Self-rescuer intelligent training network and data processing method for multi-user synchronous training

Abstract

The invention relates to the technical field of intelligent training networks and data processing, and provides a self-rescuer intelligent training network for multi-user synchronous training and a data processing method, behavior data and video data of users are acquired through a data acquisition node, the behavior data and the video data of the users are processed, and a processing result is generated; the training network can monitor the activity and operation conditions of the user in real time; the result judgment node judges the behavior result of the user according to the processing result, the behavior result of the user can be recognized, and the habit and potential problems of the user when the user uses the self-rescuer can be known; the feedback control node generates training feedback according to the behavior result of the user and judges the state of the self-rescuer, so that the user can be helped to correct errors, and the operation skill is improved; and the storage management node stores the behavior data and video data of the user and the behavior result of the user to the cloud for centralized management.

Description

Technical Field

[0001] This application relates to the technical fields of intelligent training networks and data processing, and in particular, to a self-rescuer intelligent training network and data processing method for multi-user synchronous training. Background Art

[0002] Mine production has been accompanied by high risks for a long time. Underground workers are often harmed by geological disasters, toxic gases, and enclosed oxygen-deficient environments. With the rapid development of the country, the whole society's attention to the personal safety of workers under dangerous production conditions has been increasing year by year. Among them, the correct use of mine self-rescuers is one of the essential skills. However, almost every year, there are casualties caused by improper use of self-rescuers. In terms of the proficient use of self-rescuers, there is a lack of relatively simple and quantifiable training methods. With the continuous progress of artificial intelligence and Internet of Things technologies, it is already possible to achieve a training method that is digital, intelligent, and has a low learning cost.

[0003] By simulating the self-rescue process in a real environment, it helps mine workers improve their self-rescue ability in dangerous environments. At the same time, it can also be used for safety self-rescue drills in other fields. By collecting and analyzing the operation data of users in real time, it provides instant training feedback and suggestions for users, helps to detect and correct incorrect operations in a timely manner, and can provide personalized training plans and intervention measures for users to improve the training effect. However, most of them do not solve the problem of how to obtain user data through a sensor network to identify the user's behavior pattern, output control instructions through the self-rescuer and feedback them to the user in real time, and at the same time, centralize the management of relevant data and ensure that information is obtained in a timely manner and does not interfere with each other under the conditions of batch user simultaneous operations and data concurrency.

[0004] For example, the Chinese patent with the authorization announcement number CN110162785B discloses a data method, device, readable storage medium, and computer device. The method includes: obtaining a text to be detected; inputting the context word set and candidate replacement word set corresponding to the text to be detected into a pronoun resolution neural network, respectively extracting features of the context word set and the candidate replacement word set to obtain corresponding first features and second features, performing positive example iterative processing and negative example iterative processing on the first features and the second features to obtain the positive and negative example feature vector norms, calculating the replacement probabilities corresponding to each candidate replacement word in the candidate replacement word set according to the positive and negative example feature vector norms; determining a target replacement word according to the replacement probabilities corresponding to each candidate replacement word; and inserting the target replacement word into the text to be detected according to the position corresponding to the word to be detected to obtain a target text. In addition, a pronoun resolution neural network training method, device, readable storage medium, and computer device are also provided.

[0005] As disclosed in the Chinese patent with the authorization announcement number CN113138832B, a distributed training method and system based on resetting a training data transmission network are provided. Among them, first, a method for resetting a training data transmission network during distributed training is provided. By forming a logical ring with all virtual nodes running training processes and collecting global RDMA network configurations respectively in the forward direction downstream from the starting node of the logical ring and in the reverse direction upstream from the terminating node, and resetting the training data transmission network configuration in the training cluster to an RDMA network according to the global RDMA network configuration by each node's training process, the reset of the training data transmission network during distributed training is realized. Furthermore, on this basis, an accelerated distributed training method and an accelerated distributed training system based on an efficient communication network are provided to realize accelerated distributed training with the training data transmitted through an efficient RDMA network during the training process.

[0006] The above patents have the problems raised in this background technology: The above two patents do not solve the problems of how to obtain user data through a sensor network to identify the user's behavior pattern, output control instructions through a self-rescuer and feedback them to the user in real time, and the centralized management of relevant data, and ensuring that information is obtained in a timely manner and does not interfere with each other under the conditions of batch user operations and data concurrency. Summary of the Invention

[0007] In view of the deficiencies of the prior art, the present application provides a self-rescuer intelligent training network and data processing method for multi-user synchronous training.

[0008] In a first aspect, the present application provides a self-rescuer intelligent training network for multi-user synchronous training, which includes: a data acquisition node, a result judgment node, a feedback control node, and a storage management node;

[0009] The data acquisition node is used to acquire the user's behavior data and video data, and process the user's behavior data and video data to generate a processing result;

[0010] The result judgment node is used to judge the user's behavior result according to the processing result;

[0011] The feedback control node is used to generate a training feedback according to the user's behavior result and judge the state of the self-rescuer;

[0012] The storage management node is used to store the user's behavior data, video data, and the user's behavior result in local and cloud for centralized management.

[0013] As an optional implementation manner, the generation strategy of the processing result includes:

[0014] Deploy a camera to obtain the user's video data, determine whether the user performing the current operation is the target user, analyze the user's posture, and judge the accuracy of the user's posture;

[0015] Obtain the user's behavior data through Hall elements and touch sensing, analyze the user's behavior data, compare the user's behavior data with the standard operation sequence, and verify the accuracy of the user's sequence;

[0016] Comprehensively judge the accuracy of the user's operation based on the posture accuracy and sequence accuracy of the user.

[0017] As an alternative implementation, the judgment logic of the posture accuracy includes:

[0018] Compare the user's posture with the standard posture to judge the similarity between the user's posture and the standard posture;

[0019] Configure a similarity threshold. If the similarity between the user's posture and the standard posture is less than or equal to the similarity threshold, the user's posture is inaccurate;

[0020] If the similarity between the user's posture and the standard posture is greater than the similarity threshold, the user's posture is accurate.

[0021] As an alternative implementation, the judgment logic of the similarity between the user's posture and the standard posture includes:

[0022] Identify the target nodes of the user's posture and the target nodes of the standard posture;

[0023] Calculate the Euclidean distance between the target nodes of the user's posture and the corresponding target nodes of the standard posture;

[0024] Sum the Euclidean distances between the target nodes of each user's posture and the corresponding target nodes of the standard posture after multiplying by the weights of the corresponding target nodes to obtain a comprehensive distance value;

[0025] Subtract the comprehensive distance value from 1 to obtain the similarity between the user's posture and the standard posture.

[0026] As an alternative implementation, the judgment logic of the sequence accuracy includes:

[0027] Arrange Hall elements and touch sensing at the mouthpiece and nose clip, and continuously monitor the states of the Hall elements and touch sensing;

[0028] Configure an initial baseline threshold. If the magnetic field value of the Hall element and the capacitance value of the touch sensing are less than the initial baseline threshold, it is determined that the self-rescuer is in an unused state and wait for the user to train;

[0029] When the capacitance value of the touch sensing at the mouthpiece changes, it is determined that the step of wearing the mouthpiece is started;

[0030] Configure the wearing time and capacitance threshold. If the capacitance value detected by the touch sensor at the mouthpiece is less than the capacitance threshold within the wearing time, it is determined that the mouthpiece is worn overtime, and the user is prompted by voice to speed up the training operation;

[0031] If the capacitance value detected by the touch sensor at the mouthpiece is less than the capacitance threshold, it is determined that the mouthpiece is not worn accurately, and the user is prompted by voice to adjust the posture and wear the mouthpiece again.

[0032] As an optional implementation manner, the judgment logic of the sequence accuracy further includes:

[0033] After the mouthpiece is determined to be worn correctly, continue to monitor the Hall element and touch sensor at the nose clip;

[0034] Configure the magnetic field threshold. If the magnetic field value of the Hall element at the nose clip is greater than or equal to the magnetic field threshold, it means that the nose clip has been opened, and wait for the nose clip to be worn and confirmed;

[0035] If the capacitance value detected by the touch sensor at the nose clip is greater than or equal to the capacitance threshold and the magnetic field value of the Hall element at the nose clip is greater than or equal to the magnetic field threshold, it is determined that the nose clip is worn correctly;

[0036] If the capacitance value detected by the touch sensor at the nose clip is less than the capacitance threshold or the Hall element at the nose clip is less than the magnetic field threshold, it is determined that the nose clip is worn incorrectly, and the user is prompted by voice to wear the nose clip again;

[0037] Monitor the behavior data of each step when the user performs the self-rescuer training and compare it with the standard operation sequence, and record whether the behavior data of each step of the user conforms to the standard operation sequence and the operation time of each step to judge the sequence accuracy of the user.

[0038] As an optional implementation manner, the self-rescuer status includes normal, abnormal, and to be reset;

[0039] The logic for judging the self-rescuer status includes:

[0040] When the training feedback is generated, initially check the self-rescuer status by monitoring the working parameters of the self-rescuer. If the self-rescuer status is normal, it means that the self-rescuer reset is successful;

[0041] If the self-rescuer status is abnormal or to be reset, perform the reset operation. After performing the reset operation, judge the self-rescuer status. If the reset is not successful, trigger a prompt.

[0042] As an optional implementation manner, the logic for storing to local and centralized cloud management includes:

[0043] A storage structure table is given based on the user's behavior data, video data, and the user's result pattern. A unique user ID is assigned to each user, and the storage requirements are calculated according to the storage structure table. The user's identity authentication is completed according to the storage structure table, and the user's behavior data, video data, and the user's behavior results are visually displayed.

[0044] In a second aspect, the present application provides a self-rescuer intelligent training data processing method for multi-user synchronous training. The method includes: S1. Obtain the user's behavior data and video data, and process the user's behavior data and video data to generate a processing result;

[0045] S2. Judge the user's behavior result according to the processing result;

[0046] S3. Generate training feedback according to the user's behavior result, and judge the state of the self-rescuer;

[0047] S4. Store the user's behavior data, video data, and the user's behavior results in local and cloud centralized management.

[0048] Compared with the prior art, the beneficial effects of the present application are as follows: The behavior data and video data of the user are obtained through the data acquisition node, and the behavior data and video data of the user are processed to generate a processing result, so that the training network can instantly monitor the activities and operations of the user; The result judgment node judges the user's behavior result according to the processing result, and can identify the user's behavior result, which helps to understand the user's habits and potential problems when using the self-rescuer; The feedback control node generates training feedback according to the user's behavior result and judges the state of the self-rescuer, which can help the user correct mistakes and improve operation skills; The storage management node stores the user's behavior data, video data, and the user's behavior results in cloud centralized management; enabling the training network to deeply analyze a large amount of data, thereby optimizing the intelligent level of the training network; The training network can timely detect abnormal behaviors or potential safety risks, and take measures through the feedback control node to prevent accidents from occurring. Description of the Drawings

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to these drawings without creative efforts. Among them:

[0050] Figure 1 It is the network structure diagram of the self-rescuer intelligent training network for multi-user synchronous training provided by the embodiment of the present application;

[0051] Figure 2Generate a processing result generation strategy diagram for the self-rescuer intelligent training network with multi-user synchronous training provided in the embodiments of the present application;

[0052] Figure 3 Generate a self-rescuer status judgment logic diagram for the self-rescuer intelligent training network with multi-user synchronous training provided in the embodiments of the present application;

[0053] Figure 4 Generate a method flowchart for the self-rescuer intelligent training data processing method with multi-user synchronous training provided in the embodiments of the present application. Detailed implementation manners

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present application more obvious and understandable, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings of the specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments.

[0055] Embodiment 1

[0056] As Figure 1 shown, the present application provides a network structure diagram of a self-rescuer intelligent training network with multi-user synchronous training. The network includes a data acquisition node, a result judgment node, a feedback control node, and a storage management node.

[0057] The data acquisition node is used to acquire the behavior data and video data of the user, and process the behavior data and video data of the user to generate a processing result.

[0058] The generation strategy of the processing result is as Figure 2 shown, and specifically includes:

[0059] Deploy a camera to acquire the video data of the user, determine whether the currently operating user is the target user, and analyze the user's posture to determine the accuracy of the user's posture.

[0060] It should be understood that when performing multi-user synchronous training of the self-rescuer, the camera needs to be deployed at a suitable position in the training scenario, and the video image stream (video data) of the user in the operation area needs to be acquired in real time. The camera needs to clearly capture the facial features, limb movements, and overall posture details of the user. At the same time, the acquired video data needs to be attached with timestamp information to ensure the timing accuracy of the video data; to detect whether the currently operating user is the user himself and the user's posture, and here it is used to determine the accuracy of the user's posture.

[0061] Based on video image processing technology, extract the facial key points of the user (such as the contours of eyes, nose, and mouth, etc.) from the acquired video frame sequence, compare them with the facial key points of the target user pre-recorded, identify the identity of the user, and determine whether the currently operating user is himself.

[0062] When determining whether the current operating user is the same person based on the facial key points of the user, through a comparison method based on feature vectors, assuming that the features extracted from the facial image are represented in vector form, then the facial feature vector of the current user extracted from the video frame image is The facial feature vector of the pre-recorded target user is Here, t represents the feature dimension. Then, by calculating the similarity between the two facial feature vectors to determine whether the current operating user is the same person, the functional expression of the similarity between the two facial feature vectors is as follows:

[0063]

[0064] In the formula, represents the similarity between the two facial feature vectors, The value range of is between -1 and 1, The closer the value of is to 1, it indicates that the directions of the two facial feature vectors are more similar, that is, the facial features of the current user and the target user are more similar. And When the value of is -1, it means that the directions of the two facial feature vectors are completely opposite, When the value of is 0, it means that the two facial feature vectors are orthogonal. In the user identity recognition scenario, usually pay attention to The situation where the value of is close to 1, represents the norm of the facial feature vector of the current user, and When all elements in the facial feature vector of the current user are 0, the norm is 0, otherwise it is greater than 0, represents the norm of the facial feature vector of the pre-recorded target user, and When all elements in the facial feature vector of the pre-recorded target user are 0, the norm is 0, otherwise it is greater than 0, f j refers to the value on the j-th dimension of the facial feature vector of the current user, representing the quantization value of a specific feature of the current user's face, such as the coordinates of facial key points, the texture feature value of a specific area, etc. Through the facial feature extraction algorithm, a series of information that can characterize facial features can be extracted from the facial image in the video frame and quantified into numerical values, and arranged in order to form a vector, f j The value range of depends on the specific extracted feature type and quantization method. The value ranges of different features are different. For example, the facial key point coordinates may take values within the image pixel coordinate range. Similarly, g j refers to the value on the j-th dimension of the facial feature vector of the pre-recorded target user, representing the quantization value of the same feature of the target user's face. Through the same facial feature extraction algorithm, features are extracted from the facial image of the target user and quantified into numerical values to form a vector, g jIts value range depends on the specific types of features extracted and the quantization methods. Different features have different value ranges. For example, the coordinates of facial key points may be within the range of image pixel coordinates. t represents the feature dimension, that is, the length of the feature vector, which is also the number of facial features extracted and is determined by the adopted facial feature extraction algorithm. Different algorithms select different numbers of facial features. t is a positive integer, and common values range from dozens to hundreds. For example, when there are 68 facial key points, t = 68.

[0065] By setting the similarity threshold T (usually with a value between 0 and 1), when the similarity between two calculated facial feature vectors is greater than this similarity threshold, the user of the current operation is determined to be the person himself / herself; otherwise, it is determined to be a non - person.

[0066] The judgment logic of pose accuracy includes:

[0067] Compare the user's pose with the standard pose to judge the similarity between the user's pose and the standard pose;

[0068] The judgment logic of the similarity between the user's pose and the standard pose includes:

[0069] Identify the target nodes of the user's pose and the target nodes of the standard pose;

[0070] Calculate the Euclidean distance between the target nodes of the user's pose and the corresponding target nodes of the standard pose;

[0071] Sum the Euclidean distances between the target nodes of each user's pose and the corresponding target nodes of the standard pose after multiplying them by the weights of the corresponding target nodes to obtain a comprehensive distance value;

[0072] Subtract the comprehensive distance value (after normalization to make the range between 0 and 1) from 1 to obtain the similarity between the user's pose and the standard pose.

[0073] Configure the similarity threshold. If the similarity between the user's pose and the standard pose is less than or equal to the similarity threshold, the user's pose is inaccurate;

[0074] If the similarity between the user's pose and the standard pose is greater than the similarity threshold, the user's pose is accurate.

[0075] Among them, it is assumed that the set of target nodes of the user's pose is U = {u 1 , u 2 ,..u. m ,}, the set of target nodes of the standard pose is S = {s 1 , s 2 ,..s. m ,}, and the set of weights of the corresponding target nodes is W = {w 1 , w 2 ,...w,m}, the calculation formula for the Euclidean distance between the target node of the user's posture and the target node of the corresponding standard posture is as follows:

[0076]

[0077] In the formula, di represents the Euclidean distance between the target node of the user's posture and the target node of the corresponding standard posture, and the value range is di≥0. Theoretically, when the target node of the user's posture completely coincides with the target node of the standard posture, di = 0. As the spatial position difference between the two target nodes increases, the value of di will also increase. (u ix , u iy , u iz ) and (s ix , s iy , s iz ) represent the coordinates of the target node ui of the user's posture and the target node si of the standard posture in three-dimensional space respectively.

[0078] Then, the calculation formula for the comprehensive distance value is as follows:

[0079]

[0080] In the formula, D represents the comprehensive distance value, di represents the Euclidean distance between the target node of the user's posture and the target node of the corresponding standard posture, m represents the total number of target nodes, w i represents the weight of the i-th target node, which is used to measure the importance of this target node in the calculation of the overall posture similarity, and the value range is 0≤w i ≤1.

[0081] Finally, the function expression for the similarity between the user's posture and the standard posture is as follows:

[0082] sim = 1 - D norm ;

[0083] In the formula, sim represents the similarity between the user's posture and the standard posture, and the value range is 0≤sim≤1. When sim = 1, it means that the user's posture is exactly the same as the standard posture. When sim = 0, it means that the difference between the user's posture and the standard posture is the largest. D norm represents the value after normalizing the comprehensive distance value. The specific normalization method needs to be determined according to the actual data range.

[0084] At the same time, for the target nodes (key limb parts, such as arms, palms, heads, and trunks, etc.) of the user in the video frame, posture estimation is performed to identify the positions of each limb joint point. The real-time posture of the user is compared with the preset standard posture to obtain the similarity between the two, and it is judged whether the user's posture is accurate. Each posture estimation cycle is 0.5 seconds.

[0085] In each 0.5-second posture estimation cycle, a video frame image is acquired through the camera. The video frame image needs to clearly capture the overall posture details of the user including key limbs (arms, palms, head and torso, etc.), and the video frame image is processed using a deep learning-based posture estimation algorithm (such as OpenPose, etc.) to identify the position coordinates of each limb joint in the video frame image. For example, for the arm, the coordinates of joints such as the shoulder, elbow and wrist can be identified, for the palm, the coordinates of key points such as the base of the palm and finger joints can be identified, for the head, the coordinates of key parts such as the top of the head, chin and ears can be identified, and for the torso, the coordinates of key points such as the neck and waist can be identified.

[0086] The user's behavior data is obtained through Hall elements and touch sensing, analyzed, compared with the standard operation sequence, and the user's sequence accuracy is verified.

[0087] It should be understood that electronic components (including Hall elements, touch sensors, magnetic sensors and airflow sensors) are arranged at the self-rescuer to determine whether the user's operating behavior is accurate.

[0088] The judgment logic of sequence accuracy includes:

[0089] Arrange Hall elements and touch sensors at the mouthpiece and nose clip, and continuously monitor the status of the Hall elements and touch sensors;

[0090] Configure the initial baseline threshold. If the magnetic field value of the Hall element and the capacitance value of the touch sensor are less than the initial baseline threshold, the self-rescuer is determined to be in an unused state and is waiting for user training.

[0091] When the capacitance value of the touch sensor at the mouthpiece changes, the step of determining whether the mouthpiece is worn is initiated;

[0092] Configure the wearing time (e.g. 5 seconds) and the capacitance threshold. If the capacitance value of the touch sensor at the mouthpiece is not detected to be greater than or equal to the capacitance threshold within the wearing time, the mouthpiece wearing timeout is determined, and the user is prompted by voice to speed up the training operation.

[0093] If the capacitance value of the touch sensor at the mouthpiece is detected to be less than the capacitance threshold, it is determined that the mouthpiece is not worn correctly, and the user is prompted by voice to adjust the posture and put on the mouthpiece again;

[0094] After the mouthpiece is judged to be correctly worn, the Hall element and touch sensing at the nose clip continue to be monitored;

[0095] Configure the magnetic field threshold. If the magnetic field value of the Hall element at the nose clip is greater than or equal to the magnetic field threshold, it means that the nose clip has been opened and the wearing confirmation of the nose clip is waiting.

[0096] If the capacitance value of the touch sensor at the nose clip is greater than or equal to the capacitance threshold and the magnetic field value of the Hall element at the nose clip is greater than or equal to the magnetic field threshold, it is determined that the nose clip is worn correctly;

[0097] If the capacitance value of the touch sensor at the nose clip is less than the capacitance threshold or the Hall element at the nose clip is less than the magnetic field threshold, it is determined that the nose clip is worn incorrectly, and the user is prompted by voice to wear the nose clip again;

[0098] Monitor the behavior data of each step when the user performs the self-rescuer training and compare it with the standard operation sequence, and record whether the behavior data of each step of the user conforms to the standard operation sequence and the operation time of each step to judge the sequence accuracy of the user.

[0099] The standard operation sequence of the self-rescuer includes opening the upper cover of the self-rescuer, unfolding the airbag and opening the mouthpiece plug, wearing the mouthpiece, opening the air valve, pressing the air replenishment plate, wearing the nose clip to clamp the nose, and breathing smoothly.

[0100] It should be understood that Hall elements are usually used to detect magnetic field changes. A common application is to judge whether a magnet is approaching. By arranging Hall elements and touch sensor elements at the mouthpiece and nose clip, it is possible to sense whether the mouthpiece plug is correctly installed; whether the mouthpiece is completely covered by the lips and whether the nose is clamped. This is based on the change in the magnetic field response of the Hall element and the capacitance induction change of the touch when the contact position between the mouthpiece and the lips and the contact position between the nose and the nose clip change; the use of the above electronic components can identify whether the user correctly wears the mouthpiece and clamps the nose clip according to the voice prompt of the self-rescuer, and then judge the completion of each step of the self-rescuer operation.

[0101] Hall elements and touch sensor elements are arranged at the mouthpiece and nose clip of the self-rescuer. In practice, there may be two or more touch sensor elements at the mouthpiece to avoid misjudgment of the self-rescuer caused by accidental finger touch, resulting in triggering the voice prompt of the next step. When these two or more touch sensor elements both detect that the human touch is greater than or equal to a predetermined threshold (indicating that an object is approaching or in contact), it is considered that the mouthpiece has been completely covered by the lips; the nose clip is equipped with a Hall element and a human touch sensor element. When the nose contacts the nose clip, the Hall element will sense a change in the magnetic field. If the magnetic field intensity is greater than or equal to a predetermined threshold, it means that the nose clip is opened. When the nose clip is correctly worn on the nose, the human touch sensor element will generate a capacitance change greater than or equal to the predetermined threshold, thus identifying the correct wearing of the user; if the magnetic field value detected by the electronic components of the mouthpiece is less than the predetermined threshold, or the relevant electronic components of the nose clip do not sense a magnetic field or capacitance change, it is considered that the user's posture is inaccurate and the user needs to be prompted by voice to adjust. On the contrary, if the magnetic field values detected by all electronic components are greater than or equal to the predetermined threshold, then it is considered that the user's posture is accurate and the next step can be triggered.

[0102] As the user sequentially completes the training steps of the self-rescuer, for example, magnetic induction is set at the upper cover of the self-rescuer, and when the upper cover of the self-rescuer is opened, the magnetic induction will change accordingly. Airflow induction is set at the air valve, and when the air valve is opened, the airflow induction will change accordingly. According to the behavior data of each step compared with the standard operation sequence, when the user skips a certain step, repeats an operation of a certain step, or reverses the operation sequence, it is determined that the sequence is incorrect, and the operation time of each step is recorded to determine the sequence accuracy of the user.

[0103] The operation accuracy of the user is comprehensively judged based on the user's posture accuracy and sequence accuracy.

[0104] According to the video data, it is judged whether the user is the person himself. After the user is the person himself, the user's posture is analyzed to obtain the user's posture accuracy; according to the behavior data, it is judged whether the training steps of the user conform to the standard operation sequence and whether each training step conforms to the specified operation time; when the user's posture is accurate, conforms to the standard operation sequence, and each training step conforms to the specified operation time, it indicates that the user's operation is accurate; when the user's posture is inaccurate, does not conform to the standard operation sequence, or each training step does not conform to the specified operation time, it indicates that the user's operation is inaccurate.

[0105] Specifically, the processing result is sent to the result judgment node through the wireless ZIGBEE gateway and WIFI module to identify the user's behavior result, and at the same time, it is summarized and saved locally and forwarded to the cloud.

[0106] The result judgment node is used to judge the user's behavior result according to the processing result.

[0107] The user's behavior result is judged according to the user's operation accuracy. When the user's operation is accurate, it is determined that the user's behavior is qualified; when the user's operation is inaccurate, it is determined that the user's behavior is unqualified. At the same time, the behavior results of all users are counted and arranged in ascending order of operation time.

[0108] The feedback control node is used to generate training feedback according to the user's behavior result and judge the state of the self-rescuer.

[0109] The states of the self-rescuer include normal, abnormal, and to be reset.

[0110] The logic for judging the state of the self-rescuer is as Figure 3 shown, specifically including:

[0111] After generating the training feedback, the state of the self-rescuer is initially checked by monitoring the working parameters of the self-rescuer. If the state of the self-rescuer is normal, it indicates that the self-rescuer has been successfully reset;

[0112] If the self-rescuer status is abnormal or awaiting reset, perform the reset operation. After performing the reset operation, check the self-rescuer status. If the reset is unsuccessful, trigger a prompt.

[0113] Training feedback includes a summary of operation results, error analysis, and improvement suggestions. The summary of operation results is for each user, summarizing each user's behavior (qualified or unqualified) and operation time (here, the standard operation time for performing the entire standard operation sequence is 30 seconds, and if it exceeds 30 seconds, it is unqualified). For example, User 1's operation is qualified, with an operation time of 25 seconds, User 2's operation is unqualified, with an operation time of 35 seconds, and User 3's operation is unqualified, with an operation time of 26 seconds.

[0114] Error analysis is for users with unqualified operations, analyzing the errors that occurred in their training steps. For example, "User 2 exceeded the specified time by 5 seconds during the mouthpiece wearing step of the training" and "User 3 did not operate according to the standard operation sequence during the training steps and skipped the step of opening the air valve."

[0115] Improvement suggestions refer to providing improvement suggestions for users based on the results of error analysis. For example, "User 2, please speed up the operation of wearing the mouthpiece to ensure that the training steps are completed according to the standard operation time" and "User 3, please re-learn the standard operation sequence and perform the self-rescuer training according to the standard operation sequence."

[0116] Initial check of the self-rescuer status by monitoring the working parameters of the self-rescuer (including the battery power of the self-rescuer, whether the upper cover of the self-rescuer is closed, whether each electronic component is reset, the accuracy of the user's posture, the accuracy of the user's sequence, and whether each electronic component is working properly). When all working parameters are within the preset normal range, it indicates that the self-rescuer status is normal, meaning the self-rescuer has been successfully reset and is in a working state; when some working parameters have problems (such as low battery power, network connection failure, inaccurate user posture, inaccurate user operation, or each electronic component of the self-rescuer not reaching the preset threshold, etc.), it indicates that the self-rescuer is in an abnormal state. At this time, a reset operation needs to be performed according to the working parameters with problems; when the working parameters of the self-rescuer are within a certain range (such as slightly low battery power, inaccurate user posture, and the sensor network is not completely normal, etc.), it is determined to be in a state awaiting reset, indicating that a reset operation is required, but it is not completely unable to continue working.

[0117] The reset operation includes, but is not limited to, restarting the training network (restarting the management system, clearing the error status), calibrating the sensor network (such as re-detecting the distance of the Hall element, the capacitance value of the touch sensor, and confirming the status of the upper cover of the self-rescuer), and re-verifying the wearing status (confirming whether the mouthpiece and nose clip are correct).

[0118] Specifically, if the reset operation is successful, send a message to the user such as "A certain component has been reset" or "The device has returned to normal", prompting the user that the self-rescuer is ready for continued use; if the reset operation fails, prompt the user that there are more serious problems (such as low battery power or hardware failure), and recommend further inspection or repair of the self-rescuer.

[0119] Among them, after the steps according to the voice prompt are completed, judge whether the self-rescuer is reset, whether the battery is sufficient, and whether the network connection status is normal, and perform manual reset, charging, and reconnection to the network as needed to ensure the reset, sufficient battery power, and network connection of the self-rescuer, and enable communication with the cloud.

[0120] Specifically, detect the user's posture accuracy and sequence accuracy through electronic components such as Hall elements, touch sensing, magnetic induction, and airflow induction to obtain the user's operation accuracy; accurately identify the above training steps, judge the user's operation accuracy and sort them in ascending order of operation time to obtain the user's learning performance.

[0121] The storage management node is used to store the user's behavior data, video data, and the user's behavior results in local and cloud for centralized management.

[0122] The logic of storing in local and cloud for centralized management includes:

[0123] Give a storage structure table according to the user's behavior data, video data, and the user's result pattern, assign a unique user ID to each user, calculate the storage requirements according to the storage structure table, complete the user's identity verification according to the storage structure table, and visually display the user's behavior data, video data, and the user's behavior results.

[0124] The calculation formula for storage requirements is as follows:

[0125] r=n×(b p +b o +b c );

[0126] In the formula, r represents the storage requirement, n represents the number of users, b o represents the average size of each user's operation data, b p represents the average size of each user's video data, b c represents the average size of each user's behavior pattern.

[0127] It should be explained that: the average size b o of each user's operation data is obtained by statistically analyzing the acquisition frequency of operation data and the amount of data acquired each time; the average size b pIt is obtained by statistically analyzing the acquisition frequency of video data and the amount of data acquired each time; the average size b of each user behavior pattern c is determined by analyzing the complexity and diversity of user behavior patterns.

[0128] It should be understood that the storage structure table includes a user data table, an operation data table, a video data table, and a behavior pattern table. The user data table includes a user ID, the last login time, and authentication information. The operation data table includes an operation ID, a user ID, an operation type, and an operation time. The video data table includes a video ID, a user ID, a distance value, a blinking frequency, and a recording time. The behavior pattern table includes a pattern ID, a user ID, a behavior pattern, and a pattern recognition time. The above user ID is a unique identification code assigned to each user, and all storage structure tables are connected through the user ID.

[0129] Specifically, the control system can satisfy the learning and training of multiple people at the same time, transmit data through a transmission protocol, ensure no packet loss during concurrent training of multiple people, assign a unique identification code to each user for easy data tracking and analysis, and ensure the encryption of operation data and video data during transmission to prevent operation data and video data from being stolen. The operation data and video data of users should comply with the principle of privacy protection when stored to ensure user information security. When a user logs in to use the control system, identity authentication is required, and the result of the identity authentication is compared with the content in the storage structure table. If the identity authentication is successful, the login is successful. At the same time, visual display includes the operation data and video data of the user and the behavior pattern of the user.

[0130] Embodiment 2

[0131] As Figure 4 shown, the method flowchart of the intelligent training data processing method for self-rescuers with multi-user synchronous training provided by the embodiment of the present application is as follows. The method includes:

[0132] S1. Obtain the operation data and video data of the user through the sensor network, and process the operation data and video data of the user to generate a processing result;

[0133] S2. Judge the behavior pattern of the user according to the processing result;

[0134] S3. Generate training feedback according to the behavior pattern of the user and judge the state of the self-rescuer;

[0135] S4. Store the operation data and video data of the user and the behavior pattern of the user in the cloud for centralized management.

[0136] Since the principle of solving problems by the method in the embodiment of the present application is similar to that of the system in the above embodiment of the present application, the implementation of the method refers to the implementation of the system, and the repeated parts will not be described again.

Claims

1. An intelligent training network for self-rescuers with multi-user synchronous training, characterized in that: include: Data acquisition node, result judgment node, feedback control node and storage management node; The data acquisition node is used to acquire the user's behavior data and video data, and process the user's behavior data and video data to generate a processing result; The result judgment node is used to judge the user's behavior result according to the processing result; The feedback control node is used to generate training feedback according to the user's behavior results and determine the state of the self-rescuer; The storage management node is used to store the user's behavior data and video data as well as the user's behavior results locally and in the cloud for centralized management.

2. The self-rescuer intelligent training network for multi-user synchronous training as claimed in claim 1, characterized in that: The generation strategy of the processing result includes: Deploy cameras to obtain user video data, determine whether the current operating user is the target user, and analyze the user's posture to determine the accuracy of the user's posture; Acquire user behavior data through Hall elements and touch sensing, analyze the user behavior data, compare the user behavior data with the standard operation sequence, and verify the accuracy of the user's sequence; The user's operation accuracy is comprehensively judged based on the user's posture accuracy and sequence accuracy.

3. The self-rescuer intelligent training network for multi-user synchronous training as claimed in claim 2, characterized in that: The judgment logic of the posture accuracy includes: By comparing the user posture with the standard posture, the similarity between the user posture and the standard posture is determined; Configure a similarity threshold. If the similarity between the user's posture and the standard posture is less than or equal to the similarity threshold, the user's posture is inaccurate. If the similarity between the user's posture and the standard posture is greater than the similarity threshold, the user's posture is accurate.

4. The self-rescuer intelligent training network for multi-user synchronous training as claimed in claim 3, characterized in that: The judgment logic of the similarity between the user posture and the standard posture includes: Identify the target node of the user posture and the target node of the standard posture; Calculate the Euclidean distance between the target node of the user posture and the target node of the corresponding standard posture; The Euclidean distance between the target node of each user posture and the target node of the corresponding standard posture is multiplied by the weight of the corresponding target node and then summed to obtain a comprehensive distance value; Subtract the comprehensive distance value from 1 to get the similarity between the user's posture and the standard posture.

5. The self-rescuer intelligent training network for multi-user synchronous training as claimed in claim 4, characterized in that: The judgment logic of the sequence accuracy includes: Arrange Hall elements and touch sensors at the mouthpiece and nose clip, and continuously monitor the status of the Hall elements and touch sensors; Configure the initial baseline threshold. If the magnetic field value of the Hall element and the capacitance value of the touch sensor are less than the initial baseline threshold, the self-rescuer is determined to be in an unused state and is waiting for user training. When the capacitance value of the touch sensor at the mouthpiece changes, the step of determining whether the mouthpiece is worn is initiated; Configure the wearing time and capacitance threshold. If the capacitance value of the touch sensor at the mouthpiece is not greater than or equal to the capacitance threshold within the wearing time, the mouthpiece wearing timeout is determined, and the user is prompted by voice to speed up the training operation. If the capacitance value of the touch sensor at the mouthpiece is detected to be less than the capacitance threshold, it is determined that the mouthpiece is not worn correctly, and the user is prompted by voice to adjust the posture and put on the mouthpiece again.

6. The self-rescuer intelligent training network for multi-user synchronous training as claimed in claim 5, characterized in that: The judgment logic of the sequence accuracy also includes: After the mouthpiece is judged to be correctly worn, the Hall element and touch sensing at the nose clip continue to be monitored; Configure the magnetic field threshold. If the magnetic field value of the Hall element at the nose clip is greater than or equal to the magnetic field threshold, it means that the nose clip has been opened and the wearing confirmation of the nose clip is waiting. If the capacitance value of the touch sensor at the nose clip is greater than or equal to the capacitance threshold and the magnetic field value of the Hall element at the nose clip is greater than or equal to the magnetic field threshold, it is determined that the nose clip is worn correctly; If the capacitance value of the touch sensor at the nose clip is less than the capacitance threshold or the Hall element at the nose clip is less than the magnetic field threshold, it is determined that the nose clip is worn incorrectly, and a voice prompt is given to the user to put the nose clip back on; Monitor the user's behavioral data of each step when performing self-rescuer training and compare it with the standard operation sequence, record whether the user's behavioral data of each step conforms to the standard operation sequence and the operation time of each step, so as to determine the user's sequence accuracy.

7. The self-rescuer intelligent training network for multi-user synchronous training as claimed in claim 6, characterized in that: The self-rescuer status includes normal, abnormal and waiting to be reset; The logic for judging the state of the self-rescuer includes: After the training feedback is generated, the state of the self-rescuer is initially checked by monitoring the working parameters of the self-rescuer. If the state of the self-rescuer is normal, it means that the self-rescuer is reset successfully; If the state of the self-rescuer is abnormal or waiting to be reset, the reset operation is performed. After the reset operation is performed, the state of the self-rescuer is judged. If the reset is unsuccessful, a prompt is triggered.

8. The self-rescuer intelligent training network for multi-user synchronous training as claimed in claim 7, characterized in that: The logic of local storage and cloud centralized management includes: A storage structure table is given based on the user's behavior data, video data, and the user's result pattern. A unique user ID is assigned to each user, and the storage requirements are calculated based on the storage structure table. The user's identity authentication is completed based on the storage structure table, and the user's behavior data, video data, and the user's behavior results are visualized.

9. A method for processing data of a self-rescuer with multi-user synchronous training, based on a self-rescuer with multi-user synchronous training with a smart training network according to any one of claims 1 to 8, characterized in that: include: S1. Obtaining user behavior data and video data, and processing the user behavior data and video data to generate processing results; S2. Determine the user's behavior results based on the processing results; S3, generating training feedback according to the user's behavior results and judging the state of the self-rescuer; S4. Store the user's behavior data, video data, and user's behavior results locally and in the cloud for centralized management.

Images