Miner clothing wearing state intelligent monitoring method and system based on physiological feature fusion

Abstract

The invention relates to the technical field of miner clothing wearing state intelligent monitoring methods, in particular to a miner clothing wearing state intelligent monitoring method and system based on physiological feature fusion, and the method comprises the following steps: collecting heart rate, body temperature and breathing frequency signals of a wearer; acquiring action state data of the wearer; based on the heart rate signal, the body temperature signal and the respiratory frequency signal, an edge calculation module is used for carrying out localized data processing to obtain physiological feature data; based on the physiological feature data and the action state data, utilizing a multi-modal data fusion algorithm to generate fusion feature data; establishing a personalized physiological baseline model by using a deep learning algorithm according to the fused feature data; judging whether the wearing state of the wearer is abnormal or not based on the personalized physiological baseline model; when it is judged that the wearing state is abnormal, vibration reminding is triggered; the wearing state and early warning information are transmitted to a supervision center through a wireless communication module, and monitoring parameters can be dynamically adjusted according to the working environment and task characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring methods for the wearing status of miners' clothing, and more specifically, to an intelligent monitoring method and system for the wearing status of miners' clothing based on physiological feature fusion. Background Art

[0002] In the mining industry, the safety of miners has always been a primary concern. With the advancement of technology, various wearable monitoring devices have gradually been used in the safety protection of miners. However, the existing monitoring system still has many shortcomings and cannot meet the safety monitoring needs in complex mine environments.

[0003] Traditional miner safety monitoring systems usually use independent sensors to monitor physiological indicators such as heart rate and body temperature. Although this method is simple and direct, it has many problems. First, wearing multiple independent devices will cause inconvenience to miners and affect their normal work. Secondly, these devices often use fixed thresholds to judge abnormal conditions, which cannot adapt to the individual differences of different miners and the complex and changing working environment. Furthermore, independent data collection lacks integrity and it is difficult to fully reflect the health status and wearing conditions of miners.

[0004] In recent years, some researchers have tried to integrate multiple sensors into work clothes to improve the comprehensiveness and convenience of monitoring. However, such systems still have problems such as data processing lag and insufficient accuracy. Due to the lack of effective data fusion and analysis methods, these systems are difficult to accurately identify complex wearing states and potential safety hazards. At the same time, limited by the traditional centralized data processing architecture, the system has a slow response speed and is difficult to handle emergencies in a timely manner.

[0005] In addition, existing monitoring systems generally ignore the impact of the wearing status of miners' clothing on safety. Improper wearing may lead to reduced protection effect and increased safety risks. However, there is currently a lack of effective methods to intelligently monitor and evaluate the wearing status of miners' clothing.

[0006] In view of the above problems, there is an urgent need for an intelligent system that can comprehensively, accurately and in real time monitor the wearing status of miners' clothing. The system should be able to integrate multiple physiological characteristics and environmental information, adapt to individual differences and environmental changes, and respond quickly to abnormal situations. Summary of the invention

[0007] The intelligent monitoring method and system for the wearing status of miners' clothing based on physiological feature fusion proposed in the present invention are designed to address the problems existing in the prior art. This method realizes the imperceptible collection of multiple physiological indicators such as heart rate, body temperature, and respiratory rate by embedding a flexible sensor network in the miners' clothing. Combined with the motion state data, the present invention adopts a multimodal data fusion algorithm to comprehensively capture the physiological state and wearing status of the miners. Through the deep learning algorithm, the system can establish a personalized physiological baseline model to effectively adapt to the individual characteristics of different miners.

[0008] The present invention provides an intelligent monitoring method for the wearing status of miners' clothing based on physiological feature fusion, comprising the following steps:

[0009] Through the flexible sensor network embedded in the miners' clothing, the wearer's heart rate, body temperature and respiratory rate signals are collected in real time;

[0010] Obtain the wearer's motion status data;

[0011] Based on the heart rate, body temperature and respiratory rate signals, an edge computing module is used to perform localized data processing to obtain physiological characteristic data;

[0012] Based on the physiological characteristic data and the motion state data, using a multimodal data fusion algorithm to generate fused characteristic data;

[0013] Establishing a personalized physiological baseline model using a deep learning algorithm based on the fused feature data;

[0014] Based on the personalized physiological baseline model, determining whether the wearing state of the wearer is abnormal;

[0015] When it is determined that the wearing state is abnormal, a vibration reminder is triggered;

[0016] The wearing status and warning information are transmitted to the supervision center through the wireless communication module.

[0017] Preferably, the flexible sensor network comprises:

[0018] Flexible sensors are installed at the waist, chest and trouser cuffs of miners’ uniforms;

[0019] The flexible sensor is made of at least one of silicone, polydimethylsiloxane, Ecoflex, polyurethane or rubber.

[0020] Preferably, the localized data processing specifically includes:

[0021] Using a Butterworth digital filter to remove abnormal values ​​in the heart rate, body temperature and respiratory rate signals;

[0022] The sliding average filtering method is used to smooth the filtered signal;

[0023] Perform frequency domain analysis on the processed signal through Fourier fast transform to generate a spectrum diagram;

[0024] The values ​​of heart rate, body temperature and respiratory rate are calculated based on the spectrogram.

[0025] Preferably, the multimodal data fusion algorithm specifically includes:

[0026] Extracting waveform features of the physiological feature data;

[0027] Combining the action state data, constructing a multi-dimensional feature sequence;

[0028] The multi-dimensional feature sequence is fused using a long short-term memory network (LSTM) to obtain fused feature data.

[0029] Preferably, the deep learning algorithm specifically includes:

[0030] Using a convolutional neural network to extract features from the fused feature data;

[0031] A time series prediction model is established using a recurrent neural network;

[0032] The feature extraction results and the time series prediction model are combined to generate a personalized physiological baseline model.

[0033] Preferably, the method further includes an adaptive threshold adjustment step:

[0034] Obtain information about the wearer’s age, work environment, and work tasks;

[0035] constructing a comprehensive status model based on the age, work environment and work task information;

[0036] According to the comprehensive state model, the threshold for judging abnormal wearing state is dynamically adjusted.

[0037] Preferably, the method for constructing the comprehensive state model is:

[0038] S=T·(a1-T1)+E·(E1-E0)+G·(G1-G2)+f·(f1-f)

[0039] Among them: S represents the comprehensive status score, T represents the wearer's age, E represents the working environment parameters, G represents the work task parameters, f represents the number of wearing times, a1, T1, E1, E0, G1, G2, f1 are preset thresholds.

[0040] Preferably, the method for acquiring the action status data comprises:

[0041] Using the inertial measurement unit integrated into the miner suit to collect the wearer’s motion data;

[0042] The infrared beam is projected to form a light field to capture the characteristic points of the miners' clothing;

[0043] Based on the motion data and feature points, the wearer's action types are identified, including left arm open, right arm open, standing, walking, squatting, digging, picking up, raising arms and pushing.

[0044] As a preference, it also includes a wearing normative identification step:

[0045] Based on the motion state data, identify the positions of key parts of the miner's clothing, including epaulettes, belts, waist protectors, badges, work caps, helmets, gloves and mining lamps;

[0046] Determining whether the position of the key parts meets the preset wearing specifications;

[0047] When the judgment result is not in compliance, it is treated as an abnormal wearing state.

[0048] The intelligent monitoring system for miners' clothing wearing status based on physiological feature fusion includes:

[0049] A flexible sensor network, embedded in the miners’ clothing, is used to collect the wearer’s heart rate, body temperature, and respiratory rate signals in real time;

[0050] A motion recognition module, used to obtain the wearer's motion status data;

[0051] An edge computing module is connected to the flexible sensor network and is used to perform localized data processing on the heart rate, body temperature and respiratory rate signals to obtain physiological characteristic data;

[0052] A data fusion processor is connected to the edge computing module and the action recognition module, and is used to:

[0053] Based on the physiological characteristic data and the motion state data, using a multimodal data fusion algorithm to generate fused characteristic data;

[0054] Use deep learning algorithms to build personalized physiological baseline models;

[0055] Based on the personalized physiological baseline model, determining whether the wearing state of the wearer is abnormal;

[0056] A vibration reminder module, connected to the data fusion processor, for triggering a vibration reminder when it is determined that the wearing state is abnormal;

[0057] A wireless communication module is connected to the data fusion processor and is used to transmit the wearing status and warning information to a supervision center.

[0058] The present invention has the following beneficial effects:

[0059] The innovation of the present invention is also reflected in its use of edge computing architecture to achieve localized data processing. This not only greatly improves the response speed of the system, but also reduces the pressure of data transmission. At the same time, the system introduces an adaptive threshold adjustment mechanism, which can dynamically adjust the monitoring parameters according to the working environment and task characteristics, further improving the adaptability and accuracy of the system.

[0060] This multi-dimensional and multi-level monitoring method solves the problems of data fragmentation and slow response in traditional systems. Through the organic integration of physiological characteristics, motion status and environmental information, the present invention realizes intelligent and comprehensive monitoring of the wearing status of miners' clothing. The system can not only detect physiological abnormalities in a timely manner, but also identify potential risk factors such as improper wearing, greatly improving the comprehensiveness and foresight of safety monitoring.

[0061] In addition, the method of the present invention has a significant synergistic effect. For example, motion state recognition not only provides a basis for judging the wearing standard, but also helps explain the fluctuation of physiological data and improve the accuracy of abnormal detection. The combination of edge computing and adaptive threshold adjustment realizes the organic unity of rapid response and personalized monitoring, which not only ensures the real-time performance of the system, but also improves the accuracy of monitoring.

[0062] In general, the method proposed in the present invention not only overcomes the limitations of the existing technology, but also achieves functional complementarity and effect superposition through the organic combination of multiple innovative technologies. This all-round and intelligent monitoring method greatly improves the level of miners' safety protection and provides a strong guarantee for safe production in the mining industry. At the same time, the ideas and methods of the present invention also provide new solutions for other fields that require wearable monitoring, and have broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a system logic block diagram of the present invention.

[0064] Figure 2 is a flow chart of the method of the present invention.

[0065] Figure 3 This is a flow chart of the method for obtaining action status data of the present invention. DETAILED DESCRIPTION

[0066] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0067] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0068] Please refer to Figure 1-3 The present invention discloses an intelligent monitoring method and system for the wearing status of miners' clothing based on physiological feature fusion. The method embeds a flexible sensor network in the miners' clothing to collect the wearer's physiological data in real time, and combines the action status information to realize intelligent monitoring of the wearing status of miners' clothing.

[0069] First, the method includes an acquisition step. In this step, the wearer's heart rate, body temperature and respiratory rate signals are collected in real time through the flexible sensor network embedded in the miner's clothing. These physiological signals are important indicators for evaluating the miner's health status and wearing status. At the same time, the method also obtains the wearer's motion state data. These motion data can help us understand the miner's working status and wearing status more comprehensively.

[0070] Next is the processing step. In this step, the present invention first uses the edge computing module 1 to perform localized data processing on the collected heart rate, body temperature and respiratory rate signals. This localized processing can effectively reduce the amount of data transmission and improve the system response speed. After processing, we obtain physiological characteristic data. Preferably, the present invention uses a Butterworth filter to remove outliers in the signal, and then uses a sliding average filter to smooth the signal. For example, for the heart rate signal, we can set the cutoff frequency of the filter to 0.5Hz-3Hz. This range can effectively filter out most of the noise while retaining useful heart rate information.

[0071] Then, the method generates fused feature data based on the physiological feature data and the motion state data using a multimodal data fusion algorithm. This multimodal fusion method can capture the miner's state information more comprehensively and improve the accuracy of monitoring. In one embodiment of the present invention, a long short-term memory network (LSTM) is used for feature fusion, and the number of hidden layer nodes of the LSTM network can be set to 128. This setting performs well in practice and can effectively capture time series features.

[0072] Next, the present invention uses a deep learning algorithm to establish a personalized physiological baseline model. The key to this step is that we do not use a unified standard, but instead establish a personalized model for each miner. This can better adapt to the individual differences of different miners and improve the accuracy of monitoring. For example, we can use a three-layer convolutional neural network for feature extraction, and the convolution kernel size of each layer is set to 3x3. This setting can effectively capture local features.

[0073] Finally, based on the personalized physiological baseline model, the method determines whether the wearer's wearing status is abnormal. If the wearing status is judged to be abnormal, the system will trigger a vibration reminder. This instant feedback mechanism can help miners adjust their wearing status in time to prevent potential safety hazards. At the same time, the present invention transmits the wearing status and warning information to the supervision center through the wireless communication module, realizing remote monitoring.

[0074] Furthermore, the flexible sensing network of the present invention includes flexible sensors arranged at the waist, chest and trouser cuffs of the miner's uniform. This distributed sensor layout can comprehensively collect the physiological signals of the miners. The flexible sensor is made of at least one material selected from silicone, polydimethylsiloxane, Ecoflex, polyurethane or rubber. These materials have good flexibility and durability and are suitable for long-term wear. For example, the elastic modulus of silicone material is generally between 0.1-1MPa, and this softness can ensure that the sensor fits the body without affecting the miner's activities.

[0075] The localized data processing of the present invention also includes the following specific steps: First, a Butterworth digital filter is used to remove abnormal values ​​in the heart rate, body temperature and respiratory rate signals. The order of the Butterworth filter can be selected from 4 to 8 orders, which can achieve a good balance between retaining the main features of the signal and removing noise. Then, the filtered signal is smoothed using a sliding average filter. The size of the sliding window can be selected from 5 to 10 seconds, which can effectively smooth short-term fluctuations without excessively blurring the signal features.

[0076] Next, the method performs frequency domain analysis on the processed signal through Fourier fast transform to generate a spectrum. The sampling frequency of the Fourier transform can be set to twice the sampling frequency of the original signal to avoid aliasing effects. Finally, the values ​​of heart rate, body temperature, and respiratory rate are calculated based on the spectrum. For example, the heart rate can be calculated by identifying the main peak in the range of 0.5-3Hz in the spectrum, which corresponds to the normal heart rate range of 30-180 beats / minute.

[0077] The multimodal data fusion algorithm of the present invention specifically includes the following steps: First, extract the waveform features of the physiological characteristic data. Here, features such as heart rate variability and peak-to-valley ratio of the respiratory waveform can be extracted. Then, in combination with the motion state data, a multidimensional feature sequence is constructed. For example, the physiological features and motion features can be aligned in the time dimension to form a multidimensional feature vector for each time point. Finally, the multidimensional feature sequence is fused using a long short-term memory network (LSTM) to obtain fused feature data. The time step of the LSTM network can be set to 60 seconds, which can capture sufficiently long time dependencies.

[0078] In terms of deep learning algorithms, the present invention specifically includes the following steps: First, a convolutional neural network is used to extract features from the fused feature data. The convolutional network can be set to 3-5 layers, and the number of convolution kernels in each layer can start from 32 and double layer by layer. This setting can gradually extract higher-level features. Then, a recurrent neural network is used to establish a time series prediction model. The recurrent network can select LSTM or GRU units, and the hidden layer size can be set to 128-256, which performs well in most tasks. Finally, a personalized physiological baseline model is generated by combining the feature extraction results and the time series prediction model.

[0079] This intelligent monitoring method based on physiological feature fusion has many advantages. First, it realizes comprehensive monitoring of the miners' physiological state, including not only conventional heart rate and body temperature, but also indicators such as respiratory rate, which can more comprehensively evaluate the miners' health status. Secondly, by fusing motion state data, this method can more accurately judge the wearing status and reduce false positives and false negatives. Furthermore, the personalized physiological baseline model can adapt to the individual differences of different miners and improve the accuracy of monitoring. Finally, the application of localized data processing and edge computing greatly improves the response speed of the system, so that abnormal conditions can be discovered and processed in a timely manner.

[0080] In general, the present invention provides an innovative intelligent monitoring method for the wearing status of miners' clothing, which combines advanced sensing technology, data processing algorithms and artificial intelligence technology to provide more reliable protection for the safety of miners. This method can not only be applied to miners' clothing, but also be extended to other fields that require wearing status monitoring, and has broad application prospects. The present invention further proposes an adaptive threshold adjustment step, which enables the system to dynamically adjust the monitoring parameters according to the individual characteristics and working environment of different miners, thereby improving the universality and accuracy of the system.

[0081] Specifically, this adaptive threshold adjustment step includes the following: First, the system obtains the wearer's age, work environment, and work task information. This information is crucial for assessing the miner's physiological state. For example, older miners may require a more relaxed heart rate threshold, while miners working in a high temperature environment may require more stringent body temperature monitoring.

[0082] Next, based on this information, the system builds a comprehensive state model. The core of this model is a mathematical formula used to calculate the comprehensive state value. The formula is as follows:

[0083] S=T·(a1-T1)+E·(E1-E0)+G·(G1-G2)+f·(f1-f)

[0084] Among them: S represents the comprehensive status score, T represents the wearer's age, E represents the working environment parameters, G represents the work task parameters, f represents the number of wearing times, a1, T1, E1, E0, G1, G2, f1 are preset thresholds.

[0085] The design of this formula takes into account the impact of multiple factors on the wearable state. For example, the a1-T1 item reflects the impact of age on the threshold. The older you are, the larger the value of this item is, which means the system will give a more relaxed threshold. The E1-E0 items take into account the impact of the working environment. If the severity of the environment exceeds the preset threshold, the system will adjust the monitoring parameters accordingly.

[0086] Based on this comprehensive status model, the system can dynamically adjust the threshold for abnormal wear status judgment. For example, for heart rate monitoring, the initial threshold may be set to 60-100 beats / minute. However, if the comprehensive status score S is high, the system may relax the threshold to 55-105 beats / minute. This dynamic adjustment can reduce false alarms and improve system availability.

[0087] Then, the comprehensive state score is embedded into the Riemann manifold:

[0088] P = log I (C S )=Ulog(Λ)U T

[0089] Among them, S is the comprehensive status score, P is the Riemann representation of the status score, and C S is the covariance matrix constructed based on S. Next, the physiological features are fused with the state scores:

[0090]

[0091] Among them, Q is the fused feature vector, represents the Kronecker product, is Haar transform, and B is the physiological feature vector (including heart rate, body temperature, respiratory rate, etc.).

[0092] Then, the action state information is integrated into the feature space:

[0093] R=Φ k (Q,A)

[0094] Among them, R is the feature vector for further fusion, Φ k represents the k-th order continuous homology operator, is the action state vector. In order to capture complex nonlinear relationships, this method introduces the modified Ramanujan summation formula:

[0095]

[0096] Where S is the extracted nonlinear feature, (a) n′ (b) n′ (c) n is the Pochhammer symbol, a, b, c are unknown parameters, and N is the cutoff parameter used to control the convergence of the series.

[0097] Finally, this method combines the threshold adjustment mechanism and uses an improved discrimination method to identify abnormal states:

[0098]

[0099] Where D is the degree of abnormality, erfc is the complementary error function, is the feature vector of the normal state, and σ(S) is the standard deviation dynamically adjusted based on the comprehensive state score S. The comprehensive state score is embedded in the Riemann manifold, which improves the ability to capture nonlinear relationships, realizes the deep fusion of physiological characteristics, action state and comprehensive state score, introduces the exponential decay term, ensures the convergence of the series, and makes the algorithm more stable. The abnormal state discrimination method of the present invention takes the comprehensive state score into consideration and realizes adaptive threshold adjustment.

[0100] The method of the present invention realizes the deep fusion of multi-source information, enhances the nonlinear expression ability and environmental adaptability of the algorithm, can better cope with the complex mine environment, and improves the accuracy and reliability of wearable status monitoring.

[0101] The present invention also introduces a wearing normative identification step, which is intended to ensure that miners wear protective equipment correctly and further improve safety. This step includes the following:

[0102] First, the system identifies the locations of key parts of the miner's uniform based on the action state data. These key parts include epaulettes, belts, waist guards, badges, work caps, helmets, gloves, and miner's lamps. During the recognition process, the system may use pre-trained image recognition models, such as object detection algorithms based on convolutional neural networks. For example, the YOLO (You Only Look Once) algorithm can be used, which performs well in real-time object detection tasks and can quickly and accurately locate key parts.

[0103] Next, the system determines whether the positions of these key parts meet the preset wearing specifications. The judgment criteria here may include the accuracy and stability of the position. For example, for a hard hat, the system may check whether it covers the head correctly and whether the hat strap is fastened. To achieve this function, the system may use a deep learning-based posture estimation algorithm, such as OpenPose, to accurately locate the key points of the miner's body, and then compare them with the preset wearing standards.

[0104] If the system determines that a certain part of the wear is not in compliance with the specification, it will treat this as an abnormal wearing status. This may include triggering an immediate vibration reminder to prompt the miner to adjust the corresponding equipment. At the same time, the system will also transmit this information to the supervision center so that the management personnel can understand the miner's wearing status in a timely manner.

[0105] Another important feature of the present invention is a method for obtaining action state data. This method includes the following aspects:

[0106] First, the system uses the inertial measurement unit integrated into the miner's suit to collect the wearer's motion data. The inertial measurement unit usually includes an accelerometer and a gyroscope, which can accurately capture the miner's motion state. For example, the accelerometer's measurement range may be set to ±16g and the sampling frequency is 100Hz, which can capture most human body movements.

[0107] Secondly, the system forms a light field through infrared beam projection to capture the characteristic points of the miner's clothing. This method can accurately capture the miner's movements even in a low-light mine environment. The wavelength of the infrared light source may be selected to be around 850nm, which is safe for the human eye and can be effectively captured by ordinary CMOS sensors.

[0108] Based on these data, the system identifies the wearer's action types, including left arm open, right arm open, standing, walking, squatting, digging, picking up objects, raising arms, and pushing objects. In order to achieve accurate action recognition, the system may use deep learning algorithms such as long short-term memory networks (LSTM) or temporal convolutional networks (TCN). These algorithms can effectively process time series data and capture the temporal characteristics of actions.

[0109] Finally, the present invention also includes an intelligent monitoring system for the wearing status of miners' clothing based on the fusion of physiological characteristics. This system consists of multiple modules, each of which has its specific function:

[0110] The flexible sensor network module 2 is embedded in the miner's clothing to collect the wearer's heart rate, body temperature and respiratory rate signals in real time. This network may consist of multiple flexible sensors distributed in different locations, each of which is specially processed to adapt to the harsh environment of the mine.

[0111] The motion recognition module 3 is used to obtain the wearer's motion state data. This module may include the inertial measurement unit and infrared projection system mentioned above.

[0112] The edge computing module 1 is connected to the flexible sensor network module 2 to perform local data processing on the heart rate, body temperature and respiratory rate signals to obtain physiological characteristic data. This module may use a low-power microprocessor, such as the ARMCortex-M series, to achieve efficient local computing.

[0113] The data fusion processor 4 is the core of the system, which is connected to the edge computing module 1 and the motion recognition module 3. This processor is responsible for executing the multimodal data fusion algorithm, establishing a personalized physiological baseline model, and determining whether the wearing state is abnormal. Considering the computing requirements, this processor may use a high-performance embedded processor, such as the Nvidia Jetson series.

[0114] The vibration reminder module 5 is connected to the data fusion processor 4 and is used to trigger a vibration reminder when the wearing state is judged to be abnormal. This module may use a small vibration motor to generate vibrations of different intensities and patterns to convey different levels of warnings.

[0115] The wireless communication module 6 is also connected to the data fusion processor 4 to transmit the wear status and warning information to the supervision center. Considering the particularity of the mine environment, this module may use low-power wide area network technology, such as LoRa or NB-IoT, to ensure reliable communication in complex environments.

[0116] Through the collaborative work of these modules, the system can comprehensively, accurately and in real time monitor the wearing status of miners, timely discover potential safety hazards, and provide strong protection for the life safety of miners. At the same time, the modular design of the system also facilitates future upgrades and expansions, allowing the system to continuously adapt to new needs and technological developments. In order to verify the superiority of the present invention, we conducted a series of experiments, and the following will introduce in detail the settings, test methods and result analysis of the embodiments and comparative examples.

[0117] Example 1: We selected a typical coal mine working environment for testing. The test subjects were 50 male miners aged between 28 and 45 years old, with an average age of 35.6 years old. These miners were randomly divided into two groups, with 25 people in each group. The experimental group used the intelligent monitoring system of the present invention, and the control group used the traditional fixed threshold monitoring system.

[0118] The test environment was a coal mining face 800 meters underground, with an ambient temperature of about 28°C, a relative humidity of 85%, and a dust concentration of 3 mg / m 3 The test duration is 30 consecutive working days, 8 working hours per day.

[0119] The system configuration of the present invention is as follows: the flexible sensor network is made of polydimethylsiloxane material and integrated in the chest, waist and wrist of the miner's uniform. The edge computing module uses an ARM Cortex-M4 processor with an operating frequency of 80MHz. The data fusion processor uses Nvidia JetsonNano with 128 CUDA cores. The wireless communication module uses LoRa technology with an operating frequency band of 470MHz.

[0120] Comparative Example 1: The traditional fixed threshold monitoring system used in the control group includes independent heart rate monitors, thermometers, and respiratory monitors. These devices are worn on different parts of the body, and no data fusion is performed. The system uses fixed thresholds for abnormal judgment, such as a heart rate threshold of 60-100 beats / minute and a body temperature threshold of 36.5-37.5℃.

[0121] Test indicators and methods:

[0122] 1. Accuracy of abnormal wearing status detection: Manually observe and record actual abnormal wearing status and compare them with the abnormalities reported by the system.

[0123] 2. System response time: the time interval from the occurrence of an abnormal state to the system issuing an alarm.

[0124] 3. False alarm rate: The proportion of abnormalities reported by the system but actually normal.

[0125] 4. False negative rate: the proportion of abnormalities that are not reported by the system but actually exist.

[0126] 5. Battery life: The time the system can continue to work under normal working conditions.

[0127] 6. Wearing comfort: Through a questionnaire survey, miners are asked to rate the comfort of the system on a scale of 1-10.

[0128] The test results are shown in the following table:

[0129] Test indicators Example 1 Comparative Example 1 Accuracy of abnormal detection of wearing status 95.8 82.3 System response time 0.5 2.3 False Positive Rate 3.2 12.7 False negative rate 1 5 Battery life 16 10 Wearing comfort (1-10 points) 8.5 6.2

[0130] By comparing the test results, it can be seen that the intelligent monitoring system of the present invention is significantly superior to the traditional system in all indicators. In particular, in terms of the accuracy of abnormal wearing state detection, the system of the present invention achieved a high accuracy of 95.8%, which is 13.5 percentage points higher than the traditional system. This result fully proves the effectiveness of the multimodal data fusion and deep learning algorithm in the present invention.

[0131] System response time is another outstanding advantage. The system of the present invention can respond to abnormal conditions in just 0.5 seconds, while the traditional system takes 2.3 seconds. This rapid response capability may save lives in emergency situations. This is due to the edge computing architecture adopted by the present invention, which greatly reduces the delay of data transmission and processing.

[0132] The significant reduction in false alarm rate and missed alarm rate (from 12.7% to 3.2% and from 5.0% to 1.0% respectively) demonstrates the superiority of the adaptive threshold adjustment mechanism of the present invention. This mechanism can dynamically adjust the judgment criteria according to individual differences and environmental changes, greatly improving the reliability of the system.

[0133] It is worth noting that despite the fact that the system of the present invention integrates more functions, its battery life is 6 hours longer than that of the traditional system. This shows that the present invention has also made optimizations in terms of energy efficiency, which may be achieved through low-power design and intelligent power management.

[0134] Finally, the improvement of wearing comfort (from 6.2 points to 8.5 points) proves the consideration of the invention in human-computer interaction design. The application of flexible sensor network makes the whole system more portable and comfortable, and will not affect the normal work of miners.

[0135] In general, these test results fully demonstrate the superiority of the system of the present invention in terms of accuracy, real-time performance, reliability, energy efficiency, and user experience. This comprehensive improvement can not only better protect the safety of miners, but also improve their work efficiency and comfort. Especially in complex mine environments, these advantages of the system of the present invention may play a decisive role at critical moments and provide more reliable protection for the life safety of miners.

[0136] In addition, these results also confirm the effectiveness of many innovations in this invention, such as flexible sensor networks, multimodal data fusion, edge computing architecture, adaptive threshold adjustment, etc. These innovations are not only of great significance in the field of miner safety monitoring, but may also provide new ideas and solutions for other fields that require wearable monitoring.

[0137] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent monitoring method for the wearing status of miners' clothing based on physiological feature fusion, characterized in that: The following steps are involved: Through the flexible sensor network embedded in the miners' clothing, the wearer's heart rate, body temperature and respiratory rate signals are collected in real time; Obtain the wearer's motion status data; Based on the heart rate, body temperature and respiratory rate signals, an edge computing module is used to perform localized data processing to obtain physiological characteristic data; Based on the physiological characteristic data and the motion state data, using a multimodal data fusion algorithm to generate fused characteristic data; Establishing a personalized physiological baseline model using a deep learning algorithm based on the fused feature data; Based on the personalized physiological baseline model, determining whether the wearing state of the wearer is abnormal; When it is determined that the wearing state is abnormal, a vibration reminder is triggered; The wearing status and warning information are transmitted to the supervision center through the wireless communication module.

2. The method according to claim 1, characterized in that The flexible sensing network comprises: Flexible sensors are installed at the waist, chest and trouser cuffs of miners’ uniforms; The flexible sensor is made of at least one of silicone, polydimethylsiloxane, Ecoflex, polyurethane or rubber.

3. The method according to claim 1, characterized in that The localized data processing specifically includes: Using a Butterworth digital filter to remove abnormal values ​​in the heart rate, body temperature and respiratory rate signals; The sliding average filtering method is used to smooth the filtered signal; Perform frequency domain analysis on the processed signal through Fourier fast transform to generate a spectrum diagram; The values ​​of heart rate, body temperature and respiratory rate are calculated based on the spectrogram.

4. The method according to claim 1, characterized in that The multimodal data fusion algorithm specifically includes: Extracting waveform features of the physiological feature data; Combining the action state data, constructing a multi-dimensional feature sequence; The multi-dimensional feature sequence is fused using a long short-term memory network (LSTM) to obtain fused feature data.

5. The method according to claim 1, characterized in that The deep learning algorithm specifically includes: Using a convolutional neural network to extract features from the fused feature data; A time series prediction model is established using a recurrent neural network; The feature extraction results and the time series prediction model are combined to generate a personalized physiological baseline model.

6. The method according to claim 1, characterized in that It also includes an adaptive threshold adjustment step: Obtain information about the wearer’s age, work environment, and work tasks; constructing a comprehensive status model based on the age, work environment and work task information; According to the comprehensive state model, the threshold for judging abnormal wearing state is dynamically adjusted.

7. The method according to claim 6, characterized in that The construction method of the comprehensive state model is: S=T·(a1-T1)+E·(E1-E0)+G·(G1-G2)+f·(f1-f) Among them: S represents the comprehensive status score, T represents the wearer's age, E represents the working environment parameters, G represents the work task parameters, f represents the number of wearing times, a1, T1, E1, E0, G1, G2, f1 are preset thresholds.

8. The method according to claim 1, characterized in that The method for acquiring the action status data comprises: Using the inertial measurement unit integrated into the miner suit to collect the wearer’s motion data; The infrared beam is projected to form a light field to capture the characteristic points of the miners' clothing; Based on the motion data and feature points, the wearer's action types are identified, including left arm open, right arm open, standing, walking, squatting, digging, picking up, raising arms and pushing.

9. The method according to claim 1, characterized in that: It also includes wearing normative identification steps: Based on the motion state data, identify the positions of key parts of the miner's clothing, including epaulettes, belts, waist protectors, badges, work caps, helmets, gloves and mining lamps; Determining whether the position of the key parts meets the preset wearing specifications; When the judgment result is not in compliance, it is treated as an abnormal wearing state.

10. An intelligent monitoring system for the wearing status of miners' clothing based on physiological feature fusion, which implements the method described in any one of claims 1 to 9, characterized in that: include: A flexible sensor network, embedded in the miners’ clothing, is used to collect the wearer’s heart rate, body temperature, and respiratory rate signals in real time; A motion recognition module, used to obtain the wearer's motion status data; An edge computing module is connected to the flexible sensor network and is used to perform localized data processing on the heart rate, body temperature and respiratory rate signals to obtain physiological characteristic data; A data fusion processor is connected to the edge computing module and the action recognition module, and is used to: Based on the physiological characteristic data and the motion state data, using a multimodal data fusion algorithm to generate fused characteristic data; Use deep learning algorithms to build personalized physiological baseline models; Based on the personalized physiological baseline model, determining whether the wearing state of the wearer is abnormal; A vibration reminder module, connected to the data fusion processor, for triggering a vibration reminder when it is determined that the wearing state is abnormal; A wireless communication module is connected to the data fusion processor and is used to transmit the wearing status and warning information to a supervision center.

Images