood oxygen saturation of personnel working underground, remote photoplethysmography (PPG) is used to acquire blood oxygen parameters. Changes in blood volume within the blood vessels cause changes in the intensity of light reflected from the face; this change is used to calculate the current blood oxygen concentration. Let c1 and c2 represent the concentrations of hemoglobin and oxygenated hemoglobin in the blood, respectively. and These represent the extinction coefficients of hemoglobin and oxyhemoglobin in the blood, respectively. Let t represent the volume of blood in the blood vessel. The relationship between the signal intensity obtained from the video and the light-absorbing substances in the blood is as follows:

[0036] Here, A and B are coefficients independent of substances in the blood. In the measurement, the blood oxygen concentration is considered constant, and irrelevant variables causing changes in light intensity are filtered out, allowing for the estimation of blood oxygen levels. Thus, during cardiac diastole, blood vessels constrict to their minimum, and the reflected light intensity is at its maximum.

[0037] When the heart contracts, the blood vessels dilate to their maximum size, and the intensity of reflected light is at its minimum. Next, comparing the two and taking the logarithm, we get: At this point, the emitted light changes very little, and the signal can be divided into an AC component and a DC component, i.e. Since hemoglobin and oxyhemoglobin have different extinction coefficients, measurements are taken at two different wavelengths, i.e., two from the RGB channels of the video. Comparing the two: Then, the blood oxygen concentration is calculated using the formula: get: By collecting a large amount of video data, the coefficients 'a' and 'b' in the blood oxygen estimation formula can be obtained through linear regression. Specifically, by collecting video data, obtaining the video signal, averaging the signal from the face region to remove measurement errors, and then performing detrending and Butterworth filtering on the signal to obtain a denoised signal, the AC and DC components of the denoised signal are obtained and substituted into the blood oxygen calculation formula obtained through linear regression to estimate the current blood oxygen level.

[0038] Video signals are collected via cameras, and facial data is used to estimate blood oxygen levels every 10 seconds. This data, combined with other physiological parameters, is used to assess the workers' condition and determine their suitability to continue mining work. The collected data is then uploaded to a cloud server for long-term recording of the workers' physical and mental health.

[0039] Example 2, Standard configuration includes: a millimeter-wave radar, model FMCW-60GHz, installed at the center of the top of the passageway to collect heart rate, respiratory rate, and subcutaneous blood flow; two infrared cameras symmetrically installed on both sides of the passageway to collect facial temperature and micro-expressions; an imaging photoplethysmography device integrated into the infrared cameras to collect blood oxygen signals via the RGB channel; a four-channel microphone array deployed at the entrance of the passageway to collect voice signals; a behavior recognition camera coaxially installed with the infrared cameras to track motion features; gas sensors, temperature and humidity sensors, and dust sensors evenly distributed along the top edge of the passageway; an edge computing server deployed in the monitoring room, connected to the data acquisition module via gigabit Ethernet; a cloud platform module built on Alibaba Cloud ECS instances, communicating with the edge computing server via a 5G network; and an audible and visual alarm and display screen deployed in the monitoring room, electrically connected to the edge computing server.

[0040] Furthermore, the workflow is triggered as follows: when a worker enters the passage, the infrared camera is activated, the system starts, the millimeter-wave radar outputs heart rate and respiratory waveforms, the infrared camera simultaneously collects facial temperature and blood oxygen signals, the microphone array triggers voice prompts such as "Please say 'Prepare to go down the well'" to collect voiceprint features, environmental sensors upload data in real time, during data processing, Butterworth filter processes blood oxygen signals, updates blood oxygen values ​​every 10 seconds, and calculates core body temperature, CNN model analyzes micro-expressions, LSTM model processes voice features, the cloud platform compares real-time data with the individual baseline, calculates the health decay index, classifies low-risk levels, low-risk displays as normal, medium-risk yellow warning suggests rest and re-examination, high-risk red warning prohibits going down the well, and simultaneously locks down the well access.

[0041] Example 3, The lightweight configuration solution deploys a simplified physiological monitoring unit, retaining only one infrared camera and a simplified millimeter-wave radar to measure heart rate and respiration. The environmental monitoring unit retains only gas and temperature and humidity sensors. Data processing still uses an edge computing gateway, simplifying some calculations, such as using moving average instead of Butterworth filtering. The workflow data acquisition cycle is extended to 15 seconds. Blood oxygen calculation uses a simplified model, and risk levels are only divided into normal and abnormal. When an abnormality occurs, a manual review process is triggered.

[0042] Furthermore, the lightweight configuration solution is specifically optimized for cost control and scenario adaptability: From a hardware cost perspective, the number of sensors is reduced after simplification, and the edge computing gateway adopts low-power, low-computing-power devices, reducing the overall hardware procurement and deployment costs. This adapts to the basic monitoring needs of coal mine enterprises in auxiliary wellheads and scenarios with limited budgets. In terms of algorithm efficiency, moving average replaces Butterworth filtering, reducing computational complexity and shortening data processing time by 20% to 30%. Although there is a slight loss in accuracy, it can still meet the needs of rapid initial screening. The manual review process introduces a dual-person, dual-position mechanism. After an anomaly is triggered, the system automatically pushes the worker's basic data to the wellhead safety officer's terminal. The safety officer completes the review and judgment within 5 minutes through rapid video review or verbal inquiry. This not only uses human experience to compensate for the shortcomings of the simplified algorithm, but also ensures the reliability of anomaly identification through process constraints, achieving a balance between cost, efficiency, and safety, and providing differentiated monitoring solutions for coal mine wellheads of different sizes and with different needs.

[0043] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A non-intrusive physical and mental health monitoring and early warning system for coal miners before they go underground, characterized in that, The system includes a data acquisition module, an edge computing module, a cloud platform module, and an early warning module. The data acquisition module is used to synchronously acquire physiological, psychological, and environmental data. The edge computing module uses an edge computing server to preprocess and fuse the collected data. The cloud platform module is used to store historical data and establish an individual health baseline, and compares real-time data with the individual health baseline to generate a comprehensive risk level. The early warning module outputs an early warning signal based on the risk level. The edge computing module and the data acquisition module are wired together, the cloud platform module and the edge computing module are wirelessly connected, and the early warning module and the edge computing module are electrically connected. The data acquisition module includes a physiological monitoring unit, a psychological assessment unit, and an environmental monitoring unit. The physiological monitoring unit includes a millimeter-wave radar, an infrared camera, and an imaging photoplethysmography device. The millimeter-wave radar is installed at the top center of the monitoring channel, and the infrared cameras are symmetrically installed on both sides of the channel. The imaging photoplethysmography device is integrated into the infrared camera and is used to simultaneously collect core body temperature and blood oxygen saturation. The integrated imaging photoplethysmography device focuses on the forehead and cheek areas through the RGB channel to collect the reflected light intensity fluctuation signal caused by changes in blood volume in blood vessels.

2. The non-intrusive physical and mental health monitoring and early warning system for coal miners before they go underground, as described in claim 1, is characterized in that... The psychological assessment unit includes a microphone array and a behavior recognition camera. The microphone array is deployed on the entrance side of the passage to collect voice signals and extract voiceprint features. The behavior recognition camera is coaxially mounted with an infrared camera to track the frequency of worker movements and the time spent at rest.

3. The non-intrusive physical and mental health monitoring and early warning system for coal miners before they go underground, as described in claim 1, is characterized in that... The environmental monitoring unit includes a gas sensor, a temperature and humidity sensor, and a dust concentration sensor, all of which are installed on the top edge of the channel for real-time monitoring of underground environmental risk parameters.

4. The non-intrusive physical and mental health monitoring and early warning system for coal miners before they go underground, as described in claim 1, is characterized in that... The millimeter-wave radar uses frequency-modulated continuous wave technology to capture chest cavity fluctuations and vascular pulsation in real time, outputting heart rate, respiratory rate, subcutaneous blood flow changes and blood pressure fluctuation trends, and simultaneously generating respiratory waveforms.

5. A non-intrusive physical and mental health monitoring and early warning system for coal miners before they go underground, as described in claim 1, is characterized in that... The cloud platform module is built on Alibaba Cloud ECS instances.

6. The non-intrusive physical and mental health monitoring and early warning system for coal miners before they go underground, as described in claim 1, is characterized in that... The early warning module includes an audible and visual alarm and a display screen. The audible and visual alarm emits different colored lights and prompts according to the risk level. The display screen shows the worker's health status and abnormal indicators, and when there is a high risk, it is linked to the well access management system to lock the well access.

7. A method for monitoring and warning the physical and mental health of coal miners before they go underground without being noticed, applicable to the monitoring and warning system for the physical and mental health of coal miners before they go underground as described in any one of claims 1-6, characterized in that, The following are methods for monitoring and providing early warning of physical and mental health without being noticed: S1. When a coal miner naturally enters the pre-set monitoring channel at the mine entrance, the infrared camera at the channel entrance is blocked, automatically triggering the system to start. At this time, the data acquisition module and edge computing module are activated simultaneously and enter standby mode. The worker does not need to stop or touch any equipment. S2. After the system starts, the physiological monitoring unit, psychological assessment unit, and environmental monitoring unit of the data acquisition module work simultaneously to complete the synchronous acquisition of multi-dimensional data during the worker's passage, ensuring that the data timestamps are consistent. The multi-dimensional data includes physiological data, psychological data, and environmental data. S3. The collected raw data is transmitted to the edge computing module for targeted preprocessing to eliminate noise and interference; S4, the edge computing module performs multi-dimensional fusion analysis on the preprocessed data and uploads the fused real-time data to the cloud platform module via the 5G network. This data is then compared with the worker's personal health baseline. The cloud platform calculates the difference between the real-time data and the baseline using a deviation algorithm, and, combined with environmental risks, generates a health decline index and a comprehensive risk level. The risk level includes: Low risk: Health decline index <30, no environmental risk, and physiological / psychological indicators are all within ±10% of baseline; Medium risk: Health decline index 30-60, or a single indicator deviating from the baseline by 10% to 20%, or there is a slight environmental risk; High risk: Health decline index > 60, or single indicator deviates from baseline by > 20%, or environmental parameters exceed the standard; S5 and cloud platform will feed the comprehensive risk level back to edge computing module. Edge computing module will trigger the early warning module to take action. When the risk is low, the channel exit display screen will show that the health status is normal and you can go down the well without any audio or visual prompts. When the risk level is medium, the audible and visual alarm will emit a flashing yellow light and a low-frequency warning sound, and the abnormal indicators will be displayed on the monitoring room screen simultaneously. In cases of high risk, the audible and visual alarm will emit a solid red light and a high-frequency alarm sound. The system will automatically link with the wellhead access management system to temporarily lock the worker's access to the well. A detailed risk report will pop up in the monitoring room, prompting management personnel to intervene. S6. All raw data, processing results, risk levels, and early warning records are automatically stored in the cloud platform to update personal health records for long-term health trend analysis.

8. A method for monitoring and early warning of the physical and mental health of coal miners before they go underground, as described in claim 7, is characterized in that... In S3, the specific data processing procedure is as follows: Physiological data: The photoplethysmography signal was averaged, detrended, and Butterworth filtered for the face region. The millimeter-wave radar data was filtered to remove limb swaying interference and extract pure physiological fluctuation components. Psychological data: Environmental noise is removed from the speech signal, and effective speech segments are extracted; behavioral data is used to remove fixed background interference within the channel and focus on the individual action characteristics of workers. Environmental data: Outlier removal is performed to retain continuous and stable environmental parameter sequences.

9. A method for monitoring and early warning of the physical and mental health of coal miners before they go underground, as described in claim 7, is characterized in that... In S4, multi-dimensional fusion analysis includes: S4-1, Physiological state analysis: Based on subcutaneous temperature measured by millimeter-wave radar, epidermal temperature from infrared cameras, and blood flow rate change, core body temperature is calculated using a heat conduction model; Blood oxygen saturation calculation: Extract the AC and DC components of the photoplethysmography signal, substitute them into the linear regression model, and update the blood oxygen value every 10 seconds; Basic physiological indicators integration: summarizing heart rate, respiratory rate, and blood pressure trends to generate a physiological state vector; S4-2, Psychological state analysis uses a CNN plus LSTM deep learning model to fuse micro-expressions, speech features and behavioral patterns, and outputs fatigue index and emotional stability score. S4-3, Environmental risk assessment compares environmental parameters with safety thresholds in real time, serving as a reference factor for subsequent comprehensive assessment.

10. A method for monitoring and early warning of the physical and mental health of coal miners before they go underground, as described in claim 7, is characterized in that... In S4, an individual's health baseline is established during the first test, including physiological and psychological benchmark values, and is dynamically updated every N tests thereafter, where N≥5, to ensure the baseline's adaptability.

Images