39 results about "Abnormal posture" patented technology

The invention relates to the technical field of myopia protection, in particular to a posture monitoring method and system for preventing myopia and protecting eyes. During application, image data of a user is collected, and dynamic spatial positions of key nodes such as head eyes, neck point locations and trunk shoulders are extracted and processed. In a single second preset time period, the deviation intensity of the head posture, the neck posture and the trunk posture relative to the respective standard posture in the second preset time period is monitored, prompt information of the corresponding intensity is generated, a user can be reminded to correct unconscious posture deformation in time, the prompt intensity is matched with the deviation intensity, and the user experience is improved. And a prompting effect can be achieved, and the user cannot be excessively interfered, so that myopia inducements such as eyeball ciliary muscle fatigue, crystalline lens adjustment spasm and binocular adjustment imbalance caused by posture abnormity are avoided. By monitoring the deviation frequency of the three postures in the third preset time period, when the frequency exceeds a safety value, forced intervention information is generated, a dual protection mechanism is formed, and the myopia prevention effect is further enhanced.

The invention discloses a knee joint protection device based on IMU (inertial measurement unit) and abnormal posture recognition, which comprises a protector body worn at the knee joint of a human body, an IMU posture sensing module is integrated on the protector body, and a nine-axis inertial sensor is used for collecting knee posture parameters; the main control unit is used for attitude data acquisition, identification and control logic processing; the electric control damping mechanism realizes action limitation based on a miniature electromagnetic brake or an adjustable friction material; a rule base is embedded in the abnormal posture recognition algorithm module, and abnormal standards are defined according to different application scenes; and the power supply and interface module is used for power management and data connection transmission. The knee joint protection method based on the IMU and the abnormal posture recognition applies the knee joint protection device based on the IMU and the abnormal posture recognition to carry out knee joint protection. Rigid constraint is replaced by electromagnetic damping'soft limiting ', and the comfort problem is solved; and the lightweight design is matched with a dual-guarantee mechanism to ensure the use safety, and functionality and practicability are achieved.

The present application relates to the technical field of rehabilitation equipment control, and particularly relates to a control method and system of a lower limb movement rehabilitation training machine. The method comprises the following steps: starting the brushless motor lower limb movement rehabilitation training machine, collecting the force detection parameters of the user's stepping and the operation feedback parameters of the brushless motor in real time; sending the resistance adaptation instruction to the brushless motor of the lower limb movement rehabilitation training machine based on the force detection parameters and in combination with the operation feedback parameters, and real-time regulating and controlling the output impedance of the brushless motor; detecting the posture signal of the lower limb movement rehabilitation training machine in real time during the continuous operation of the output impedance regulation and control of the brushless motor; and triggering the emergency stop instruction of the brushless motor immediately when the abnormal posture signal is detected. The present application adopts the brushless motor, combines the standby protection countdown with the zoned, angularly linked collection of the force parameters and the dynamic impedance regulation and control, so as to realize the portable and silent operation of the equipment and the personalized resistance adaptation, and to enhance the operation stability and the training adaptation accuracy.

The invention discloses a scoliosis degree evaluation and prediction method and system based on spine information, and the method comprises the steps: carrying out the classification processing of spine data and a spine full-length X-ray picture, determining the type of scoliosis, generating the scoliosis data of a patient, and carrying out the evaluation and prediction of the scoliosis degree according to the type of scoliosis and the scoliosis data of the patient. Determining scoliosis exercise training data, generating a spinal exercise training model and performing visual display; responding to an abnormal posture according to a daily posture real-time monitoring result of the patient and the scoliosis data of the patient; responding to a physiotherapy request in real time, based on a target partition and the scoliosis data of the patient, judging and determining the convex side position and the concave side position of scoliosis according to the scoliosis type, a target mode and the target partition, sending physiotherapy data to the target partition and executing a corresponding execution program, therefore, the accuracy, the real-time performance, the informatization level and the individuation level of scoliosis degree monitoring, evaluation and prediction are improved.

The invention relates to the technical field of rehabilitation assistance, and particularly discloses a lower limb rehabilitation training system which is used for solving the problems that an existing system is difficult to accurately judge and adapt to gradient change, and cannot provide enough assistance, so that the risk of slipping or secondary injury of a user is increased. The method comprises the following steps: integrating a plantar pressure sensor, an inertial measurement unit, a tilt angle sensor and a joint angle and torque sensor to obtain data, performing multi-point data fusion, and judging whether foot stress is uneven or not according to a first parameter; the plantar pressure in the whole gait tends to be balanced, the local fatigue or abnormal stress phenomenon is reduced, the slope walking risk coefficient is obtained through multi-sensor data fusion, the slope walking risk is judged in real time, the posture abnormal risk of a user is detected according to the slope walking risk coefficient and the first inclination parameter, plantar stress can be effectively balanced, and the user experience is improved. And the slope walking risk is judged in time, and the abnormal posture risk is detected, so that accurate auxiliary protection is realized.

The invention provides intelligent resistance training equipment and method. The method comprises the following steps: synchronously collecting resistance signals of a neck and lower limbs when a user executes a collaborative resistance action; determining force application delay time between the neck and the lower limbs according to the resistance signals of the neck and the lower limbs; according to the angular velocity component of the neck of the user and the anti-resistance signal of the neck, determining a posture-force coupling coefficient when the user executes the collaborative anti-resistance action; determining a load adjustment coefficient of a lower limb target load based on the force application delay time and the posture-force coupling coefficient in combination with a preset load adjustment rule; and adjusting the lower limb load applied by the equipment according to the load adjustment coefficient, and generating prompt information containing posture coordination abnormity at the same time. By adopting the scheme of the invention, the training load conforming to the current coordination capability of the user can be dynamically allocated.

The application provides a kind of adolescent spinal abnormal posture monitoring method based on behavior characteristic identification, it is related to image recognition field, in key point detection stage introduces key point confidence mechanism;Nonlinear spine projection function, torso behavior torsion degree factor and shoulder hip asymmetric disturbance function are constructed, and joint behavior abnormal feature vector is constructed according to nonlinear spine projection function, torso behavior torsion degree factor and shoulder hip asymmetric disturbance function;In the analysis of adolescent behavior image region, the spatial posture abnormal feature based on gray offset and edge modulation is proposed, the frequency abnormal feature of two-dimensional frequency transformation and nonlinear enhancement is combined, and the multiscale enhanced feature vector is formed by multiscale fusion, to realize the synchronous modeling of global structure and local details;In the abnormal identification and evaluation stage, classification evaluation function and abnormal severity scoring function are constructed by joint behavior abnormal feature vector and multiscale enhanced feature vector, the abnormal posture type is identified, and the grading result is output.

This invention relates to the field of intelligent sports training equipment technology. It discloses an intelligent sports training machine with motion posture recognition, comprising: an equipment frame; a drive assembly mounted on the equipment frame, the drive assembly including: a motor; a reel connected to the output shaft of the motor; and a rope wound around the reel, the free end of which extends to the outside of the equipment frame. This invention offers significant advantages over traditional training machines. It achieves stepless resistance adjustment through motor drive, and with closed-loop control, ensures precise and stable resistance that can be dynamically adjusted according to posture. It achieves three-dimensional recognition of key points of the human skeleton through visual acquisition and posture analysis, and combined with audio-visual feedback, allows users to correct their movements in real time, improving training effectiveness. It automatically reduces resistance in abnormal postures and provides emergency braking in dangerous situations, significantly reducing the risk of injury. It can also record training data for digital management, providing support for personalized training.

The invention discloses a physical health monitoring method and system based on big data, and relates to the technical field of intelligent sensors, and the method comprises the steps: obtaining a posture data sequence related to physiological postures in daily activities of a user, and extracting the distribution characteristics of abnormal postures based on the posture data sequence; determining a quantized value of accumulated pressure based on the distribution characteristics of the abnormal postures; if the quantized value of the accumulated pressure exceeds a preset accumulated pressure quantized threshold value, judging a risk level of physiological health of the user based on the quantized value of the accumulated pressure, and generating a trend index reflecting a pressure accumulation rule based on the risk level; and based on the trend index, carrying out weighted adjustment on the distribution characteristics of the abnormal postures, and generating a health monitoring feedback sequence based on the adjusted distribution characteristics of the abnormal postures, so that the physical health monitoring accuracy can be improved.

The invention discloses an IMU and abnormal posture recognition-based ankle joint sprain prevention device, which comprises a flexible wearing structure worn at the ankle joint of a human body, and is characterized in that a plurality of IMU posture sensing modules are integrated on the flexible wearing structure to collect the posture angle, angular velocity and acceleration data at the ankle in real time; the central control unit is electrically connected with the plurality of IMU attitude sensing modules and is used for judging abnormal actions and outputting control signals; the dynamic limiting execution mechanism is electrically connected with the central control unit and is used for receiving the abnormal action signal and executing dynamic intervention; and the power supply and communication module is arranged on the flexible wearable structure to supply power to the device, and performs wireless data communication with an external mobile terminal for a user to check and train the model. According to the control method of the ankle joint sprain prevention device based on the IMU and abnormal posture recognition, the posture of the ankle joint can be monitored in real time, high-risk twisting actions can be recognized, and a protection mechanism can be started in real time, so that ankle joint sprain is effectively prevented, and the exercise safety is improved.

The invention provides a hospital real-time navigation path planning method and system based on deep learning, and relates to the technical field of wisdom medices.The walking posture of a patient is analyzed in real time through a deep learning network model, firstly, the patient with the normal walking posture and the patient with the abnormal walking posture are recognized and distinguished, and according to the actual environment condition of a hospital, the real-time navigation path of the hospital is planned; according to the method, a walking path line suitable for patient demands is planned, an optimal navigation walking path suitable for special crowds such as elderly patients, pregnant women and handicapped people can be provided for the special crowds such as the elderly patients, the pregnant women and the handicapped people, the problems of walking difficulty and cross infection of the special crowds are solved, and the doctor-seeing experience satisfaction degree of the patients and the hospital service efficiency are improved. Meanwhile, the walking posture of the patient is tracked, a secondary abnormity alarm mechanism is set, the path is re-planned in case of an abnormal posture, and an alarm is given, so that the patients of special crowds are rescued at the first time, and the service quality is further improved.

To provide an abnormal posture detection apparatus configured to prevent adverse effect on detection processing without requiring manual labeling work or image data photography or limiting detectable abnormal postures.SOLUTION: An abnormal posture detection apparatus includes: a skeletal position acquisition unit which acquires three-dimensional skeletal position of a person; a vector acquisition unit which acquires a perpendicular vector, a right-leg vector and a left-leg vector based on the three-dimensional skeletal position; a calculation unit which calculates a right-leg angle formed by the perpendicular vector and the right-leg vector, and a left-leg angle formed by the perpendicular vector and the left-leg vector; a pre-processing unit which performs pre-processing to adjust the three-dimensional skeletal position to be aligned to a predefined direction and size; and a determination unit which determines whether the person is in an abnormal posture, using a learning model generated by machine learning using the pre-processed three-dimensional skeletal position, the right-leg angle and the left-leg angle.SELECTED DRAWING: Figure 1

The embodiment of the invention relates to the technical field of auxiliary judgment, and discloses a method and equipment for constructing an auxiliary judgment model for an abnormal body position of a living body and a storage medium, and the method comprises the following steps: performing data analysis on video data of the living body in different emotions or different disease states to obtain a plurality of three-dimensional skeletons with continuous time; calculating the three-dimensional information of the three-dimensional skeleton to obtain a plurality of dynamic motion parameters, and marking the dynamic motion parameters as corresponding emotion or disease states; a living body behavior feature training set is established; and by taking the kinetic parameters as input and the corresponding emotion or disease state as output, training a living body abnormal body position auxiliary judgment model through the living body behavior feature training set to obtain a single living body abnormal body position auxiliary judgment model corresponding to the current species, so that the single living body abnormal body position auxiliary judgment model corresponding to the current species can be obtained through the analysis process. The emotion or disease state is connected with the behavior characteristics, so that the emotion or disease state of the living body can be analyzed and determined through the abnormal body position.

The application discloses a spastic cerebral palsy infant model for rehabilitation nursing skill teaching, which comprises a head, upper limbs, a trunk, a hip and lower limbs, and a head turning device is connected to the top inner side of the trunk, the upper end of the head turning device is connected with the head, and the head turning device is connected with a head tightness adjusting mechanism; the upper limbs and the lower limbs are respectively provided with limb simulation structures; the upper limbs are provided with upper limb pressure detection modules, and the lower limbs are provided with lower limb pressure detection modules above and below. The child model can simulate abnormal postures and positions of spastic cerebral palsy children during spasm, such as frog postures, knife-fold postures and the like, and further assists operators to carry out corresponding operation practice. In the implementation, the tightness adjuster of different grades can be adjusted to adjust the spasticity degree, so that the change of muscle tension of spastic cerebral palsy children during spasm is simulated, and thus the simulation teaching of cerebral palsy rehabilitation nursing skills is realized.

The invention discloses a lower limb ultrasonic stimulation method and device based on low-intensity focused ultrasound, and the method comprises the steps: obtaining the multi-channel pressure signal data of the sole of a patient in the movement process, and carrying out the real-time sampling and preprocessing, and obtaining the target multi-channel pressure signal data; calculating motion state parameters of the patient according to the target multi-channel pressure signal data by using a data processing algorithm, identifying an abnormal posture according to the motion state parameters, and marking a target area needing to be corrected; generating an initial parameter of a focused ultrasonic stimulation signal according to the motion state parameter analysis result, and generating focused ultrasonic waves according to the initial parameter; adjusting focused ultrasonic waves to accurately stimulate a target area which is marked on the lower limb and needs to be corrected, so as to activate or inhibit related muscles and correct the motion posture of the patient; and collecting rehabilitation progress data of the patient, and generating a personalized training scheme for rehabilitation of the patient according to the rehabilitation progress data. The efficiency and safety of rehabilitation training are improved.

The application relates to the technical field of human body impedance measurement, and provides a human body impedance abnormality measurement posture recognition method, an electronic device and a storage medium. The human body impedance value of a target object is measured according to a plurality of excitation current signals of different frequencies, and a plurality of target impedance values formed by the excitation current signals of different frequencies are obtained. Whether the measurement posture of the target object is an abnormal posture is determined according to the plurality of target impedance values. Through the recognition of the abnormal posture, the problem that the measurement accuracy is not improved is solved, and the measurement accuracy is improved.

The invention relates to the technical field of medical rehabilitation, in particular to a training device for postoperative rehabilitation of knee arthritis patients, which comprises a fixing plate, leg lifting assemblies are symmetrically arranged on the fixing plate, sliding groove assemblies are arranged between the leg lifting assemblies and the fixing plate, and pedal assemblies are symmetrically arranged on the sliding groove assemblies; symmetrical air bag assemblies are arranged on the inner sides of the leg lifting assemblies, a reset assembly is further fixedly connected to the fixing plate, the air bag assemblies communicate with the two ends of the reset assembly correspondingly, and the reset assembly is used for pushing the air bag assemblies to reset based on sensing information of the air bag assemblies so as to correct the posture of a patient. Elastic assemblies used for sensing the inward turning or outward turning condition of the foot sole of the patient during rehabilitation training are symmetrically arranged on the side, close to the pedal assembly, of the fixing plate, one end of each elastic assembly is fixedly connected with the pedal assembly, and the other end of each elastic assembly is hinged to the reset assembly. Through a closed-loop feedback mechanism of air bag induction and mechanical linkage, real-time detection and dynamic correction of abnormal postures of multiple joints of the lower limb are achieved.

The application relates to a mixed reality office scene bad sitting posture correction method and device, equipment and medium, which are suitable for a mixed reality head-mounted device. The method continuously acquires current spatial posture information of a user's head and shoulders, calculates head position offset, left shoulder position offset, right shoulder position offset and head angle offset based on the current spatial posture information and a reference posture model, and judges whether the user is in an abnormal posture state according to each offset and a corresponding threshold value. When the abnormal posture state lasts for more than a preset time length, the mixed reality head-mounted device is controlled to visually modulate display content to prompt the user to adjust the posture. The application can improve the accuracy of bad sitting posture recognition and reduce the interference on the user's office process.

The invention discloses a novel lightweight human body abnormal posture detection method based on YOLOv8n, and relates to the technical field of human body posture detection, and the method comprises the steps: adding a PoseMSA attention module in a backbone network of the original YOLOv8n, and fusing the attention module with a C2f module in the backbone network; a KA-Sample up-sampling module is adopted to replace an original up-sampling module of YOLOv8n; a Detect-PSA detection head is adopted to replace a detection head of an original YOLOv8n, and a lightweight human body abnormal posture detection model PSD-YOLOv8n is obtained. The method solves the problems that in the prior art, in an abnormal detection scene with a complex background, shielding interference and obvious human body posture difference, recognition is difficult, the real-time performance is difficult to guarantee, and the boundary is fuzzy.

The invention relates to the technical field of auxiliary rehabilitation, and discloses a joint movement detection method and system, computer equipment and a storage medium, and the method comprises the steps: obtaining joint movement video data of a to-be-measured user; joint two-dimensional coordinate information in one-to-one correspondence with the joint points is obtained; converting the joint two-dimensional coordinate information into joint three-dimensional coordinate information based on depth information in the joint activity video data; determining joint motion range measurement information of the user to be measured based on the joint three-dimensional coordinate information corresponding to each frame of image; judging whether the to-be-measured user has an abnormal posture or action or not based on the joint motion range measurement information; and when the user to be measured has the abnormal posture or action, generating an abnormal analysis report. According to the invention, full automation of joint motion detection is effectively realized, the joint motion range measurement efficiency and the judgment efficiency of abnormal user motion conditions are improved, personal errors are reduced, and the accuracy of joint motion range and abnormal posture or motion detection is ensured.

This invention provides an interactive sports training assessment method and system, belonging to the field of sports training assessment technology. The method involves: distributing an assessment item set in a virtual training assessment scenario; sorting the items according to the trainee's exercise habits and psychological state for each assessment item; sequentially displaying the items and capturing training behavior posture sets; globally comparing these with a standard posture set to construct joint point abnormal posture vectors; and combining joint point behavior trajectories with standard trajectories to construct point abnormal vectors; combining and analyzing these results to obtain single-item assessment results; and finally, processing all results and sending them to the assessment terminal. This invention achieves standardized quantification of assessments, improves the objectivity and accuracy of assessments, and provides a scientific reference for sports training.

The invention relates to the technical field of spine protection, in particular to a posture correcting and spine protecting table and chair regulation and control method and system. According to the posture correcting and spine protecting table and chair adjusting and controlling method, the pressure sensor is arranged on the back of the spine protecting chair, the standard pressure data of the cervical vertebra area, the thoracic vertebra area and the lumbar vertebra area are collected when the user sits and is in the reference posture, and in the learning process of the user, the posture correcting and spine protecting table and chair adjusting and controlling method is achieved based on comparison of the real-time pressure data and the standard data. And abnormal postures such as abnormal neck, humpback, heeling and waist strain can be identified. According to different abnormal postures, the postures of the spine protection table and the spine protection chair are dynamically adjusted, so that a reasonable strategy is provided for the abnormal postures of the user, and spine correction is carried out. Finally, the spine protection table and the spine protection chair are slowly adjusted to the initial state for the next posture correction, the user can be guided to naturally return to the reference posture meeting the physiological curvature of the spine in the adjusting process, and the strain risk of the spine caused by the abnormal posture is effectively eliminated.

The invention discloses a posture sensing and fatigue early warning method and system based on an intelligent sportswear, and the method comprises the steps: collecting posture data and electromyographic signals in real time when a user moves through a flexible stretching sensor and an electromyographic sensor on the intelligent sportswear; after processing the collected data, analyzing attitude normalization based on a biomechanical algorithm, and evaluating the muscle fatigue degree in combination with the electromyographic signals; when the abnormal posture exists or the fatigue degree exceeds a preset threshold value, an early warning action is triggered, specifically, tactile feedback is output through the flexible vibration motor, and early warning information containing a posture correction suggestion and a fatigue recovery scheme is sent to the user terminal. The system comprises a sensing module, a processing analysis module, a feedback execution module and a user terminal interface module, can realize integration of exercise data acquisition, analysis and early warning feedback, provides personalized exercise protection guidance for a user, and is suitable for daily exercise training and rehabilitation scenes.

This invention relates to the field of motion analysis technology, specifically disclosing a badminton motion data analysis method and system that combines big data. It captures athletes' joint angles, speeds, and accelerations in real time, and uses continuous wavelet transform and phase synchronization coefficients to calculate the consistency coefficient on the time series, ensuring data accuracy and reliability. Based on accurate data, it extracts short-term fine-tuning features and long-term trend features to construct comprehensive time-dimensional features. It also extracts joint angle change features and body posture features to form comprehensive spatial-dimensional features. These features are integrated into a comprehensive feature representation, input into a clustering model to generate an athlete's posture recognition score. Based on the score results, the athlete's posture is classified. For severely abnormal postures, it provides real-time feedback instructions based on spatiotemporal location to guide athletes in adjusting their movements, and establishes a closed-loop feedback system to monitor the correction effect.

The invention provides a correcting shoe with foot valgus and valgus and claw-shaped toe interior eight correcting functions, and belongs to the technical field of shoes, the correcting shoe comprises a sole, a vamp, an insole, a shoelace and an insole, and further comprises a supporting part which is arranged at a welt in the vamp, the bottoms of the two ends of the supporting part are sleeved with two symmetrically-arranged correcting assemblies, and the top of the supporting part is fixedly connected with a limiting piece; adjusting assemblies which correspond to the correcting assemblies in position and are the same as the correcting assemblies in number are arranged in the limiting piece. The correcting assembly comprises a hollow piece, the top of the hollow piece is connected with the supporting part, and the bottom of the hollow piece penetrates through the vamp and is connected with the sole; the adjusting belt is arranged in the hollow part, the bottom of the adjusting belt penetrates through the side wall of the hollow part and then is connected with the bottom of the insole, and the top of the adjusting belt is connected with the adjusting assembly through an adjusting rope; through the cooperation of the adjusting assembly and the adjusting rope, the contraction of the adjusting belts in the correcting assembly can be adjusted, and the abnormal posture of the foot can be corrected in a targeted manner by respectively controlling the tension of the adjusting belts on the inner side and the outer side of the shoe.

The invention discloses a human body abnormal posture recognition method used in an assembly work man-machine cooperation environment, and relates to the technical field of safety monitoring. According to the method, normal skeleton data and repaired skeleton data are subjected to data enhancement on human body posture key point related data by using a Cycle-GAN network, and human body posture recognition is performed by using a double-flow network; wherein a spatial flow branch in the double-flow network can obtain global feature representation of human body posture key points and high-order node features of rich topological structure information, and a time flow branch can obtain time sequence features with a global time sequence relation and dynamic posture change feature representation; and the self-attention layer performs weighted aggregation on the time sequence features and the high-order node features, endows the high-order node features corresponding to key time steps in the time sequence features with higher weights, and outputs joint feature representation with time dimension information and space dimension information. The average recognition precision obtained by the method reaches up to 84.1%.

Embodiments of the present application relate to the technical field of information processing, and disclose a scene intelligent adjustment method and device, equipment and a storage medium, comprising: periodically identifying a user posture appearing in a place; judging whether an abnormal posture meeting a preset abnormal posture condition exists in the identified user posture; and when it is determined that the identified user posture includes an abnormal posture, intelligently adjusting a scene in the place where the abnormal posture appears. The scene intelligent adjustment method, device, equipment and storage medium disclosed by the present application solve the problem that the existing method requires a user to manually select a scene and set a trigger condition, which is inconvenient to use and cannot adjust the scene according to the actual use, so that the scene in the place is intelligently adjusted based on the user posture, without the need for manual selection of the scene or setting of a scene adjustment trigger instruction, and the scene is adjusted according to the actual situation, thereby effectively improving the convenience and flexibility of the user in adjusting the scene.

A method and device for describing the human body structure relationship based on spatial transformation, belonging to the field of medical information processing. It is set that each rigid component of the human body is represented by a spatial triangular surface formed by three adjacent bone feature points. Typical rigid components include the upper triangular plane of the shoulder, the left and right scapula planes, the thoracic vertebrae, the lumbar vertebrae, the pelvic plane, and the left and right lower limb planes. An independent local coordinate system of its own is established for each rigid component. The local coordinate system of its own of the rigid component pelvic plane is established as the base coordinate system, and the coordinate system where the upper camera is located is the global world coordinate system. The spatial rotation transformation matrix of each rigid component relative to the pelvic base coordinate system is obtained through the coordinate transformation under the local coordinate system of its own of each rigid component and the global world coordinate system. The singular value decomposition method is used to solve the relative spatial rotation transformation matrix, and a mathematical expression for describing the spatial position relationship between the bone parts of the human body with abnormal postures is constructed. The present invention is used to accurately detect the abnormal body postures and structures of the human body.

The application discloses a ward patient abnormal posture AI visual detection method and device, performs privacy protection skeleton extraction on a monitoring image frame sequence to determine a posture monitoring area, forms a monitoring priority distribution based on a static time length reverse priority layering, combines an activity restriction medical order to match to execute individualized baseline correction and limited joint attenuation weight calibration, forms a weighted posture feature and establishes a posture reference atlas, continuously recognizes a unilateral cumulative deviation of a shoulder and hip connecting line inclination angle after a turning-over standard is reached based on the posture reference atlas, performs risk level marking on a detected deviation event to generate a pre-trigger frame group, captures a deterioration trend on the pre-trigger frame group and forms a pre-warning start parameter to construct a posture disposal strategy, and finally corrects an anchor point position and adequacy by bed inspection in place recording and patient historical deviation recurrence interval respectively to generate a nursing pre-warning dispatch instruction, which improves the individualized adaptation degree of a patient body position deviation detection and the timing matching accuracy of a pre-warning response.

The invention provides a teenager spine abnormal posture monitoring method based on behavior feature recognition, and relates to the field of image recognition. A key point confidence calculation mechanism is introduced in a key point detection stage; constructing a nonlinear spine projection function, a trunk behavior torsion factor and a shoulder and hip asymmetric disturbance function, and constructing a joint behavior anomaly feature vector according to the nonlinear spine projection function, the trunk behavior torsion factor and the shoulder and hip asymmetric disturbance function; according to the method, spatial attitude anomaly features based on gray offset and edge modulation are put forward in adolescent behavior image region analysis, frequency anomaly features of two-dimensional frequency conversion and nonlinear enhancement are combined, multi-scale enhancement feature vectors are formed through multi-scale fusion, and synchronous modeling of a global structure and local details is achieved; and in the exception identification and evaluation stage, a classification evaluation function and an exception severity scoring function are constructed by combining the behavior exception feature vector and the multi-scale enhanced feature vector, and an exception posture type is identified and a grading result is output.