Human physiology and motion state anomaly detection method and device

Abstract

The invention is suitable for the technical field of machine learning, and provides a human physiology and motion state anomaly detection method and device. The human physiology and motion state anomaly detection method comprises the steps of obtaining real-time sign data collected by the sign collector and standard sign data which are obtained through pre-training and respectively correspond to multiple preset states; performing semantic classification on the real-time physical sign data to obtain real-time semantic information; each piece of real-time semantic information corresponds to one or more preset states; and according to standard physical sign data corresponding to the real-time semantic information, determining whether the real-time physical sign data is in an abnormal state or not. According to the technical scheme, by introducing the anomaly detection method based on semantic classification, the accuracy of physical sign data analysis in the wearable device is remarkably improved, and the real-time performance and the intelligent level of the device are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning, and in particular relates to a method and device for detecting abnormal physiological and motor states of a human body. Background Art

[0002] Wearable devices, as mechanical devices that can sense and enhance human movement, have been widely used in the fields of medical rehabilitation, assisted walking, and physical strength enhancement in recent years. Especially in the field of medical rehabilitation, wearable devices can monitor the state of the human body and help patients with limited mobility to carry out daily activities, thereby improving their quality of life.

[0003] However, in the process of using wearable devices, how to effectively detect and deal with abnormal conditions that may occur during human movement has become an important issue that needs to be solved urgently. Relying on traditional vital sign data analysis methods, it is often difficult to accurately identify a small number of abnormal conditions, such as (backward) falls, (sudden) turns, and (accidental) falls. These abnormal conditions not only involve changes in movement patterns, but are also accompanied by comprehensive changes in multiple individual physiological and movement parameters, resulting in low accuracy of traditional abnormal vital sign data analysis methods. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a method and device for detecting abnormal physiological and motion states of a human body, so as to solve the technical problem of low accuracy of traditional abnormal vital sign data analysis methods.

[0005] A first aspect of an embodiment of the present invention provides a method for detecting abnormal physiological and motion states of a human body, wherein the method is applied to a processor in a wearable device, wherein the wearable device includes a vital sign collector and a processor, and the method includes:

[0006] Acquire the real-time vital sign data collected by the vital sign collector and the standard vital sign data corresponding to the various preset states obtained by pre-training;

[0007] Performing semantic classification on the real-time vital sign data to obtain real-time semantic information; each piece of the real-time semantic information corresponds to one or more of the preset states;

[0008] According to the standard vital sign data corresponding to the real-time semantic information, it is determined whether the real-time vital sign data is in an abnormal state.

[0009] Furthermore, the step of determining whether the real-time vital sign data is in an abnormal state according to the standard vital sign data corresponding to the real-time semantic information includes:

[0010] Acquiring standard vital sign data corresponding to the real-time semantic information;

[0011] Calculating the distance between the real-time vital sign data and the standard vital sign data;

[0012] If the distance is greater than a threshold, it is determined that the real-time vital sign data is in an abnormal state.

[0013] Furthermore, before the step of acquiring the real-time vital sign data collected by the vital sign collector and the standard vital sign data corresponding to the various preset states obtained by pre-training, the method further includes:

[0014] Acquire first training physical sign data collected by the physical sign collector; the first training physical sign data is physical sign data of a target user in a normal state collected by the physical sign collector;

[0015] Acquire first historical vital sign data corresponding to each of a plurality of preset states; the first historical vital sign data is obtained by collecting vital sign data of other users in the preset states by a vital sign collector;

[0016] The first training vital sign data and the first historical vital sign data are input into a target generation model to obtain the standard vital sign data output by the generation model.

[0017] Furthermore, the step of inputting the first training vital sign data and the first historical vital sign data into the target generation model to obtain the standard vital sign data output by the generation model comprises:

[0018] Performing noise processing on the first historical vital sign data through a noise adding network in the target generation model to obtain first feature data;

[0019] Matching first training vital sign data corresponding to the real-time semantic information;

[0020] fusing the first feature data with first training vital sign data corresponding to the real-time semantic information to obtain second feature data;

[0021] The second feature data is denoised by a denoising network in the target generation model to obtain standard vital sign data.

[0022] Furthermore, before the step of inputting the first training vital sign data and the first historical vital sign data into the target generation model to obtain the standard vital sign data output by the target generation model, the step further includes:

[0023] Acquire second training vital sign data collected by the vital sign collector;

[0024] Acquire second historical vital sign data corresponding to each of a plurality of preset states;

[0025] Obtain the initial generation model, initial semantic information and label data;

[0026] The initial generation model is trained according to the second historical vital sign data, the initial semantic information and the label data to obtain the target generation model.

[0027] Furthermore, the step of training the initial generation model according to the second historical vital sign data, the initial semantic information and the label data to obtain the target generation model includes:

[0028] Performing noise processing on the second historical vital sign data through the noise adding network in the initial generation model to obtain third feature data;

[0029] Acquire second training vital sign data corresponding to the initial semantic information;

[0030] fusing the third feature data with the second training vital sign data corresponding to the initial semantic information to obtain fourth feature data;

[0031] Performing denoising processing on the fourth feature data through a denoising network in the initial generation model to obtain output data;

[0032] Calculating an error between the output data and the label data, and adjusting the initial generation model based on the error;

[0033] Based on different first historical vital sign data, the step of performing noise processing on the second historical vital sign data through the noise adding network in the initial generation model to obtain first feature data and subsequent steps are iteratively performed to obtain the target generation model.

[0034] Furthermore, the step of performing semantic classification on the real-time vital sign data to obtain semantic information includes:

[0035] The real-time vital sign data is input into a supervision model to obtain the semantic information output by the supervision model.

[0036] A second aspect of an embodiment of the present invention provides a device for detecting abnormal physiological and motion states of a human body, comprising:

[0037] An acquisition unit, used to acquire the real-time vital sign data collected by the vital sign collector and the standard vital sign data corresponding to each of the multiple preset states obtained by pre-training;

[0038] A classification unit, used for semantically classifying the real-time vital sign data to obtain real-time semantic information; each piece of the real-time semantic information corresponds to one or more of the preset states;

[0039] The judgment unit is used to determine whether the real-time vital sign data is in an abnormal state according to the standard vital sign data corresponding to the real-time semantic information.

[0040] The third aspect of an embodiment of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for detecting abnormal physiological and motor states of the human body described in the first aspect are implemented.

[0041] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the method for detecting abnormal physiological and motion states of the human body described in the first aspect are implemented.

[0042] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: by using the standard vital sign data corresponding to the various preset states obtained by pre-training, the method can accurately identify and classify various abnormal motion states, such as falling, turning and falling. This detection method based on semantic classification is more targeted than the traditional method, reducing the possibility of misjudgment and missed judgment. By semantically classifying the real-time vital sign data collected by the vital sign collector, real-time semantic information of the preset state can be quickly generated, so as to instantly judge whether the vital sign data is abnormal. This real-time detection capability is of great significance for the safety and monitoring of the user's motion state, especially in some high-risk sports or rehabilitation training. The method uses the pre-trained model to accurately identify complex motion states through semantic classification, reflecting a high level of intelligent processing capabilities. The processor can autonomously analyze and judge the motion state. The method is applicable to various types of wearable devices and can meet the needs of different users in various sports and daily activity scenarios. Whether it is used for rehabilitation training, assisted walking or enhanced sports ability, reliable abnormality detection support can be provided. Through high-precision and real-time abnormality detection, the user's motion state can be fed back in time to prevent the occurrence of potential dangers, thereby enhancing the user's sense of security and user experience. At the same time, it reduces false alarms and improves users' trust and dependence on the device. In summary, this technical solution significantly improves the accuracy of vital sign data analysis in wearable devices by introducing an anomaly detection method based on semantic classification, and enhances the real-time and intelligent level of the device. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 A schematic flow chart of a method for detecting abnormal physiological and motion states of a human body provided by the present invention is shown;

[0045] Figure 2 A schematic flow chart of another method for detecting abnormal physiological and motion states of a human body provided by the present invention is shown;

[0046] Figure 3 A schematic flow chart of another method for detecting abnormal physiological and motion states of a human body provided by the present invention is shown;

[0047] Figure 4 A schematic diagram of a device for detecting abnormal physiological and motion states of a human body provided by an embodiment of the present invention is shown;

[0048] Figure 5 A schematic diagram of a terminal device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0049] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0050] The embodiment of the present invention provides a method and device for detecting abnormal physiological and motion states of a human body, so as to solve the technical problem of low accuracy of traditional abnormal vital sign data analysis methods.

[0051] First, the present invention provides a method for detecting abnormal physiological and motion states of a human body. Figure 1 , Figure 1 FIG. 1 is a schematic flow chart of a method for detecting abnormal physiological and motor states of a human body provided by the present invention. Figure 1 As shown, the method for detecting abnormal physiological and motion states of a human body may include the following steps: Step 101: acquiring real-time vital sign data collected by the vital sign collector and standard vital sign data corresponding to a plurality of preset states obtained by pre-training;

[0052] The method for detecting abnormal physiological and motion states of the human body is applied to the processor in a wearable device, and the wearable device includes a vital sign collector and a processor. The vital sign collector is used to collect physiological data and vital sign data of the human body. The application scenarios of the vital sign collector include but are not limited to motion data collection in exoskeleton instruments and physiological (snoring and apnea, etc.) and motion detection (leg movement and sleepwalking, etc.) in sleep monitoring. That is, the application scenarios of the present application include but are not limited to the field of abnormal state detection of exoskeletons and the field of abnormal sleep monitoring, etc. There is no limitation on the application scenarios here, and it can be any application scenario applicable to the technical solution of the present application.

[0053] Vital sign collectors include but are not limited to motion sensors and physiological monitors. Vital sign collectors include but are not limited to inertial sensors, force sensors, and optical sensors. Physiological monitors include but are not limited to EEG monitors, electrooculogram monitors, electromyography monitors, blood oxygen monitors, blood pressure monitors, electrocardiogram monitors, respiratory monitors, and nasal airflow monitors.

[0054] Vital sign data include real-time vital sign data and standard vital sign data. Vital sign data include but are not limited to physiological data, dynamic data and kinematic data. Kinematic data include but are not limited to one or a combination of multiple data such as position, velocity, acceleration and angular velocity. Dynamic data include but are not limited to one or a combination of multiple data such as force, moment, center of mass position and torque. Physiological data include but are not limited to one or a combination of multiple data such as physiological electrical signals, respiration, heart rate, blood oxygen, blood pressure, airflow pressure and airflow temperature. The parts of the human body monitored by kinematics include but are not limited to toes, heels, calves, knees, thighs, palms, forearms, elbows, upper arms, shoulders and hips.

[0055] For example, when the application scenario is in the field of exoskeleton abnormal state detection, the vital sign data is kinematic data and / or dynamic data. When the application scenario is in the field of sleep abnormality monitoring, the vital sign data is kinematic data and physiological data. The vital sign data corresponding to different application scenarios can be selected according to actual needs, and no limitation is made here.

[0056] Wearable devices include but are not limited to exoskeleton devices, physiological monitoring devices, mechanical monitoring devices, and combinations of multiple devices, which can be based on actual application scenarios and are not limited here.

[0057] Different preset states correspond to different standard vital signs data. Standard vital signs data include but are not limited to standard motion data, standard physiological data and standard mechanical data. Standard motion data include but are not limited to normal walking data, turning data, jumping data, squatting data and going up and down stairs data. Standard physiological data include normal breathing data and normal heart rate data. Standard mechanical data include normal torque data and normal moment data.

[0058] Among them, taking standard motion data as an example, for different users, normal walking data is easier to obtain (users can be guided to walk normally for the first time to obtain corresponding walking data), while turning data, jumping data, squatting data and up and down stairs data are more difficult to obtain, so this application generates personalized data (standard vital signs data) for different users through data synthesis technology. For example, the normal walking data of the current user is combined with the jumping data of other users to obtain the jumping data of the current user. Personalized data regards each sample as an individual with independent characteristics, aiming to synthesize data across semantics, such as using a large amount of normal walking data of users and other people's standard vital signs data with target semantics (target semantics are not limited to turning, jumping and squatting, etc.) to synthesize the standard vital signs data of the user. The advantage of doing this is that it can combine the normal vital signs data of individuals and a large number of standard vital signs data of other individuals with target semantics, without the need to collect all the standard vital signs data of individuals. This is different from the traditional prediction model, which is often committed to generating data with the same semantics within a short time window. For example, for a patient with hip compensatory force, his walking posture may be considered abnormal under conventional standards, but it is normal for him personally. Therefore, it is necessary to use the walking data containing his unique walking posture to predict the standard vital sign data, which can effectively reduce the problems of poor quality and lack of continuity of the generated data.

[0059] The synthesis method of standard vital signs data is as follows:

[0060] Optionally, before step 101, steps A1 to A3 are also included:

[0061] Step A1: Acquire first training physical sign data collected by the physical sign collector; the first training physical sign data is physical sign data of a target user in a normal state collected by the physical sign collector;

[0062] It is worth noting that the execution time node of step 101 is to collect real-time vital sign data for abnormal judgment during the daily use of the wearable device. Step A1 is used to initialize the wearable device to train the deep learning model in the wearable device to adapt to the physiological characteristics and walking habits of different users.

[0063] The first training physical sign data includes but is not limited to motion characteristic data, physiological physical sign data, mechanical data, and the like.

[0064] Taking the wearable device as an exoskeleton as an example, in the initialization stage of the wearable device, the user needs to wear the wearable device and walk normally according to his or her own walking habits. At this time, the first training vital sign data is collected by the vital sign collector.

[0065] Step A2: acquiring first historical vital sign data corresponding to each of a plurality of preset states; the first historical vital sign data is obtained by collecting vital sign data of other users in the preset states by a vital sign collector;

[0066] The first historical vital sign data is historical standard vital sign data of other users, or may be artificially set standard vital sign data, which is used to fuse and generate the standard vital sign data of the current user.

[0067] Step A3: input the first training vital sign data and the first historical vital sign data into the target generation model to obtain the standard vital sign data output by the target generation model.

[0068] The first training vital sign data and the first historical vital sign data are fused through the target generation model. The specific logic is as follows:

[0069] Specifically, step A3 includes steps A31 to A34:

[0070] Step A31: performing noise processing on the first historical vital sign data through the noise adding network in the target generation model to obtain first feature data;

[0071] Noise is gradually added to the first historical vital sign data through a noise adding network to obtain first characteristic data, making it gradually become random.

[0072] Step A32: matching the first training vital sign data corresponding to the real-time semantic information;

[0073] Step A33: fusing the first feature data with the first training vital sign data corresponding to the real-time semantic information to obtain second feature data;

[0074] Step A34: De-noising the second feature data through the denoising network in the target generation model to obtain standard vital sign data.

[0075] The denoising network gradually removes noise from the second feature data to obtain standard vital sign data.

[0076] The first training vital sign data of the target user during normal walking is collected by the vital sign collector, and the first historical vital sign data corresponding to each of the multiple preset states are obtained, and these historical data are composed of the vital sign data of other users in the preset state. By collecting the vital sign data of different users in the preset state, a diversified standard vital sign data set can be established. The first historical vital sign data is processed by the noise adding network in the target generation model to generate the first feature data. Through the noise adding process, the uncertainty and noise characteristics in the data can be effectively simulated, so that the model can better cope with various complex situations in practical applications. The first training vital sign data corresponding to the real-time semantic information is matched and fused with the first feature data after noise addition to obtain the second feature data. This process can combine the normal vital sign data of the user with the standard vital sign data of others by combining the characteristics of the normal vital sign data and the standard vital sign data, and then generate the user's personalized standard vital sign data. The second feature data is denoised by the denoising network in the target generation model to obtain the final standard vital sign data. The denoising process can filter out the noise and interference in the data, making the extracted standard vital sign data more accurate and reliable. In summary, this technical solution can effectively improve the accuracy and robustness of abnormal motion state detection by integrating normal walking training data with historical data of preset states and processing with the help of noise addition and denoising networks. This method can adapt to individual differences of different users.

[0077] The target generation model is able to generate samples with high diversity through a gradual denoising process. Due to the randomness of its generation process, the target generation model can generate different standard vital sign data with the same latent variable as input. The target generation model supports guided diffusion, which means that the generation process can be guided by adding additional conditional information (such as semantic information) to improve the quality and consistency of the generated quantity. Compared with generative adversarial networks (GANs), the training process of the target generation model is more stable. The training of GANs often requires complex balancing techniques to counter the generator and the discriminator, while the target generation model is simpler and more stable to train through a clear loss function and gradual denoising, reducing the difficulty and instability of training. The training logic of the target generation model is as follows:

[0078] Optionally, before step A34, steps B1 to B4 are also included:

[0079] Step B1: acquiring second training vital sign data collected by the vital sign collector;

[0080] It is worth noting that the execution time node of step 101 is to collect real-time vital sign data for abnormal judgment during the daily use of the wearable device. Step B1 is used to train the target generation model. The second training vital sign data can be the training vital sign data of any other user, when the user wears the wearable device and walks normally according to his own walking habits, and the second training vital sign data collected by the vital sign collector.

[0081] Step B2: Acquire second historical vital sign data corresponding to each of a plurality of preset states;

[0082] The second historical vital sign data may be historical vital sign data of any other user.

[0083] Step B3: Obtain the initial generation model, initial semantic information and label data;

[0084] The initial generation model refers to the generation model in which the parameters are initialization parameters. The initial semantic information is a plurality of preset states. The label data is the labels corresponding to the various initial semantic information.

[0085] Step B4: training the initial generation model according to the second historical vital sign data, the initial semantic information and the label data to obtain the target generation model.

[0086] Specifically, step B4 includes steps B41 to B46:

[0087] Step B41: performing noise processing on the second historical vital sign data through the noise adding network in the initial generation model to obtain third feature data;

[0088] Step B42: Acquire second training vital sign data corresponding to the initial semantic information;

[0089] Step B43: fusing the third feature data with the second training vital sign data corresponding to the initial semantic information to obtain fourth feature data;

[0090] Step B44: performing denoising processing on the fourth feature data through the denoising network in the initial generation model to obtain output data;

[0091] Step B45: calculating the error between the output data and the label data, and adjusting the initial generation model based on the error;

[0092] Step B46: Based on different first historical vital sign data, iteratively execute the step of performing noise processing on the second historical vital sign data through the noise adding network in the initial generation model to obtain the first feature data and subsequent steps to obtain the target generation model.

[0093] Among them, the target generation model can also be VAE, GAN, diffusion model and its variants, and the feature extraction network includes but is not limited to CNN, LSTM, U-net, Transformer and its variants.

[0094] As an optional embodiment of the present application, for the target generation model, during subsequent user use, daily vital sign data can be collected and used as the second training vital sign data in step B43 to perform model update training so that the target generation model better meets user usage needs.

[0095] In this embodiment, by acquiring the second historical vital sign data corresponding to multiple preset states and performing noise processing on them, the model can learn and adapt to different preset states during the training process, thereby improving the robustness and generalization ability of the model when facing abnormal situations in real application scenarios. By introducing the noise adding network and the denoising network in the initial generation model, the model can effectively remove noise and extract useful features when processing noisy data. This process improves the adaptability and accuracy of the model in complex environments. By fusing the second training vital sign data corresponding to the third feature data and the initial semantic information, the model can more accurately capture and reflect the motion characteristics collected by the vital sign collector, and improve the model's understanding and generation ability of vital sign data. During the training process, the error between the output data and the label data is calculated, and the initial generation model is adjusted based on the error, so that the model can continuously optimize and improve its own performance and gradually approach the ideal state of the target generation model. By iterative training based on different first historical vital sign data, the model can better utilize the information in the historical data and improve the model's understanding and processing ability of the diversity and complexity of vital sign data. In summary, the technical solution significantly improves the performance and effect of the target generation model in processing vital sign data through a series of innovative processing and optimization steps.

[0096] Step 102: semantically classify the real-time vital sign data to obtain real-time semantic information; each piece of the real-time semantic information corresponds to one or more of the preset states;

[0097] The real-time vital sign data is input into a supervision model to obtain semantic information output by the supervision model. The supervision model includes but is not limited to CNN, LSTM, GNN, Transformer and its variants.

[0098] Step 103: Determine whether the real-time vital sign data is in an abnormal state according to the standard vital sign data corresponding to the real-time semantic information.

[0099] This application performs semantic classification on real-time vital sign data to obtain real-time semantic information of preset states. Then, multiple corresponding standard vital sign data (data subsets) are matched according to the real-time semantic information. Different preset states correspond to different subsets. This classification of data subsets can significantly reduce the amount of calculation (no need to match all data). The data subsets are generated based on the training data through the target generation model, that is, the target generation model pre-classifies different preset states into multiple data subsets, and the data distribution of the data subsets has high consistency and stability. Among them, the semantic classification logic of real-time vital sign data is as follows:

[0100] Specifically, step 103 specifically includes steps 1031 to 1033:

[0101] Step 1031: Acquire standard vital sign data corresponding to the real-time semantic information;

[0102] The standard vital sign data corresponding to different preset states are obtained by pre-training the target generation model. Step 1032: Calculate the distance between the real-time vital sign data and the standard vital sign data;

[0103] Methods for calculating the distance between real-time vital sign data and standard vital sign data include but are not limited to Euclidean distance, cosine similarity, Manhattan distance, Minkowski distance, Mahalanobis distance, Chebyshev distance combined with threshold determination, and also include Weibull distribution, parameterized Sigmoid distribution, generalized Pareto distribution, Frechet distribution, Gumbel distribution combined with confidence determination.

[0104] Step 1033: If the distance is greater than a threshold, it is determined that the real-time vital sign data is in an abnormal state.

[0105] By calculating the distance between real-time vital sign data and pre-defined standard vital sign data, this solution can accurately determine whether there is an abnormal state. Using distance as the judgment standard ensures the objectivity of the judgment, reduces the probability of misjudgment and missed judgment, and improves the reliability of the detection results. Real-time acquisition of vital sign data and distance calculation can quickly determine whether there is an abnormality at the moment of movement. This real-time detection capability plays a key role in the monitoring of user safety and movement status, especially in high-risk sports or rehabilitation training, and can provide timely warnings and take corresponding measures. By setting appropriate thresholds, the detection sensitivity can be flexibly adjusted according to different application scenarios and user needs. For example, a lower threshold can be set in a high-risk environment to quickly respond to potential dangers; a higher threshold can be set in general sports to avoid too many false alarms, thereby optimizing the user experience. The solution uses real-time semantic information and standard vital sign data for calculation, achieving a high level of intelligent processing. The processor can autonomously analyze and judge the movement state, which not only improves the degree of automation of detection, but also reduces the dependence on manual monitoring, reflecting a high degree of intelligence. This technical solution is applicable to various types of wearable devices and other sports monitoring devices, and can meet the needs of different users in various sports and daily activity scenarios. Whether it is used for rehabilitation training, assisted walking or enhanced athletic ability, it can provide reliable abnormality detection support. Through high-precision and real-time abnormality detection, the user's movement status can be fed back in time to prevent the occurrence of potential dangers, thereby enhancing the user's sense of security and usage experience. At the same time, false alarms are reduced, and the user's trust and dependence on the device are improved. In summary, this technical solution significantly improves the accuracy and real-time performance of abnormal state detection by calculating the distance between real-time vital sign data and standard vital sign data, thereby enhancing the intelligence level and applicability of the equipment.

[0106] Abnormal conditions include but are not limited to abnormal movement, physiological abnormalities and mechanical abnormalities. Abnormal movement includes but is not limited to falling, sudden stop and falling. Physiological abnormalities include but are not limited to abnormal breathing and abnormal heart rate. Mechanical abnormalities include but are not limited to abnormal moment and abnormal torque.

[0107] The abnormal data corresponding to the abnormal state is usually too small to be trained independently as a category. It has the characteristics of rarity and lack of cases, which will have a negative impact on the operation of many systems. For example, unnatural and sudden motion data appear in the gait signal of the lower limb exoskeleton due to the human body state and external environment.

[0108] It should be noted that in actual application scenarios, for preset falls, preset emergency stops, preset drops and preset pathological conditions, if data collection and wearable device correction can be performed, this part that can be processed is also regarded as "normal", that is, as standard vital sign data. For example, taking falls as an example, the preset falls are often forward falls, and the probability of falling backwards is extremely small, resulting in the inability to collect data. Therefore, this part of the abnormalities is usually too small to be used as a class for independent training, and has characteristics such as rarity and no cases.

[0109] In this embodiment, by using the standard vital sign data corresponding to a plurality of preset states obtained by pre-training, the method can accurately identify and classify a plurality of abnormal motion states, such as falling, turning and falling. This detection method based on semantic classification is more targeted than the traditional method, and reduces the possibility of misjudgment and missed judgment. By semantically classifying the real-time vital sign data collected by the vital sign collector, the real-time semantic information of the preset state can be quickly generated, so as to instantly judge whether the vital sign data is abnormal. This real-time detection capability is of great significance for the safety and monitoring of the user's motion state, especially in some high-risk sports or rehabilitation training. The method uses a pre-trained model to accurately identify complex motion states through semantic classification, reflecting a high level of intelligent processing capabilities. The processor can autonomously analyze and judge the motion state. The method is applicable to various types of wearable devices and can meet the needs of different users in various sports and daily activity scenarios. Whether it is used for rehabilitation training, assisted walking or enhanced athletic ability, reliable abnormality detection support can be provided. Through high-precision and real-time abnormality detection, the user's motion state can be fed back in time to prevent the occurrence of potential dangers, thereby enhancing the user's sense of security and user experience. At the same time, it reduces false alarms and improves users' trust and dependence on the device. In summary, this technical solution significantly improves the accuracy of vital sign data analysis in wearable devices by introducing an anomaly detection method based on semantic classification, and enhances the real-time and intelligent level of the device.

[0110] like Figure 4 The present invention provides a device for detecting abnormal physiological and motor status of human body, see Figure 4 , Figure 4 A schematic diagram of a device for detecting abnormal physiological and motion states of a human body provided by the present invention is shown. Figure 4 A device for detecting abnormal physiological and motion status of a human body includes:

[0111] An acquisition unit 21 is used to acquire the real-time vital sign data collected by the vital sign collector and the standard vital sign data corresponding to the various preset states obtained by pre-training;

[0112] A classification unit 22, configured to perform semantic classification on the real-time vital sign data to obtain real-time semantic information; each of the real-time semantic information corresponds to one or more of the preset states;

[0113] The judgment unit 23 is used to determine whether the real-time vital sign data is in an abnormal state according to the standard vital sign data corresponding to the real-time semantic information.

[0114] The present invention provides a human physiological and motion state abnormality detection device. By using the standard vital sign data corresponding to multiple preset states obtained by pre-training, the method can accurately identify and classify multiple abnormal motion states, such as falling, turning and falling. This detection method based on semantic classification is more targeted than the traditional method and reduces the possibility of misjudgment and missed judgment. By semantically classifying the real-time vital sign data collected by the vital sign collector, real-time semantic information of the preset state can be quickly generated, so as to instantly judge whether the vital sign data is abnormal. This real-time detection capability is of great significance for the user's safety and motion state monitoring, especially in some high-risk sports or rehabilitation training. The method uses a pre-trained model to accurately identify complex motion states through semantic classification, reflecting a high level of intelligent processing capabilities. The processor can autonomously analyze and judge the motion state. The method is applicable to various types of wearable devices and can meet the needs of different users in various sports and daily activity scenarios. Whether it is used for rehabilitation training, assisted walking or enhanced sports ability, reliable abnormality detection support can be provided. Through high-precision and real-time abnormality detection, the user's motion state can be fed back in time to prevent the occurrence of potential dangers, thereby enhancing the user's sense of security and user experience. At the same time, it reduces false alarms and improves users' trust and dependence on the device. In summary, this technical solution significantly improves the accuracy of vital sign data analysis in wearable devices by introducing an anomaly detection method based on semantic classification, and enhances the real-time and intelligent level of the device.

[0115] Figure 5 Schematic diagram of a terminal device provided by an embodiment of the present invention. Figure 5 As shown, a terminal device 3 of this embodiment includes: a processor 30, a memory 31, and a computer program 32 stored in the memory 31 and executable on the processor 30, such as a human physiological and motion state abnormality detection program. When the processor 30 executes the computer program 32, the steps of each of the above-mentioned human physiological and motion state abnormality detection method embodiments are implemented, such as Figure 1 Alternatively, when the processor 30 executes the computer program 32, the functions of each unit in the above-mentioned device embodiments are realized, for example, Figure 2 Function of the unit shown.

[0116] Exemplarily, the computer program 32 may be divided into one or more units, which are stored in the memory 31 and executed by the processor 30 to complete the present invention. The one or more units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 32 in the terminal device 3. For example, the computer program 32 may be divided into the following specific functions of each unit:

[0117] An acquisition unit, used to acquire the real-time vital sign data collected by the vital sign collector and the standard vital sign data corresponding to each of the multiple preset states obtained by pre-training;

[0118] A classification unit, used for semantically classifying the real-time vital sign data to obtain real-time semantic information; each piece of the real-time semantic information corresponds to one or more of the preset states;

[0119] The judgment unit is used to determine whether the real-time vital sign data is in an abnormal state according to the standard vital sign data corresponding to the real-time semantic information.

[0120] The terminal device includes but is not limited to a processor 30 and a memory 31. Those skilled in the art will appreciate that Figure 3 It is only an example of a terminal device 3 and does not constitute a limitation on the terminal device 3. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device may also include input and output devices, network access devices, buses, etc.

[0121] The processor 30 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0122] The memory 31 may be an internal storage unit of the terminal device 3, such as a hard disk or memory of the terminal device 3. The memory 31 may also be an external storage device of the terminal device 3, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device 3. Further, the memory 31 may also include both an internal storage unit and an external storage device of the terminal device 3. The memory 31 is used to store the computer program and other programs and data required by the roaming control device. The memory 31 may also be used to temporarily store data that has been output or is to be output.

[0123] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0124] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present invention. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0125] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0126] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.

[0127] An embodiment of the present invention provides a computer program product. When the computer program product runs on a mobile terminal, the mobile terminal can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0128] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the camera device / terminal device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk.

[0129] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0130] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0131] In the embodiments provided by the present invention, it should be understood that the disclosed devices / network equipment and methods can be implemented in other ways. For example, the device / network equipment embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0132] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed over multiple network units.

[0133] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0134] It should also be understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0135] As used in the present specification and the appended claims, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to monitoring, depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is monitored" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is monitored" or "in response to monitoring [described condition or event]", depending on the context.

[0136] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0137] References to "one embodiment" or "some embodiments" etc. described in the present specification mean that one or more embodiments of the present invention include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0138] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for detecting abnormal physiological and motor states of a human body, characterized in that: The method for detecting abnormal physiological and motion states of a human body is applied to a processor in a wearable device, wherein the wearable device includes a vital sign collector and a processor, and the method for detecting abnormal physiological and motion states of a human body includes: Acquire the real-time vital sign data collected by the vital sign collector and the standard vital sign data corresponding to the various preset states obtained by pre-training; Performing semantic classification on the real-time vital sign data to obtain real-time semantic information; each piece of the real-time semantic information corresponds to one or more of the preset states; According to the standard vital sign data corresponding to the real-time semantic information, it is determined whether the real-time vital sign data is in an abnormal state.

2. The method for detecting abnormal physiological and motor status of a human body as claimed in claim 1, characterized in that: The step of determining whether the real-time vital sign data is in an abnormal state according to the standard vital sign data corresponding to the real-time semantic information comprises: Acquiring standard vital sign data corresponding to the real-time semantic information; Calculating the distance between the real-time vital sign data and the standard vital sign data; If the distance is greater than a threshold, it is determined that the real-time vital sign data is in an abnormal state.

3. The method for detecting abnormal physiological and motor status of a human body as claimed in claim 1, characterized in that: Before the step of acquiring the real-time vital sign data collected by the vital sign collector and the standard vital sign data corresponding to the various preset states obtained through pre-training, the method further includes: Acquire first training physical sign data collected by the physical sign collector; the first training physical sign data is physical sign data of the target user collected by the physical sign collector when the target user is in a normal state; Acquire first historical vital sign data corresponding to each of a plurality of preset states; the first historical vital sign data is obtained by collecting vital sign data of other users in the preset states by a vital sign collector; The first training vital sign data and the first historical vital sign data are input into a target generation model to obtain the standard vital sign data output by the target generation model.

4. The method for detecting abnormal physiological and motor status of a human body as claimed in claim 3, characterized in that: The step of inputting the first training vital sign data and the first historical vital sign data into a target generation model to obtain the standard vital sign data output by the generation model comprises: Performing noise processing on the first historical vital sign data through a noise adding network in the target generation model to obtain first feature data; Matching first training vital sign data corresponding to the real-time semantic information; fusing the first feature data with first training vital sign data corresponding to the real-time semantic information to obtain second feature data; The second feature data is denoised by a denoising network in the target generation model to obtain standard vital sign data.

5. The method for detecting abnormal physiological and motor status of a human body as claimed in claim 3, characterized in that: Before the step of inputting the first training vital sign data and the first historical vital sign data into the target generation model to obtain the standard vital sign data output by the target generation model, the method further includes: Acquire second training vital sign data collected by the vital sign collector; Acquire second historical vital sign data corresponding to each of a plurality of preset states; Obtain the initial generation model, initial semantic information and label data; The initial generation model is trained according to the second historical vital sign data, the initial semantic information and the label data to obtain the target generation model.

6. The method for detecting abnormal physiological and motor status of a human body as claimed in claim 5, characterized in that: The step of training the initial generation model according to the second historical vital sign data, the initial semantic information and the label data to obtain the target generation model comprises: Performing noise processing on the second historical vital sign data through the noise adding network in the initial generation model to obtain third feature data; Acquire second training vital sign data corresponding to the initial semantic information; fusing the third feature data with the second training vital sign data corresponding to the initial semantic information to obtain fourth feature data; Performing denoising processing on the fourth feature data through a denoising network in the initial generation model to obtain output data; Calculating an error between the output data and the label data, and adjusting the initial generation model based on the error; Based on different first historical vital sign data, the step of performing noise processing on the second historical vital sign data through the noise adding network in the initial generation model to obtain first feature data and subsequent steps are iteratively performed to obtain the target generation model.

7. The method for detecting abnormal physiological and motor status of a human body as claimed in claim 1, characterized in that: The step of semantically classifying the real-time vital sign data to obtain semantic information comprises: The real-time vital sign data is input into a supervision model to obtain the semantic information output by the supervision model.

8. A device for detecting abnormal physiological and motor status of a human body, characterized in that: The device for detecting abnormal physiological and motion status of human body comprises: An acquisition unit, used to acquire the real-time vital sign data collected by the vital sign collector and the standard vital sign data corresponding to each of the multiple preset states obtained by pre-training; A classification unit, used for semantically classifying the real-time vital sign data to obtain real-time semantic information; each piece of the real-time semantic information corresponds to one or more of the preset states; The judgment unit is used to determine whether the real-time vital sign data is in an abnormal state according to the standard vital sign data corresponding to the real-time semantic information.

9. A terminal device, characterized in that: The terminal device includes: a memory, a processor, and a human physiological and motion state abnormality detection program stored in the memory and executable on the processor, wherein the human physiological and motion state abnormality detection program is configured to implement the steps in the human physiological and motion state abnormality detection method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for detecting abnormal physiological and motion states of a human body as described in any one of claims 1 to 7 are implemented.

Images