Nursing bed capable of monitoring health state of patient

Abstract

The invention discloses a nursing bed capable of monitoring the health state of a patient, and relates to the field of nursing beds, the nursing bed comprises a bed frame, the two sides of the bed frame are provided with blocking frames, the top end of the bed frame is provided with a mattress, the two ends of the bed frame are provided with a display control unit and an infrared blood oxygen detector respectively, and the bottom end of the bed frame is provided with a placing plate through four sets of supporting legs; an excrement collecting box is arranged at the top end of the placing plate, a defecation auxiliary assembly is arranged at the bottom end of the bed frame, and the display control unit is used for receiving and displaying monitoring data of a patient, recognizing the state of the patient according to a monitoring data result and adjusting the control state of the nursing bed based on the requirement of the patient. Facial expressions and dining behavior data can be continuously collected through the camera, matched intervention content can be found and triggered in time when a patient has negative emotion, psychological stress of the patient is effectively relieved, meanwhile, the diet structure and regular diet rhythm of the patient can be monitored based on dining information, and the patient experience is improved. And timeliness and effectiveness of personalized nutrition intervention are guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of nursing beds, in particular to a nursing bed capable of monitoring the health status of a patient. BACKGROUND

[0002] A nursing bed is a multifunctional medical bed specially designed for bedridden patients or people with physical disabilities. It combines ergonomics, safe support, nursing convenience, and intelligent technology, and can be used in hospital, health care institutions or home care scenarios. With the development of artificial intelligence, sensing technology and intelligent hardware, modern nursing beds have gradually evolved from traditional mechanical nursing devices to intelligent nursing platforms that integrate monitoring, identification, intervention, communication and data management.

[0003] After integrating AI and multi-modal sensing technology, the nursing bed has become a key node device in the intelligent medical system that connects patients and nursing services, and has important significance for building a people-centered continuous health care ecosystem.

[0004] However, the nursing bed in the prior art lacks patient emotion recognition and diet rationality recognition functions during application, which makes it impossible to dynamically determine whether the patient's diet is reasonable according to the patient's specific condition, recovery stage or medical order, so as to provide a truly personalized intervention plan, and cannot actively identify the patient's negative emotional state, which is easy to cause the patient's emotional abnormality to be ignored for a long time, affecting the patient's recovery process.

[0005] In view of the problems in the related art, no effective solution has been proposed so far. SUMMARY

[0006] In view of the problems in the related art, the present application proposes a nursing bed capable of monitoring the health status of a patient to overcome the above technical problems existing in the prior art.

[0007] To this end, the specific technical solutions adopted by the present application are as follows: A nursing bed capable of monitoring the health status of a patient, comprising a bed frame, the two sides of the bed frame are provided with a blocking frame, the top end of the bed frame is provided with a mattress, the two ends of the bed frame are respectively provided with a display control unit and an infrared blood oxygen detector, the bottom end of the bed frame is provided with a placing plate through four groups of supporting legs, the top end of the placing plate is provided with a bedpan, and the bottom end of the bed frame is provided with a defecation assisting assembly, wherein the display control unit is used to receive and display the monitoring data of the patient, and identify the state of the patient according to the monitoring data result, and adjust the control state of the nursing bed based on the patient's needs.

[0008] Preferably, in order to utilize the temperature sensor in the mattress, the patient's body temperature is measured in real time and accurately, and the collected body temperature data is transmitted to the display control unit for health status detection. At the same time, the air pump can be automatically controlled to inflate and deflate different areas in sequence according to the preset time interval, to relieve the support force of the mattress surface on each part of the patient's body, avoid long-term pressure on local tissues, effectively reduce the occurrence of pressure ulcers, and install high-precision pressure sensors on the four supporting legs of the nursing bed. When the patient lies on the bed, the pressure sensor measures the pressure borne by the bed body, calculates the patient's body weight, and transmits it to the display control unit for health status detection. The inside of the mattress is provided with two groups of placement cavities, one group of placement cavities is provided with a plurality of groups of temperature sensors, and the other group of placement cavities is provided with a plurality of groups of air pumps. The top middle of the mattress is provided with a toilet gap. The inside bottom of the supporting leg is provided with a pressure sensor. The top of the collection box is provided with an excrement inlet. The inside of the collection box is provided with a collection barrel. The inner wall of the collection box is provided with a double-layer activated carbon filter screen. One side of the infrared blood oxygen detector is provided with a camera.

[0009] Preferably, in order to start the electric push rod one to push the bed frame to complete the comfortable toilet position adjustment when the patient needs, and use the electric push rod two to control the state of the mattress patch, it is convenient for the patient to use the toilet, and the excrement directly falls into the inside of the collection box to realize the collection of the excrement. The collection box is provided with a double-layer activated carbon filter screen to adsorb odors in real time to avoid affecting the environment. The toilet auxiliary assembly includes two groups of electric push rods one with opposite directions installed on the top of the bed frame. One side of the electric push rod one is connected with a connecting rod, and the connecting rod is connected with the bottom of the bed frame. The outside of the bed frame is provided with a servo motor. The output shaft of the servo motor is provided with a lead screw. One side of the lead screw and located at the bottom of the bed frame is provided with a limiting rod. The outside of the lead screw and the limiting rod is provided with an adjuster. The adjuster includes a sliding frame installed on the outside of the lead screw and the limiting rod. The top of the sliding frame is provided with a sliding rail. One side of the sliding frame is provided with an electric push rod two. One side of the electric push rod two is connected with a sliding rod one. One side of the sliding rod one is connected with a sliding rod two. One side of the sliding rod one is connected with the sliding rail. The top of the sliding rod one and the sliding rod two is provided with a mattress patch. The size of the mattress patch is the same as the size of the toilet gap.

[0010] Preferably, the display control unit comprises: a health assessment alarm module for obtaining the monitoring results of the temperature sensor, the infrared blood oxygen detector and the pressure sensor, and judging the health status of the patient according to the detection results, and generating alarm information based on the state and sending it to the user end; a nursing bed use adjustment module for controlling the use state of the air pump and the toilet auxiliary assembly according to the preset time and the patient's needs; an emotional state evaluation module, configured to identify an emotional state of the patient based on the monitoring video of the camera, and dynamically match intervention content according to the identification result to adjust the emotion of the patient; a nutritional status evaluation module, configured to extract meal information of the patient, evaluate nutritional intake of the patient based on the meal information, and determine rationality of eating of the patient in combination with the medical record of the patient, and generate a voice prompt according to the rationality result; a control state data display module, configured to display result data of the health evaluation alarm module, the nursing bed use adjustment module, the emotional state evaluation module and the nutritional status evaluation module.

[0011] Preferably, the emotional state evaluation module comprises: a face key frame output module, configured to collect video data of the patient by using the camera at regular time intervals, extract a key frame from the video data according to a pre-set time interval, and output a face key frame of the patient according to a face detection result in the key frame; a face image determination module, configured to generate a face structure vertex based on a superpixel block of the face key frame of the patient, and determine a face image region of the patient by using clustering learning and Markov technology according to density information of the face structure vertex; an emotional state identification module, configured to extract contour information in the face image region of the patient, generate a visual saliency map in combination with a visual saliency model, and identify a variance distance between the contour information by using a second-order oscillation technology to output the emotional state of the patient; a patient emotion adjustment module, configured to establish a matching relationship between the emotional state and a local video tag of the nursing bed, determine the local video of the nursing bed corresponding to the emotional state of the patient based on the matching relationship, and extract the local video of the nursing bed to adjust the emotion of the patient.

[0012] Preferably, the face structure vertex is generated based on the superpixel block of the face key frame of the patient, and the face image region of the patient is determined by using clustering learning and Markov technology according to the density information of the face structure vertex, which comprises: uniformly arranging seed points in the face key frame of the patient, searching for image pixel points in a pre-set range of the seed points according to the arrangement result, and completing image seed processing after the seed points coincide with the minimum gradient position in the pre-set range; performing color conversion on the face key frame of the patient, calculating color distance and spatial distance between the image pixel points and the seed points according to the conversion result, generating a distance measure value, and outputting the superpixel block based on the distance measure value; extracting contour distribution of the face key frame according to the superpixel block, performing initial segmentation on the superpixel block by using the extraction result, and judging similarity between each superpixel block in the face key frame based on the segmentation result; Based on the similarity results, the density information of each superpixel block vertex is estimated, and the face structure vertex is output according to the estimation results. The density information is re-clustered using clustering learning technology and combined with the face structure vertex to obtain the face region connection map. The gradient operator of the face region connectivity graph is analyzed using Markov extraction technology. The gradient components of the patient's face points in different directions are obtained based on the gradient operator, and the directional gradient histogram is generated to determine the patient's face image region.

[0013] Preferably, the formula for calculating density information is: ; In the formula, delta a Indicates the first a Density information estimation of the vertices of a superpixel block, H This indicates the total number of superpixel blocks. S b Indicates the first b The similarity between a superpixel block and its neighboring superpixel blocks. c Indicates the index of the superpixel block. D ε represents the distance metric. l,l+1 Represents superpixel block segmentation points l Segmentation points with adjacent superpixel blocks l Weights between +1, h This represents the bandwidth of the contour distribution.

[0014] Preferably, contour information is extracted from the patient's facial image region, combined with a visual saliency model to generate a visual saliency map, and second-order oscillation technique is used to identify the variance distance between contour information. The patient's emotional state is then output, including: The patient's face image is scaled to the desired size, and the coordinates of any pixel in the adjusted face image are obtained. The translation amount of the pixel is determined based on the coordinates, and a first-order approximation of the translation amount is obtained through Taylor expansion. The feature values ​​of the first-order approximation of each position in the patient's face image are obtained, and the sum matrix and difference matrix are constructed according to the absolute value of the sum and difference of any feature values. The sum matrix and difference matrix are normalized, and the facial point saliency map and contour edge saliency map are output based on the processing results. Based on the visual saliency model, the facial point saliency map and the contour edge saliency map are linearly combined to output the patient's facial visual saliency map, and the edge detection technology is used to extract the patient's organ shape information from the patient's facial visual saliency map. The dynamic change trend of the organ shape information is captured by fitting the oscillation curve, the contour change amplitude of each organ is measured based on the dynamic change trend, and the structure variance feature is output by implementing variance extraction processing on the contour change amplitude.

[0015] Preferably, the nutritional status evaluation module comprises: The meal information extraction module is configured to extract the patient's meal food image from the monitoring video of the camera and obtain the patient's dining behavior in the meal food image based on image recognition technology. The evaluation index construction module is configured to generate a comprehensive evaluation index system according to the patient's medical record data, analyze the membership degree of the comprehensive evaluation index system by using a fuzzy hierarchical algorithm, and determine the patient's intake rationality standard. The diet rationality judgment module is configured to quantify the patient's intake nutritional structure and diet behavior according to the dining behavior, and match the quantification result with the rationality standard to output the patient's eating rationality judgment result. The voice prompt generation module is configured to adaptively match the corresponding diet suggestion content according to the rationality judgment result, and transmit the diet suggestion content to the medical staff terminal through wireless technology for patient diet intervention.

[0016] Preferably, the comprehensive evaluation index system is generated according to the patient's medical record data, and the membership degree of the comprehensive evaluation index system is analyzed by using a fuzzy hierarchical algorithm to determine the patient's intake rationality standard, which comprises: Based on the patient's medical record data and disease state, the nutritional structure and eating frequency that the patient should intake are analyzed to construct a comprehensive eating evaluation index set, and the importance of each index is determined by using a subjective weighting method. The grade comments of each index are established according to the importance, the membership degree value of each index under each grade is obtained based on a trapezoidal membership function, and the fuzzy membership degree of each grade is analyzed by using the membership degree value. The fuzzy rule matrix is constructed based on the membership degree value and the fuzzy membership degree, the fuzzy rule matrix is multiplied by the weight vector to determine the comprehensive rule result of the nutritional structure and eating frequency, and the patient's intake rationality standard is output.

[0017] The present application has the following advantages: 1. The temperature sensor in the mattress measures the patient's body temperature in real time and accurately, and transmits the collected body temperature data to the display control unit for health status detection. At the same time, the air pump is automatically controlled to inflate and deflate different areas in sequence according to the preset time interval, which relieves the support force of the mattress surface on each part of the patient's body, avoids long-term pressure on local tissues, effectively reduces the occurrence of pressure injuries, and installs high-precision pressure sensors on the four supporting legs of the nursing bed. When the patient lies on the bed, the pressure sensor measures the pressure borne by the bed body, calculates the patient's weight, and transmits it to the display control unit for health status detection.

[0018] 2. The toilet auxiliary assembly is provided, which can start the electric push rod to drive the bed frame to complete comfortable toilet position adjustment when the patient needs, and control the state of the mattress patch block by the electric push rod two, so that the patient can use the toilet, and the excrement directly falls into the inside of the excrement collecting box, realizes the collection of excrement, and the double-layer activated carbon filter screen is built-in in the excrement collecting box, which can adsorb odor in real time to avoid affecting the environment.

[0019] 3. The camera can continuously collect facial expression and meal behavior data, can timely discover and trigger matched intervention content when the patient has negative emotions, effectively relieve the psychological pressure of the patient, improve the spiritual comfort, and based on the meal information, can monitor the diet structure and regular diet rhythm of the patient, combine the patient's medical record to judge the rationality of intake in real time, and output intuitive prompt through voice, help the patient to adjust independently, guarantee the timeliness and effectiveness of personalized nutrition intervention, and further not only significantly improve the physical and mental health guarantee and life quality of the patient, but also optimize the nursing process. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1 It is a structure schematic diagram of a nursing bed capable of monitoring the health status of a patient according to an embodiment of the present application; Figure 2 It is a structure schematic diagram of a nursing bed capable of monitoring the health status of a patient from another angle according to an embodiment of the present application; Figure 3 It is a structure schematic diagram of a nursing bed capable of monitoring the health status of a patient according to an embodiment of the present application; Figure 4is a structure diagram of a regulator in a nursing bed capable of monitoring a patient's health status according to an embodiment of the present application; Figure 5 is a sectional view of a mattress in a nursing bed capable of monitoring a patient's health status according to an embodiment of the present application; Figure 6 is a sectional view of a support leg in a nursing bed capable of monitoring a patient's health status according to an embodiment of the present application; Figure 7 is a sectional view of a toilet box in a nursing bed capable of monitoring a patient's health status according to an embodiment of the present application; Figure 8 is a principle block diagram of a display control unit in a nursing bed capable of monitoring a patient's health status according to an embodiment of the present application; Figure 9 is a principle block diagram of a mood state evaluation module in a nursing bed capable of monitoring a patient's health status according to an embodiment of the present application; Figure 10 is a principle block diagram of a nutrition state evaluation module in a nursing bed capable of monitoring a patient's health status according to an embodiment of the present application.

[0022] In the drawings: 1, bed frame; 2, blocking frame; 3, mattress; 4, display control unit; 401, health evaluation alarm module; 402, nursing bed use adjustment module; 403, mood state evaluation module; 4031, face key frame output module; 4032, face image determination module; 4033, mood state recognition module; 4034, patient mood adjustment module; 404, nutrition state evaluation module; 4041, meal information extraction module; 4042, evaluation index construction module; 4043, diet rationality judgment module; 4044, voice prompt generation module; 405, control state data display module; 5, infrared blood oxygen detector; 6, support leg; 7, placing plate; 8, toilet box; 9, toileting auxiliary assembly; 901, electric push rod one; 902, connecting rod; 903, servo motor; 904, screw rod; 905, limiting rod; 906, regulator; 907, sliding frame; 908, sliding rail; 909, electric push rod two; 910, sliding rod one; 911, sliding rod two; 912, mattress filling block; 10, placing cavity; 11, temperature sensor; 12, air pump; 13, toileting gap; 14, pressure sensor; 15, defecation port; 16, collection bucket; 17, double-layer activated carbon filter screen; 18, camera. DETAILED DESCRIPTION

[0023] To further illustrate the embodiments, the present application provides accompanying drawings which are part of the disclosure of the present application, mainly used to illustrate the embodiments, and can be explained in conjunction with the related description of the specification to understand the operating principle of the embodiments. Those skilled in the art should understand other possible implementations and advantages of the present application by referring to these contents.

[0024] According to an embodiment of the present application, a nursing bed capable of monitoring the health status of a patient is provided.

[0025] The present application will be further described in conjunction with the accompanying drawings and specific embodiments, as shown in Figures 1-7 The nursing bed capable of monitoring the health status of a patient according to the embodiment of the present application includes a bed frame 1, both sides of which are provided with a blocking frame 2, the top end of the bed frame 1 is provided with a mattress 3, both ends of the bed frame 1 are respectively provided with a display control unit 4 and an infrared blood oxygen detector 5, the bottom end of the bed frame 1 is provided with a placing plate 7 through four groups of supporting legs 6, the top end of the placing plate 7 is provided with a feces collecting box 8, and the bottom end of the bed frame 1 is provided with a toilet auxiliary assembly 9. The display control unit 4 is used to receive the monitoring data of the patient and identify the state of the patient according to the monitoring data results, and adjust the control state of the nursing bed based on the needs of the patient.

[0026] In one embodiment, for the above-mentioned mattress 3, two groups of placing cavities 10 are opened in the inside of the mattress 3, one group of placing cavities 10 is provided with a plurality of groups of temperature sensors 11 in the inside, and the other group of placing cavities 10 is provided with a plurality of groups of air pumps 12 in the inside, and a toilet gap 13 is opened in the middle of the top end of the mattress 3; a pressure sensor 14 is arranged at the inside bottom end of the supporting leg 6, an excrement inlet 15 is opened at the top end of the feces collecting box 8, a collecting barrel 16 is arranged in the inside of the feces collecting box 8, a double-layer activated carbon filter screen 17 is arranged on the inner wall of the feces collecting box 8, and a camera 18 is arranged on one side of the infrared blood oxygen detector 5. Thus, the temperature sensors 11 in the mattress 3 can be used to measure the body temperature of the patient in real time and accurately, and the collected body temperature data can be transmitted to the display control unit 4 for health status detection. Meanwhile, the air pumps 12 can be automatically controlled to inflate and deflate different areas in turn according to the preset time interval, so as to relieve the support force of the surface of the mattress 3 on each part of the patient's body, avoid the local tissue from being pressed for a long time, effectively reduce the occurrence of pressure injury, and install high-precision pressure sensors 14 on the four supporting legs 6 of the nursing bed respectively, so that when the patient lies on the bed, the pressure sensors 14 measure the pressure borne by the bed body, calculate the body weight of the patient, and transmit it to the display control unit 4 for health status detection.

[0027] In one embodiment, for the above-mentioned toilet auxiliary assembly 9, the toilet auxiliary assembly 9 comprises two sets of electric push rods one 901 mounted on the top end of the bed frame 1, one side of the electric push rod one 901 is connected with a connecting rod 902, and the connecting rod 902 is connected with the bottom end of the bed frame 1; the outer side of the bed frame 1 is provided with a servo motor 903, the output shaft of the servo motor 903 is provided with a lead screw 904, one side of the lead screw 904 and at the bottom end of the bed frame 1 is provided with a limiting rod 905, the outer side of the lead screw 904 and the limiting rod 905 is provided with an adjuster 906; the adjuster 906 comprises a sliding frame 907 mounted on the outer side of the lead screw 904 and the limiting rod 905, the top end of the sliding frame 907 is provided with a sliding rail 908, one side of the sliding frame 907 is provided with an electric push rod two 909, one side of the electric push rod two 909 is connected with a sliding rod one 910, one side of the sliding rod one 910 is connected with a sliding rod two 911, and one side of the sliding rod one 910 is connected with the sliding rail 908, the top end of the sliding rod one 910 and the sliding rod two 911 is provided with a mattress filling block 912, and the size of the mattress filling block 912 is the same as the size of the toilet gap 13, so that the electric push rod one 901 can be started to push the bed frame to complete the comfortable toilet position adjustment when the patient needs, and the state of the mattress filling block is controlled by the electric push rod two 909, so that the patient can use the toilet conveniently, and the excrement directly falls into the inside of the excrement collecting box 8, realizes the collection of the excrement, and the excrement collecting box 8 is provided with a double-layer activated carbon filter screen 17, which can absorb odor in real time to avoid affecting the environment.

[0028] The working principle of the toilet assisting assembly is as follows: after the display control unit 4 receives the toilet demand of the patient, the electric push rod 1 is extended outward, so that the connecting rod 902 moves to the extension direction of the electric push rod 1, and in the process of movement, the back position of the bed frame 1 is lifted to 30° to 45° (which can be adjusted adaptively according to the actual situation), at the same time, the electric push rod 1 is retracted inward, so that the connecting rod 902 moves to the retraction direction of the electric push rod 1, and in the process of movement, the leg position of the bed frame 1 is lifted and bent to 15° to 30° (which can be adjusted adaptively according to the actual situation), and then the electric push rod 2 is retracted inward, in the process of retraction, the sliding rod 2 moves to the retraction direction of the electric push rod 2, so that the distance between the mattress patch block 912 and the bed frame 1 is increased, after the mattress patch block 912 moves away from the bed frame 1, the servo motor 903 moves the mattress patch block 912 to one side, so that there is no shelter at the bottom end of the toilet gap 13, and then the patient can go to the toilet, and the excrement directly falls into the bottom end of the collection box 8 under the bed, the upper layer of the collection box 8 is the position of the feces, and the lower layer is the urine, the collection box 8 is provided with a double-layer activated carbon filter screen 17, which can adsorb odor in real time (similar to the refrigerator deodorizing box), and after the end, the voice or button is restored again, and the original position is automatically restored, the mattress patch block 912 is supplemented to the bottom end of the toilet gap 13, and in order to ensure the use effect of the collection box 8, a box cover is arranged at the top end of the collection box 8, and the box cover can automatically control the opening and closing of the box cover through a mechanical device (the technology belongs to the prior art, and will not be described in detail here), so that the box cover is automatically sealed after the toilet is completed, preventing odor from overflowing.

[0029] As shown in Figure 8 The display control unit 4 comprises: The health assessment alarm module 401 is used for acquiring the monitoring results of the temperature sensor 11, the infrared blood oxygen detector 5 and the pressure sensor 14, judging the health state of the patient according to the detection results, and sending alarm information to the user end based on the state; The nursing bed use adjustment module 402 is used for controlling the use state of the air pump 12 and the toilet assisting assembly 9 according to the preset time and the patient demand; The emotional state evaluation module 403 is used for identifying the emotional state of the patient based on the monitoring video of the camera 18, dynamically matching the intervention content according to the identification result, and adjusting the emotion of the patient; The nutritional status evaluation module 404 is used for extracting the meal information of the patient, evaluating the nutritional intake status of the patient based on the meal information, combining with the medical record of the patient to judge the rationality of the patient's eating, and generating a voice prompt according to the rationality result; The control state data display module 405 is used for displaying the result data of the health assessment alarm module 401, the nursing bed use adjustment module 402, the emotional state evaluation module 403 and the nutritional status evaluation module 404.

[0030] Specifically, the health assessment alarm module 401, the nursing bed use adjustment module 402, the emotional state assessment module 403, the nutritional status assessment module 404 and the control state data display module 405 are connected in sequence.

[0031] It needs to be explained that in the health assessment alarm module 401, the temperature sensor 11 is uniformly embedded in the mattress 3, which can accurately measure the body temperature of the patient in real time, and the collected body temperature data can be transmitted to the control state data display module 405. When the body temperature data received by the control state data display module 405 exceeds the preset normal body temperature range (such as 37.5°C), the cooling device is immediately started, and at the same time the fever information and real-time body temperature data are sent to the family mobile phone through Bluetooth, reminding the family to pay attention.

[0032] The infrared blood oxygen detector 5 can monitor the blood oxygen of the patient in real time, and record in the control state data display module 405. The blood oxygen data measured by the infrared blood oxygen detector 5 is transmitted to the display screen of the nursing bed, which is recorded in real time on the display screen at the bedside, and the data is also stored on the mobile phone, which is convenient for subsequent viewing of historical data. When the blood oxygen saturation is lower than the preset normal range (such as 90%), the control system triggers the alarm device, issues a warning, and sends an alarm message to the family mobile phone.

[0033] High-precision pressure sensors 14 are installed on the four supporting legs 6 of the nursing bed. When the patient lies on the bed, the pressure sensor 14 measures the pressure borne by the bed body, calculates the weight of the patient, and displays it in real time on the display screen (the weight can be measured every few months), and automatically records the weight data of each measurement, forming a weight change curve. Through the built-in data analysis algorithm, combined with the defecation situation of the patient and other health indicators (such as body temperature, blood oxygen, etc.), the reason for the change of weight is dynamically analyzed.

[0034] The display screen at the bedside adopts a touch design, which is convenient and easy to operate. The display screen not only displays the patient's body temperature, blood oxygen, weight, and other health indicators in real time, but also can query historical data. The historical data is presented in the form of charts, such as body temperature change curve, blood oxygen trend chart, etc., which intuitively shows the changes of the patient's health status. In addition, it can automatically answer video chat. When any health indicator exceeds the preset normal range, the alarm device is triggered immediately, a loud sound alarm and flashing light prompt are issued, and at the same time the alarm information is sent to the family mobile phone through the Bluetooth module. The alarm information includes the specific abnormal indicator name, value and current time. The family can timely understand the health status of the patient and take corresponding measures. In addition, the family can also remotely view the real-time health data and historical records of the patient through the mobile phone APP, realizing remote attention to the health status of the patient.

[0035] When assessing a patient's health status based on test results and generating alarm information to send to the user terminal, the system can collect raw data from various sensors (including body temperature, blood oxygen saturation, etc.), preprocess the raw data (including noise reduction, filtering, outlier handling, unit standardization, etc.), and perform feature extraction and signal integration processing to form a multi-parameter state vector. It then combines medical knowledge or machine learning models to evaluate the current state and determine if any abnormalities exist (such as fever, low oxygen risk, etc.). Once the state is found to exceed the normal physiological range (such as body temperature exceeding 37.5°C, blood oxygen below 90%, etc.), the alarm logic is immediately triggered, and structured alarm information (including abnormality type, severity level, recommended measures, timestamp, etc.) is generated. This alarm information is then sent to the user terminal (such as nursing terminals, family apps, etc.) via push notification.

[0036] It should be explained that in the nursing bed adjustment module 402, after the patient presses the "toilet" button or makes a request to use the toilet via voice, the back is raised and the legs are bent, automatically switching to the toilet sitting mode (simulating the feeling of defecating on a toilet). When the position is adjusted, the excretion under the buttocks is automatically opened for toilet use.

[0037] The mattress 3 is set with multiple zones, each zone is connected to a smart air pump 12. According to a preset time interval (such as every 2 hours), the air pump 12 is automatically controlled to inflate and deflate different zones in sequence, relieving the support force of the mattress 3 surface on various parts of the patient's body, avoiding prolonged pressure on local tissues, and effectively reducing the occurrence of pressure injuries.

[0038] During the control process, when the patient actively requests to use the toilet, the system judges the current environmental status (such as the bed angle) according to the execution logic. After confirming that there is no abnormality, it issues a control command, which activates the toilet assistance component 9 (including the lifting and bending of the bed frame 1, the movement of the mattress filler block 912, etc.). After the task is completed, the device automatically returns to the initial state and records the operation log and patient feedback for subsequent optimization and health assessment.

[0039] like Figure 9 As shown, the emotional state assessment module 403 includes: The face keyframe output module 4031 is used to collect video data of the patient at regular intervals using the camera 18, extract keyframes from the video data at preset time intervals, and output the patient's face keyframes based on the face detection results in the keyframes. The face image determination module 4032 is used to generate face structure vertices based on superpixel blocks of patient face keyframes, and determine the patient face image region using clustering learning and Markov techniques based on the density information of the face structure vertices. The emotion state recognition module 4033 is configured to extract contour information in the patient face image region, generate a visual saliency map in combination with a visual saliency model, and recognize a variance distance between the contour information by using a second-order oscillation technology, and output an emotion state of the patient. The patient emotion regulation module 4034 is configured to establish a matching relationship between the emotion state and a local video tag of the nursing bed, determine a local video of the nursing bed corresponding to the emotion state of the patient based on the matching relationship, and extract the local video of the nursing bed to regulate the emotion of the patient.

[0040] The face key frame output module 4031, the face image determination module 4032, the emotion state recognition module 4033, and the patient emotion regulation module 4034 are sequentially connected.

[0041] It should be explained that, in the process of outputting the patient face key frame, the face key frame output module 4031 needs to start the camera 18 and record or capture the video stream at a set acquisition frequency in a timing manner, then extract a key frame image from the video stream according to a set time interval (for example, every 5 seconds), perform face recognition and positioning on each frame of image by using a recognition algorithm (specifically, MTCNN, YOLO-Face, or RetinaFace, etc. can be used), and determine whether the image contains a clear face region; if there is a face and the confidence is higher than a set threshold, the frame is marked as a "face key frame", and the face region is cropped or labeled, then the key frame is time-stamped, numbered, and saved in a structured data format (for example, JPEG+JSON), finally all the face key frames detected are output to a specified storage path for subsequent expression recognition and health assessment.

[0042] The face image determination module 4032 generates face structure vertices based on superpixel blocks of the patient's face keyframes, and determines the patient's face image region using clustering learning and Markov techniques based on the density information of the face structure vertices. It can uniformly distribute seed points within the patient's face keyframes, search for image pixels within a preset range of seed points based on the distribution results, and complete image seeding after the seed points coincide with the minimum gradient position within the preset range. It performs color conversion on the patient's face keyframes, calculates the color distance and spatial distance between image pixels and seed points based on the conversion results, generates distance metric values, and outputs superpixel blocks based on the distance metric values. Superpixel blocks are used to extract the contour distribution of facial keyframes. The extraction results are used to perform initial segmentation of the superpixel blocks, and the similarity between superpixel blocks within the facial keyframes is determined based on the segmentation results. Based on the similarity results, the density information of the vertices of each superpixel block is estimated, and the facial structure vertices are output according to the estimation results. Clustering learning technology is used to re-cluster the density information and combine it with the facial structure vertices to obtain a facial region connectivity map. The gradient operator of the facial region connectivity map is analyzed according to the Markov extraction technology. Based on the gradient operator, the gradient components of the patient's face points in different directions in the facial region connectivity map are obtained, and an oriented gradient histogram is generated to determine the patient's face image region.

[0043] The formula for calculating density information is as follows: ; In the formula, delta a Indicates the first a Density information estimation of the vertices of a superpixel block, H This indicates the total number of superpixel blocks. S b Indicates the first b The similarity between a superpixel block and its neighboring superpixel blocks. c Indicates the index of the superpixel block. D Represents the distance metric value, ε l,l+1 Represents superpixel block segmentation points l Segmentation points with adjacent superpixel blocks l Weights between +1, h This represents the bandwidth of the contour distribution.

[0044] Specifically, in determining the patient face image region, the image needs to be pre-processed, including image size normalization, denoising and light balance, to ensure the stability and accuracy of subsequent calculations; then use the superpixel segmentation algorithm (specifically, SLIC technology can be used) to preliminarily divide the image, generate dense and approximately uniform superpixel blocks, and record the basic features of each superpixel block, such as pixel composition, color mean, spatial coordinate mean, etc. After generating superpixels, seed points are arranged in the key frame using a uniform dot arrangement method. Seed points can be arranged based on fixed grid spacing, interest point detection (such as Harris corner points), or through pre-set face component regions (such as eyes, nose, contour regions). In the preset search range of each seed point (such as a neighborhood window with a radius of r If the minimum gradient position coincides with or is close to the seed point position, the point is taken as an effective image seed and its spatial position, color value and other information are saved. Image seed processing is completed, and the key frame image is color space transformed to enhance the separability of color distribution. In the converted color space, for each image pixel, the color distance (specifically, Euclidean distance or perceptual distance) and spatial distance (such as pixel coordinate difference) between it and the nearest seed point are calculated, and the two are weighted and fused to generate the distance measure value between each pixel and the seed point. According to the measure value, the pixels are attributed to the corresponding seed points, thereby generating updated superpixel block division. After obtaining the new superpixel blocks, edge detection and contour extraction (Canny or gradient amplitude-based method can be used) are performed on each superpixel block to determine the preliminary contour distribution of the face structure, and the spatial adjacency relationship and boundary continuity between superpixel blocks are determined through the connectivity of the contour line. Thus, the similarity between adjacent superpixel blocks is calculated based on their color, texture, and position features, and the superpixels are initially clustered and divided based on the similarity.

[0045] For each clustering region, the structural vertex density distribution thereof is estimated, i.e., the density and distribution characteristics of the superpixel block vertices in the image space are estimated, and specifically, a density map can be constructed using kernel density estimation (KDE), and more accurate facial structure vertices are determined through the high-density region in the density map; the vertices are input into a clustering learning model (specifically, K-Means), and the vertex density and geometric distribution characteristics calculated previously are combined to re-cluster the facial key regions, form a region connection graph describing the topology of the face, and on this basis, Markov extraction technology is applied, i.e., a Markov random field model (MRF) is constructed, the edges between the vertices and their adjacent regions in the connection graph are taken as state transition edges, the state transition probability between the vertices in the whole facial region in the image gradient space is analyzed, the gradient components of each point on the connection graph in different directions (such as horizontal, vertical, diagonal, etc.) are calculated, a directional gradient tensor is constructed, the gradient distribution in different directions is calculated according to the directional gradient components, and finally a histogram of oriented gradients (HOG-like structure) is generated. The histogram can depict the feature changes between the regions in the facial image region in terms of edge intensity and direction distribution, so that the mode distribution in the histogram of oriented gradients is used to determine and output the position and range of the patient's facial image region.

[0046] In the process of the emotion state recognition module 4033 extracting the contour information in the patient's facial image region, generating a visual saliency map in combination with a visual saliency model, and identifying the variance distance between the contour information by using a second-order oscillation technology to output the emotion state of the patient, the patient's facial image can be scaled, the patient's facial image can be adjusted to a desired size, and the coordinates of any pixel point in the adjusted patient's facial image can be obtained. The translation amount of the pixel point is determined according to the coordinates, and the first-order approximation of the translation amount is obtained by Taylor expansion; the eigenvalues of the first-order approximation of each position in the patient's facial image are obtained, and the sum and difference matrices are constructed according to the absolute values of the sum and difference of any eigenvalues, respectively. The sum and difference matrices are normalized, and the facial point saliency map and the contour edge saliency map are output based on the processing results; the facial point saliency map and the contour edge saliency map are linearly combined based on the visual saliency model to output the patient's facial visual saliency map, and the organ shape information of the patient is extracted from the patient's facial visual saliency map by using edge detection technology; the dynamic change trend of the organ shape information is captured by fitting an oscillation curve, the contour change amplitude of each organ is measured based on the dynamic change trend, and the structure variance feature is output by performing variance extraction processing on the contour change amplitude. The emotion state of the patient is output by matching the emotion label feature.

[0047] Specifically, in the process of outputting the emotional state of the patient, the original face image of the patient is first scaled, the image is adjusted to a uniform desired size (for example, 256x256 pixels) through affine transformation or bilinear interpolation algorithm, the scaling operation ensures that the subsequent feature extraction has consistency between different images, after the image is scaled, the coordinate position of each pixel point in the image is recalculated, and the translation amount of the pixel is obtained according to the difference between the pixel positions before and after scaling, the translation amount is first approximated by using Taylor expansion, the displacement response of the image at the position is obtained by calculating the partial derivative of the image function with respect to the coordinate, the first order approximation feature of each pixel position is further constructed, the first order approximation of the whole image is decomposed to obtain the local structure strength feature in the neighborhood of each pixel point, then for these eigenvalues, the absolute value of the sum of the eigenvalues and the absolute value of the difference between the eigenvalues of any two points in the image are calculated respectively, and a "sum matrix" and a "difference matrix" are constructed respectively, the two matrices respectively reflect the concentration degree and the contrast change degree of the features in the local region of the image.

[0048] To eliminate the influence of numerical scales, the sum matrix and the difference matrix need to be normalized. Then, the normalized sum matrix is used as the facial point saliency map, and the difference matrix is used as the contour edge saliency map. The former highlights the key point regions of facial organs such as eyes, nose, and mouth, and the latter emphasizes the organ contour and edge structure. On this basis, a visual saliency model (specifically, the Itti-Koch model) is introduced to linearly combine the two saliency maps, obtaining a comprehensive patient facial saliency image, called "facial visual saliency map", which clearly highlights the visual attention concentration area. Edge detection technology (specifically, Canny, Sobel, or Laplacian operator) is used to extract the edge and shape structure of facial organs in the facial visual saliency map, including key contour lines such as mouth corner curvature, eyelid opening and closing, and eyebrow trend. At the same time, the extracted organ contour curves are fitted, which can be modeled as oscillating curves in time series using sine function fitting, spline curve or Fourier descriptor. The curve is used to describe the dynamic deformation of the organ between different frames. By analyzing the periodicity and amplitude changes of these oscillating curves, the fluctuation range, local mutation points and continuous change trend are extracted, and the "contour change amplitude" of each organ is quantified. These contour amplitudes are used as basic data, and their statistical variance is extracted to calculate the distribution change degree in the time or space dimension, obtaining the corresponding structure variance features of each organ. The structure variance features contain the core information of facial micro-expression dynamic changes and can be used as structural representation indicators of emotional state. The extracted structure variance features are matched with pre-trained emotional label features, and a classification model (such as SVM, KNN, random forest or deep neural network) can be used to determine which emotional category the current features belong to, such as happy, sad, angry, surprised, etc. Finally, the emotional state of the patient at the current time is output as the final result of this process.

[0049] The patient emotion regulation module 4034 establishes a matching relationship between the emotional state and the local video tag of the nursing bed, determines the local video of the nursing bed corresponding to the patient's emotional state based on the matching relationship, and extracts the local video of the nursing bed to regulate the patient's emotion. In the process, the current emotional state of the patient is obtained through the emotion recognition result, such as "sadness", "anger", "happiness", "irritation", etc., and the label is structured. In the local video resource library of the nursing bed, the content label, emotional guidance direction, playing scene and physiological regulation tendency of each video are preset. These labels need to be manually annotated by mental health experts or nursing staff or automatically labeled by emotion classification algorithm during deployment. For example, a video playing natural scenery and light music is labeled as "relaxation-calm-suitable for anxiety", and a video playing a lively and interactive animation is labeled as "active-positive-suitable for depression". Then a matching mapping table or label matching model is constructed. The emotional label recognized by the patient in real time is used as a query condition to search for the video resource with the highest matching degree in the local video label library. The matching process can be determined by keyword matching and label vector similarity (such as calculating the cosine similarity of word vectors). The result is sorted in priority according to the effectiveness of emotional intervention; after the matching is completed, the corresponding video segment is scheduled and played on the patient display screen, and whether the emotional state is improved before and after the intervention is analyzed after the playing is finished, so as to record the regulation effect of the video on the specific emotional label, thereby realizing personalized and intelligent emotional regulation video service, and effectively assisting nursing staff to complete psychological care for patients.

[0050] As shown in Figure 10 , the nutritional status evaluation module 404 includes: The meal information extraction module 4041 is configured to extract the patient's meal food image from the monitoring video of the camera 18, and obtain the patient's dining behavior in the meal food image based on image recognition technology. The evaluation index construction module 4042 is configured to generate a comprehensive evaluation index system according to the patient's medical record data, and analyze the membership degree of the comprehensive evaluation index system by applying a fuzzy hierarchical algorithm to determine the patient's intake rationality standard. The diet rationality judgment module 4043 is configured to quantify the patient's intake nutrition structure and diet behavior according to the dining behavior, and match the quantization result with the rationality standard to output the patient's eating rationality judgment result. The voice prompt generation module 4044 is configured to adaptively match the corresponding diet suggestion content according to the rationality judgment result, and transmit the diet suggestion content to the medical staff end through wireless technology for patient diet intervention.

[0051] The meal information extraction module 4041, the evaluation index construction module 4042, the diet rationality judgment module 4043 and the voice prompt generation module 4044 are sequentially connected.

[0052] It should be explained that the meal information extraction module 4041 starts the camera to collect real-time video in the process of obtaining the dining behavior of the patient in the meal food image, sets a fixed time period or triggers the detection mechanism to identify the starting moment of the meal scene through the meal action, extracts the meal-related video segment from the video stream and performs frame sampling or key frame extraction to obtain the image sequence, then uses the object detection algorithm (specifically, YOLOv5, FasterR-CNN or EfficientDet technology can be used) to identify the food types and quantities in the image in each frame of image, combines the food classification model to classify and label the food in the dinner plate, and simultaneously introduces the human pose estimation algorithm in the image to obtain the skeletal key points of the patient, further judges the spatial relationship between the head, hand, elbow, etc. and the tableware or food, identifies whether there are typical meal action features such as "picking up food", "feeding", "chewing", etc., and analyzes the continuity and periodicity of these actions in the time dimension through the action recognition model (specifically, TCN, LSTM or Transformer based on time convolution can be used), extracts the dining behavior sequence, then fuses the food recognition information and the patient action data, constructs a time behavior joint feature vector, discriminates the dining state of the patient in different time periods, including whether to eat, food intake, action frequency, food intake type, etc., and finally stores these behavior features.

[0053] The evaluation index construction module 4042 can generate a comprehensive evaluation index system according to the medical record of the patient, analyze the membership degree of the comprehensive evaluation index system by applying a fuzzy hierarchical algorithm, determine the patient's intake rationality standard, based on the medical record and disease state of the patient, analyze the patient's nutritional structure and eating frequency to construct a comprehensive eating evaluation index set, and determine the importance of each index by a subjective weighting method; according to the importance, establish the grade comments of each index, and based on the trapezoidal membership function, obtain the membership values of each index under each grade, and analyze the fuzzy membership of each grade using the membership values; based on the membership values and fuzzy membership, construct a fuzzy rule matrix, multiply the fuzzy rule matrix and the weight vector to determine the comprehensive rule result of the nutritional structure and eating frequency, and output the patient's intake rationality standard.

[0054] It needs to be explained that in the process of realizing the rationality of the patient's diet, structured information needs to be extracted from the patient's electronic medical record, including basic information (age, gender, weight, height, BMI), diagnosed diseases (such as diabetes, hypertension, kidney disease, etc.), treatment stages (preoperative, postoperative, rehabilitation period) and medical order information, etc. According to the medical record information and the current disease state, the patient's nutritional structure and eating frequency requirements are constructed according to the medical nutrition standard, and a comprehensive eating evaluation index set composed of multiple dimensions is formed, which contains indexes such as total energy demand, carbohydrate intake ratio, protein intake, fat restriction, dietary fiber amount, eating frequency, average intake per meal, and night eating frequency., etc., and the importance of the index is scored and normalized to obtain a weight vector by applying a subjective weighting method.

[0055] Based on the characteristics of each index, a grading standard is set, for example, four levels of "excellent, good, medium, and poor", and a corresponding trapezoidal membership function is set for each level to describe the fuzzy membership of each index value at different levels. When a certain index is quantified, its membership value at each level can be calculated through the trapezoidal membership function, and then the grade membership of all indexes is composed into a fuzzy evaluation matrix, and on this basis, the fuzzy evaluation matrix and the weight vector are multiplied to obtain the comprehensive fuzzy evaluation value of each level. The rationality standard judgment level of the patient's nutritional structure and eating frequency is output through the maximum membership principle or weighted average method; at the same time, using front-end image recognition and behavior analysis technology, based on the video extraction of patients' dining behavior and quantification, including daily intake of food types, nutritional ingredients (identified by image recognition and searched in the nutrition database), total calorie intake, eating frequency, meal duration, food chewing action frequency and other behavior characteristics, which are mapped into quantified index values consistent with the evaluation index system, and these quantification results are matched with the previously obtained intake rationality standard. Analyze the deviation of the grade membership of each actual value and standard value, and then obtain the comprehensive eating rationality judgment result, including whether the intake is excessive, whether the structure is unbalanced, and whether the eating frequency is abnormal.

[0056] Based on the judgment result, adaptive matching of corresponding suggestion content is carried out, such as controlling oil intake, increasing dietary fiber, and suggesting fixed time and quantity, etc., and a recommended recipe or behavior adjustment suggestion adapted to the patient's taste and health status is automatically generated according to the suggestion category. Finally, through wireless communication technology (such as WiFi, Bluetooth, NB-IoT), the diet suggestion content is sent to the user end in a structured data or message push way, so that the patient's diet status can be understood in real time and intervention guidance can be carried out, thereby forming a closed-loop identification evaluation matching feedback diet intervention process.

[0057] In the present application, unless otherwise explicitly specified and limited, the terms "mounting", "setting", "connecting", "fixing", "screwing" and other terms should be understood in a broad sense, for example, can be fixedly connected, or can be detachably connected, or can be integrated; can be mechanically connected, or can be electrically connected; can be directly connected, or can be indirectly connected through an intermediate medium; can be internal communication of two elements or interaction relationship between two elements, unless otherwise explicitly limited, the above-mentioned terms in the present application can be understood according to the specific meaning of the above-mentioned terms in the present application by those skilled in the art.

[0058] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A nursing bed capable of monitoring the health state of a patient, comprising a bed frame (1), the two sides of the bed frame (1) being provided with a blocking frame (2), characterized in that, The top end of the bed frame (1) is provided with a mattress (3), and the two ends of the bed frame (1) are respectively provided with a display control unit (4) and an infrared blood oxygen detector (5); The bottom end of the bed frame (1) is provided with a placing plate (7) through four groups of supporting legs (6), the top end of the placing plate (7) is provided with a feces collecting box (8), and the bottom end of the bed frame (1) is provided with a defecation assisting assembly (9); The display control unit (4) is used for receiving display patient monitoring data and identifying the state of the patient according to the monitoring data results, and adjusting the control state of the nursing bed based on the patient's needs.

2. A care bed according to claim 1, wherein, The inside of the mattress (3) is provided with two groups of placing cavities (10), one group of the placing cavities (10) is provided with a plurality of temperature sensors (11), and the other group of the placing cavities (10) is provided with a plurality of air pumps (12), and the top middle part of the mattress (3) is provided with a defecation gap (13); The inside bottom end of the supporting leg (6) is provided with a pressure sensor (14), the top end of the feces collecting box (8) is provided with a feces inlet (15), the inside of the feces collecting box (8) is provided with a collecting barrel (16), the inner wall of the feces collecting box (8) is provided with a double-layer activated carbon filter screen (17), and one side of the infrared blood oxygen detector (5) is provided with a camera (18).

3. A care bed according to claim 2, wherein, The defecation assisting assembly (9) comprises two groups of electric push rods (901) which are installed at the top end of the bed frame (1) and have opposite directions, one side of the electric push rod (901) is connected with a connecting rod (902), and the connecting rod (902) is connected with the bottom end of the bed frame (1); The outside of the bed frame (1) is provided with a servo motor (903), the output shaft of the servo motor (903) is provided with a lead screw (904), one side of the lead screw (904) and located at the bottom end of the bed frame (1) is provided with a limiting rod (905), and the outside of the lead screw (904) and the limiting rod (905) is provided with an adjuster (906); The adjuster (906) comprises a sliding frame (907) which is installed at the outside of the lead screw (904) and the limiting rod (905), the top end of the sliding frame (907) is provided with a sliding rail (908), one side of the sliding frame (907) is provided with an electric push rod (909), one side of the electric push rod (909) is connected with a sliding rod (910), one side of the sliding rod (910) is connected with a sliding rod (911), one side of the sliding rod (910) is connected with the sliding rail (908), the top end of the sliding rod (910) and the sliding rod (911) is provided with a mattress gap filling block (912), and the size of the mattress gap filling block (912) is same with the size of the defecation gap (13).

4. The care bed of claim 2, wherein, The display control unit (4) comprises: The health assessment alarm module (401) is configured to acquire monitoring results of the temperature sensor (11), the infrared blood oxygen detector (5), and the pressure sensor (14), judge the health state of the patient according to the detection results, and generate alarm information based on the state and send the alarm information to the user end; The nursing bed use adjustment module (402) is configured to control the use state of the air pump (12) and the toilet auxiliary assembly (9) according to a preset time and the patient demand; The emotion state evaluation module (403) is configured to identify the emotion state of the patient based on the monitoring video of the camera (18), dynamically match the intervention content according to the identification result, and adjust the emotion of the patient; The nutrition state evaluation module (404) is configured to extract the meal information of the patient, evaluate the nutrition intake state of the patient based on the meal information, judge the rationality of the meal of the patient in combination with the medical record of the patient, and generate a voice prompt according to the rationality result; The control state data display module (405) is configured to display the result data of the health assessment alarm module (401), the nursing bed use adjustment module (402), the emotion state evaluation module (403), and the nutrition state evaluation module (404).

5. A care bed according to claim 4, wherein, The emotion state evaluation module (403) includes: The face key frame output module (4031) is configured to acquire video data of the patient by using the camera (18) at a regular time, extract a key frame from the video data at a preset time interval, and output a patient face key frame according to a face detection result in the key frame; The face image determination module (4032) is configured to generate a face structure vertex based on a superpixel block of the patient face key frame, and determine a patient face image region by using clustering learning and Markov technology according to density information of the face structure vertex; The emotion state recognition module (4033) is configured to extract contour information in the patient face image region, generate a visual saliency map in combination with a visual saliency model, and output the emotion state of the patient by using a second-order oscillation technology to identify a variance distance between the contour information; The patient emotion adjustment module (4034) is configured to establish a matching relationship between the emotion state and a local video tag of the nursing bed, determine a local video of the nursing bed corresponding to the emotion state of the patient based on the matching relationship, and extract the local video of the nursing bed to adjust the emotion of the patient.

6. A care bed according to claim 5, wherein, The generation of the face structure vertex based on the superpixel block of the patient face key frame and the determination of the patient face image region according to the density information of the face structure vertex by using the clustering learning and Markov technology include: Uniformly arranging seed points in the patient face key frame, searching for image pixel points in a preset range of the seed points according to the arrangement result, and completing image seed processing after the seed points coincide with the minimum gradient position in the preset range; Performing color conversion on the patient face key frame, calculating color distance and spatial distance between the image pixel points and the seed points according to the conversion result, generating a distance measure value, and outputting a superpixel block based on the distance measure value; According to the superpixel block, the contour distribution of the face key frame is extracted, the superpixel block is segmented based on the extraction result, and the similarity between the superpixel blocks in the face key frame is judged based on the segmentation result; Based on the similarity result, the density information of the vertex of each superpixel block is estimated, and the face structure vertex is output according to the estimation result, and the face region connection graph is obtained by re-clustering the density information combined with the face structure vertex by using clustering learning technology; According to the Markov extraction technology, the gradient operator of the face region connection graph is analyzed, the gradient components of the patient face points in the face region connection graph in different directions are obtained based on the gradient operator, and the direction gradient histogram is generated to determine the patient face image region.

7. A care bed according to claim 6, wherein, The calculation formula of the density information is: ； wherein, The contour information in the patient face image region is extracted, a visual saliency map is generated by combining a visual saliency model, and the variance distance between the contour information is identified by using a second-order oscillation technology, and the emotional state of the patient is output. a denotes the density information estimate of the a th superpixel block vertex, H denotes the total number of superpixel blocks, S b denotes the similarity between the b th superpixel block and its neighboring superpixel blocks, c denotes the index of the superpixel block, D denotes the distance measure value, ε l,l+1 denotes the weight between the l th superpixel block split point and the neighboring superpixel block split point l + 1, h denotes the bandwidth of the profile distribution.

8. A care bed according to claim 7, wherein, The patient face image is scaled, the patient face image is adjusted to the desired size, and the coordinates of any pixel point in the adjusted patient face image are obtained, the translation amount of the pixel point is judged according to the coordinates, and the first-order approximation of the translation amount is obtained by Taylor expansion; The eigenvalues of the first-order approximation of each position in the patient face image are obtained, and the sum and difference matrices are constructed based on the sum and difference of any eigenvalue, and the sum and difference matrices are normalized, and the face point saliency map and the contour edge saliency map are output based on the processing result; The face point saliency map and the contour edge saliency map are linearly combined based on the visual saliency model to output the patient face visual saliency map, and the organ shape information of the patient is extracted from the patient face visual saliency map by using edge detection technology; The dynamic change trend of the organ shape information is captured by fitting the oscillation curve, the contour change amplitude of each organ is measured based on the dynamic change trend, and the structure variance feature is output by implementing variance extraction processing on the contour change amplitude. The nutritional status evaluation module (404) includes:

9. The care bed of claim 4, wherein, The meal information extraction module (4041) is used to extract the patient's meal food image from the monitoring video of the camera (18), and to obtain the dining behavior of the patient in the meal food image based on image recognition technology; The evaluation index construction module (4042) is used to generate a comprehensive evaluation index system according to the patient's medical record, and to analyze the membership degree of the comprehensive evaluation index system by applying a fuzzy hierarchical algorithm to determine the patient's intake rationality standard; The diet rationality judgment module (4043) is used to quantify the patient's intake nutrition structure and eating behavior according to the dining behavior, and to match the quantization result with the rationality standard to output the patient's eating rationality judgment result; The voice prompt generation module (4044) is used to adaptively match the corresponding diet suggestion content according to the rationality judgment result, and to transmit the diet suggestion content to the medical staff terminal through wireless technology for patient diet intervention. ​ 10. A care bed according to claim 9, wherein, The comprehensive evaluation index system is generated according to the medical record of the patient, and the membership degree of the comprehensive evaluation index system is analyzed by applying a fuzzy hierarchical algorithm to determine the patient's intake rationality standard, which comprises: Based on the medical record of the patient and the disease state, the nutritional structure and eating frequency that the patient should intake are analyzed to construct a comprehensive eating evaluation index set, and the importance of each index is judged by a subjective weighting method; According to the importance, the grade comments of each index are established, and the membership degree values of each index under each grade are obtained based on a trapezoidal membership function, and the fuzzy membership degree of each grade is analyzed by using the membership degree values; Based on the membership degree values and the fuzzy membership degree, a fuzzy rule matrix is constructed, and the comprehensive rule result of the nutritional structure and eating frequency is determined by multiplying the fuzzy rule matrix and the weight vector to output the patient's intake rationality standard.

Images