Multifunctional nursing equipment suitable for aging and based on multi-modal monitoring and nursing method

Abstract

The invention belongs to the technical field of medical instruments and intelligent nursing, and discloses multifunctional nursing equipment suitable for aging and based on multi-modal monitoring and a nursing method.The nursing equipment comprises a nursing bed body, a multi-modal monitoring unit, a nursing supporting unit, an excretion supporting unit and an intelligent control unit, and the multi-modal monitoring unit, the nursing supporting unit, the excretion supporting unit and the intelligent control unit are arranged on the nursing bed body; a complete closed loop of monitoring-prediction-control-execution is constructed, the multi-mode monitoring unit comprehensively collects physiological-behavior-environment data, the intelligent control unit prejudges nursing requirements through a pre-trained AI model of a front-end computer, and a PLC drives all the units to execute nursing actions such as body position adjustment, pressure sore prevention and excretion without leaving a bed. Safety protection and remote cooperation functions are matched. The problems of few monitoring and nursing functions, response lag and the like of existing equipment are solved, the nursing accuracy and safety are improved, the nursing burden is reduced, and the system is suitable for nursing scenes such as hospitals, old-age nursing institutions and families.

Description

Technical Field

[0001] This invention belongs to the field of medical equipment and intelligent nursing technology, specifically relating to an age-friendly multifunctional nursing device and method based on multimodal monitoring. It is applicable to the accurate prediction and execution of nursing needs for the elderly and disabled patients in scenarios such as hospitals, rehabilitation and elderly care institutions, providing comprehensive nursing assistance and decision support for medical staff. Background Technology

[0002] With the improvement of social medical standards and the deepening of population aging, the demand for long-term care for some patients and elderly people, especially those who are immobile, disabled, or semi-disabled, is increasing. Traditional care methods mainly rely on manual observation and experience accumulation by medical staff. This method not only consumes a lot of manpower, but may also lead to problems such as untimely and inaccurate care and low efficiency due to human negligence.

[0003] In hospitals, rehabilitation centers, and elderly care facilities, medical staff often have to care for multiple disabled elderly people simultaneously, resulting in a heavy workload and making it difficult to monitor the specific condition of each individual in real time. When an elderly person experiences an emergency need for care, failure to detect and address it promptly can lead to serious consequences such as pressure sores, falls from bed, and deterioration of their condition. Therefore, there is an urgent need for a method and system that can comprehensively and in real-time collect relevant data on disabled elderly people and accurately predict their care needs, in order to improve the efficiency and quality of care and safeguard the health and lives of disabled elderly individuals.

[0004] Existing nursing equipment and methods of this kind, such as the method for controlling a nursing bed to perform automatic nursing disclosed in CN105534654A, and the device adaptive control method and system based on multimodal perception disclosed in CN120949582A, can solve some problems, but still have the following technical issues: 1. Complex structure and limited monitoring functions: Most nursing devices have very complex structures and are expensive. They often use a single pressure sensor or a simple physiological monitoring module, which cannot comprehensively acquire user information, physiological status, behavioral characteristics and environmental information. This leads to a one-sided judgment of nursing needs and can only achieve basic nursing functions such as body position adjustment. They lack the ability to monitor the multi-dimensional physiological parameters and behavioral status of the elderly in real time and provide multi-type nursing care.

[0005] 2. Complex computation and low level of intelligence: It lacks effective data fusion and analysis algorithms, nursing needs judgment relies on human experience, nursing actions mostly rely on manual triggering, it cannot achieve advance prediction and automatic execution of nursing needs, and lacks intelligent prediction and decision-making capabilities based on data analysis; 3. Disconnection in nursing execution: Insufficient coordination between different parts; lack of organic coordination between the nursing bed structure and various functional modules; resulting in fewer nursing actions and less precise execution control; fixed mattress support, unable to dynamically adapt to changes in user weight and position; manual assistance required for excretion care, cumbersome operation and poor privacy. 4. Inadequate safety protection: It lacks the ability to respond quickly to abnormal situations, cannot effectively prevent safety risks such as falls from bed and pressure sores, and is also difficult to deal with emergencies and prevent nursing risks; 5. Poor personalization: The nursing parameters are set in a fixed manner and cannot be dynamically adjusted according to the weight, position and health status of different users, which affects the nursing effect and user comfort, and also makes it impossible to provide customized nursing services according to the individual differences of different elderly people.

[0006] 6. Weak remote collaboration: Poor information exchange with off-site medical staff makes it impossible to remotely monitor and intervene in the nursing process, increasing the workload of on-site nursing staff. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the existing technology and provide an age-friendly multifunctional nursing system and method based on multimodal monitoring. Based on the physical carrier of the nursing bed, it integrates multimodal sensing, intelligent algorithms and automatic actuators to construct a complete closed loop of monitoring-prediction-control-execution, thus solving the above problems from multiple aspects.

[0008] To achieve the above objectives, the present invention provides the following technical solution: An age-friendly multifunctional nursing device based on multimodal monitoring is characterized by comprising a nursing bed body and a multimodal monitoring unit, a nursing support unit, an excretion support unit, and an intelligent control unit disposed on the nursing bed body; The nursing bed body includes a headboard, a footboard, sideboards, a horizontal bed panel, and an internal support frame, which provide a physical carrier for the multimodal monitoring unit, nursing support unit, excretion support unit, and intelligent control unit. The multimodal monitoring unit includes an audio and video monitoring camera and an ultrasonic radar, which are vertically mounted on the nursing bed body via an electric telescopic rod. It is used to collect user physiological parameters, behavioral data and environmental data in real time, and transmit them to the intelligent control unit through dual-mode communication. The nursing support unit includes multiple air-floating mattresses that are horizontally set on the bed surface of the nursing bed body and spliced ​​together, used to perform nursing actions such as body position adjustment, airbag pressure adaptation, and assisting in excretion support. The excretion support unit includes a smart toilet, which enables users or patients to defecate without leaving the bed. The intelligent control unit includes a PLC controller and a front-end computer. The front-end computer has a built-in intelligent control program that is electrically or wirelessly connected to the multimodal monitoring unit, nursing support unit, and excretion support unit. It is used to receive and analyze multi-source data on user physiology, behavior, and environment collected by the multimodal monitoring unit, process the data, control the nursing support unit to complete adaptive nursing actions, control the excretion support unit to complete excretion support actions, and reset the system after the actions are completed, forming a hardware and software combined monitoring-prediction-control-execution collaborative architecture.

[0009] A multi-modal monitoring-based age-friendly multifunctional nursing method, applied to the aforementioned nursing equipment, includes the following steps: S1. Equipment initialization: Move the nursing bed to the designated position and lock the silent casters. After powering on, the intelligent control unit drives the multimodal monitoring unit, nursing support unit, and excretion support unit to complete self-tests. The electric telescopic rod is adjusted to the working posture, and the ultrasonic radar sets the safety threshold of the bed edge. S2. Multi-source data acquisition: The multi-modal monitoring unit collects user physiological parameters, behavioral data and environmental data in real time, and transmits them to the intelligent control unit through dual-mode encrypted communication; S3. Data Preprocessing: The intelligent control unit cleans, standardizes, and merges the collected data, removing noise and outliers to form a user status dataset in a unified format. S4. Nursing Demand Prediction: The front-end computer has a built-in or calls the intelligent control program (STAP-Net model and pre-trained AI model) in the remote network server to analyze the dataset and output the nursing demand type and urgency level for the next 5-30 minutes. S5. Nursing Action Execution: Based on the prediction results, the intelligent control unit sends instructions to the PLC controller to control the nursing support unit to perform body position adjustment and airbag pressure adaptation, or to control the excretion support unit to perform excretion support actions. S6. Reset and Feedback: After the nursing action is completed, the PLC controller controls each actuator to reset, and the front-end computer synchronizes the execution results and real-time data to the remote network server and remote nursing terminal to update the model parameters.

[0010] The beneficial effects of the present invention include at least the following aspects: 1. This invention, through an integrated design of nursing bed, sensor, control unit, intelligent algorithm, and actuator, and through a closed-loop design of multimodal monitoring, intelligent prediction, and automatic execution, constructs a complete nursing system from monitoring and judgment to execution. It can realize intelligent and integrated nursing care throughout the entire process from data collection and demand prediction to nursing execution, significantly reducing the number of medical staff and workload required, and can solve the problems of intelligent and long-term care for the elderly and patients from multiple aspects.

[0011] 2. Comprehensive and accurate multimodal monitoring: Integrating multiple sensors such as bio-radar and pressure sensors, it achieves full coverage of physiological, behavioral, and environmental data, improving monitoring accuracy by 40% compared to traditional equipment; 3. Intelligent prediction and proactive intervention: Through pre-trained AI models deployed on remote servers, nursing needs can be predicted 5-30 minutes in advance, with an emergency response time of ≤0.8 seconds, solving the problem of delayed nursing care; 4. Humanized excretion care: The patient does not need to leave the bed. Through the linkage of the sliding bed panel, support board and foot pedal, patients who are not completely disabled can complete the excretion care themselves, which greatly reduces the workload of on-site caregivers and reduces the exposure of the patient's privacy. 5. Efficient and convenient remote collaboration: Supports remote monitoring and intervention, significantly improving the work efficiency of nursing staff by 45.3%, and increasing the number of manageable users for on-site medical staff from an average of 6 to more than 10. Attached Figure Description

[0012] Figure 1 is a schematic diagram of the overall structure of the nursing equipment according to an embodiment of the present invention; Figure 2 is a schematic diagram of the overall three-dimensional shape of the nursing device according to an embodiment of the present invention (the smart toilet is in working state, while the multimodal monitoring unit, electric lifting tray and electric telescopic pedal are in non-working state). Figure 3 is another overall three-dimensional structural diagram of the nursing device according to an embodiment of the present invention (the smart toilet, electric lifting tray and electric telescopic pedal are all in working state, and the multimodal monitoring unit is in non-working state). Figure 4 is another overall three-dimensional structural diagram of the nursing device according to an embodiment of the present invention (the multimodal monitoring unit is in working state, while the smart toilet, electric lifting tray and electric telescopic pedal are in non-working state). Figure 5 is another overall three-dimensional structural diagram of the nursing equipment according to an embodiment of the present invention (the multimodal monitoring unit, smart toilet, electric lifting tray and electric telescopic pedal are all in working state). Figure 6 is a three-dimensional structural diagram of the electric telescopic rod in an embodiment of the present invention; Figure 7 is a top view of the nursing device in an embodiment of the present invention; Figure 8 is a schematic diagram of the front view structure of the nursing device in an embodiment of the present invention; Figure 9 is a schematic diagram of the left-side structure of the nursing device in an embodiment of the present invention; Figure 10 is a schematic diagram of the right-side structure of the nursing device in an embodiment of the present invention; Figure 11 is a top view of the nursing support unit in an embodiment of the present invention; Figure 12 is a top sectional view of the air-floating mattress in an embodiment of the present invention.

[0013] In the picture: 1. Nursing bed body; 11. Headboard; 12. Footboard; 121. Smart toilet inlet / outlet; 13. Side panel; 14. Horizontal bed panel; 141. Head level bed panel; 142. Lumbar level bed panel; 143. Foot level bed panel; 1431. Fixed foot level bed panel; 1432. Sliding foot level bed panel; 144. Receiving groove; 15. Casters; 16. Vertical mounting holes; 2. Excretion support unit; 21. Smart toilet; 22. Electric lifting tray; 23. Electric retractable footrest; 3. Nursing support unit; 31. Air-floating mattress; 311. Front air-floating mattress; 312. Rear fixed air-floating mattress; 313. Rear sliding air-floating mattress; 314. Base support layer; 315. Inflatable airbag adjustment layer; 316. Pressure sensor array layer; 317. Sealed skin-friendly surface layer; 318. Inlet and outlet pipes; 319. Air pump; 320. Airbag bag; 4. Intelligent control unit; 5. Multimodal monitoring unit; 51. Electric telescopic pole; 52. Ultrasonic radar; 53. Audio monitoring camera. Detailed Implementation

[0014] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0015] Basic Implementation like Figure 1 As shown, the multi-modal monitoring-based age-friendly multifunctional nursing device provided in this embodiment includes a nursing bed body 1 and a multi-modal monitoring unit 5, a nursing support unit 3, an excretion support unit 2 and an intelligent control unit 4 installed on the nursing bed body 1; The nursing bed body 1 includes a headboard panel 11, a footboard panel 12, a sideboard panel 13, a horizontal bed panel 14, and an internal support frame (not shown), which provides a physical carrier for the multimodal monitoring unit 5, the nursing support unit 3, the excretion support unit 2, and the intelligent control unit 4, forming an integrated design scheme that is easy to manufacture, use, and maintain. The overall architecture of the nursing equipment adopts a layered structure of "bed frame body + functional modules". The main body of the bed frame is made of high-strength aluminum alloy with a load-bearing capacity of 250kg. It achieves corrosion resistance and wear resistance through electrostatic spraying, making it suitable for long-term, high-frequency use. The bottom of the bed frame (legs) is equipped with silent universal casters 15, with a wheel diameter of 12cm and a braking device, which facilitates the movement and adjustment of the nursing bed body 1 while ensuring stability during use.

[0016] The multimodal monitoring unit 5 includes an audio and video monitoring camera 53 and an ultrasonic radar 52, which are vertically installed on the nursing bed body 1 via an electric telescopic rod 51. It is used to collect user physiological parameters, behavioral data and environmental data in real time, and transmit them to the intelligent control unit 4 through dual-mode communication for analysis and processing. The nursing support unit 3 includes multiple air-floating mattresses 31 that are horizontally set on the horizontal bed surface of the nursing bed body 1 and spliced ​​together, for performing nursing actions such as body position adjustment, airbag pressure adaptation, and assisting in excretion support. The excretion support unit 2 includes a smart toilet 21, which is used to enable users or patients to defecate without leaving the bed. The intelligent control unit 4 includes a PLC controller and a front-end computer. The front-end computer has a built-in intelligent control program (STAP-Net model + pre-trained AI model), which is electrically or wirelessly connected to the multimodal monitoring unit 5, nursing support unit 3, and excretion support unit 2, respectively. It is used to receive and analyze multi-source data on user physiology, behavior, and environment collected by the multimodal monitoring unit, process the data, control the nursing support unit to complete adaptive nursing actions, control the excretion support unit to complete excretion support actions, and reset the system after the actions are completed, forming a hardware and software combined monitoring-prediction-control-execution collaborative architecture.

[0017] The horizontal bed panel 14 of the nursing bed body 1 includes a head horizontal bed panel 141, a lumbar horizontal bed panel 142, and a foot horizontal bed panel 143; the foot horizontal bed panel 143 includes a foot horizontal fixed bed panel 1431 and a foot horizontal sliding bed panel 1432; a receiving groove 144 is provided between the foot horizontal sliding bed panel 1432, the foot horizontal fixed bed panel 1431, and the lumbar horizontal bed panel 142, and a smart toilet 21 is provided in the receiving groove 144. The opening of the smart toilet faces upward, and its axis is perpendicular to the axis of the horizontal bed panel 14.

[0018] The multimodal monitoring unit also includes a bio-radar array, a flexible pressure sensing grid, and a noise-suppressed speech module. The bio-radar array adopts a MIMO architecture, is deployed at the four corners of the horizontal bed surface, emits millimeter waves, and has a detection range of 0.5-1.5m, used to collect heart rate, respiratory rate, and micro-motion characteristics. The flexible pressure sensing grid is used to generate dynamic pressure heat maps. The noise-suppressed speech module adopts a dual-microphone array and a blind source separation algorithm.

[0019] The electric telescopic rod includes a fixed lower section and a telescopic upper section. The lower section is vertically fixed in the vertical mounting hole of the bed foot panel, and the upper section extends and retracts vertically. In the working state, the upper section extends upward, so that the working center of the audio-visual monitoring camera and the ultrasonic radar forms an angle of 30°~45° with the horizontal bed surface. In the non-working state, the upper section retracts into the lower section, and the upper end face of the ultrasonic radar is flush with or slightly lower than the bed foot panel. The audio-visual monitoring camera is equipped with a 2-megapixel CMOS sensor and supports infrared night vision. The ultrasonic radar has a detection distance of 0.3-5m, and a safe distance threshold of 30-50cm from the bed edge is set.

[0020] The air-supported mattress 31 of the nursing support unit further includes a base support layer, an inflatable airbag adjustment layer, a pressure sensor array layer, a sealed skin-friendly surface layer, and an air pump assembly; each layer is stacked sequentially from bottom to top and the edges are sealed and fixed by a heat-sealing process; the base support layer is a double-layer composite sponge structure, with a thick memory foam upper layer and a thick high-density sponge lower layer, covering the entire horizontal bed panel; the inflatable airbag adjustment layer is composed of rows and columns of independent airbag modules, each airbag module has a built-in elastic airbag bag, which is connected to the electromagnetic switch valve of the air pump assembly one by one through branch inlet and outlet pipes; the pressure sensor array layer is distributed in a mesh and is evenly embedded between the inflatable airbag adjustment layer and the sealed skin-friendly surface layer.

[0021] The pressure sensor array layer integrates an optical heart rate sensor, a bioelectrical impedance sensor, and a temperature and humidity sensor; the optical heart rate sensor is used to monitor heart rate, the bioelectrical impedance sensor is used to monitor respiratory rate, and the temperature and humidity sensor monitors temperature and humidity; the air pump assembly includes an air pump, a main intake and exhaust pipe, and multiple electromagnetic switching valves, which are integrated and installed in the bed support accommodating area below the horizontal bed panel, and are connected to the branch intake and exhaust pipes of each airbag module through the main intake and exhaust pipe.

[0022] A multi-modal monitoring-based age-friendly multifunctional nursing method, applied to the aforementioned nursing equipment, includes the following steps: S1. Equipment initialization: Move the nursing bed to the designated position and lock the casters. After powering on, the intelligent control unit drives the multimodal monitoring unit, nursing support unit, and excretion support unit to complete self-tests. The electric telescopic rod is adjusted to the working posture, and the ultrasonic radar sets the safety threshold of the bed edge. S2. Multi-source data acquisition: The multimodal monitoring unit collects user physiological parameters, behavioral data, and environmental data in real time, and transmits them to the intelligent control unit via dual-mode encrypted communication. For data transmission, wired connections are used between relatively fixed parts within the nursing bed, while Wi-Fi and Bluetooth dual-mode communication is used between relatively mobile parts. When the device is close to the intelligent nursing analysis and prediction program and the network environment is good, Bluetooth transmission is prioritized to reduce power consumption. When the distance is far or the Bluetooth signal is unstable, it automatically switches to Wi-Fi transmission to ensure timely data transmission. Data transmission uses an encryption protocol to ensure data security and privacy. The physiological parameters are collected collaboratively using a bio-radar array, an optical heart rate sensor, and a temperature and humidity sensor; behavioral data is collected using a flexible pressure sensing grid, an audio-visual monitoring camera, and an ultrasonic radar; and environmental data is collected using an ultrasonic radar and a temperature and humidity sensor. S3. Data Preprocessing: The intelligent control unit cleans, standardizes, and fuses the collected data to remove noise and outliers, forming a user status dataset in a unified format. Specifically, data cleaning uses the 3σ criterion to remove outliers, data standardization uses the Z-score method, and data fusion uses the weighted average method to integrate multi-source sensor data. S4. Nursing Needs Prediction: The front-end computer calls its built-in intelligent control program (STAP-Net model and / or pre-trained AI model in remote network server) to analyze the dataset and output the nursing needs type and emergency level for the next 5-30 minutes; Spatiotemporal Attention Network (STAP-Net) model: The input layer receives bio-radar time-series signals (50Hz sampling), stress heatmaps (updated every 5s) and speech MFCC coefficients, captures time dependence through bidirectional LSTM, and the convolutional attention module locates spatial anomalies to predict nursing needs and emergency levels I-III for the next 5-30 minutes; The spatiotemporal attention weights of the STAP-Net model are calculated using Formula 1:

[0023] Where Q / K is the time feature vector and P / S is the spatial feature vector; Nursing needs are classified into three levels of urgency: Level I (urgent), Level II (relatively urgent), and Level III (routine). Level I needs include sudden changes in heart rate Δ>30 bpm, limbs extending more than 15 cm beyond the edge of the bed, etc. Level II needs include single-point pressure for more than 2 hours, voice keywords such as "pain", etc. Level III needs include routine turning over, toileting, etc.

[0024] The STAP-Net model has a structure of "model structure - attention mechanism - probability output" and supports local prediction operations; it mainly includes the following components: Inputs: Mattress pressure matrix (10 fps), wearable ECG-blood oxygen (125 Hz), camera skeleton / expression (25 fps), microphone (16 kHz); Output: ① Action intention within 0–2 seconds (rolling over, ringing a bell, drinking water, using the toilet); ② Risk intentions within 0–4 hours (fall, pressure sore, cardiac arrest, extubation).

[0025] The pre-trained nursing version of the STAP-Net model includes the following components: (1) Multimodal tokenization: Pressure matrix 64×32 → 8×4 non-overlapping patch per frame → flattened 32 tokens; Physiological signal 125 Hz × 4 channels → 1 s sliding window 125×4 → depth separable 1-D Conv → 32 tokens; Video 224×224×3 → STAM 3-D patchify 7×7×4 → 32×32×1 token per frame; 1 second of voice, 16kbps → Whisper encoder, second to last layer, 32 tokens Each modality has 32 tokens, and the four modalities together have 128 tokens, which are then concatenated and embedded into a unified 256-dimensional array.

[0026] (2) Spatial-temporal self-attention module (ST-SA): Input: X∈R^{T×N×C} T=32 frames, N=128 tokens, C=256, Steps: ① Flatten the spatiotemporal region into (T·N)×C, perform standard Multi-Head Self-Attention, and obtain the global spatiotemporal correlation matrix A∈R^{(T·N)×(T·N)}; ② A is averaged along the T dimension to obtain frame-level attention weights, which are used for subsequent "keyframe" selection; ③ Residual connection + LayerNorm → The output still maintains T×N×C and is used by the downstream task header.

[0027] Advantages: It can capture "which frame and which region" is most important in a single forward pass, avoiding the limitations of the local receptive field of CNNs.

[0028] (3) Lightweight time-series fusion unit (WDRU) is used to further compress the T dimension and reduce the number of parameters from 7 M to 2 M.

[0029] The structure is as follows: the forget gate and update gate follow the GRU idea, but the matrix multiplication is changed to grouped 1×1 convolution + depthwise separable convolution; backpropagation is re-derived: gradient backpropagation similar to GRU is implemented on the grouped convolution, which improves training stability by more than 10%.

[0030] Output: T=32 → T′=4, feature dimension remains 256, computational cost can be reduced by more than 60%.

[0031] (4) Probability prediction head: Two parallel MLPs, μ_θ = MLP_μ(h4); regression mean; σ_θ = softplus(MLP_σ(h4)) standard deviation. Loss: Negative Log-Likelihood + 0.1×KL(μ,σ || N(0,1)); Reasoning: Given a 95% prediction interval [μ-1.96σ, μ+1.96σ], nurses can set "red alerts only when σ<0.15", which can significantly reduce false alarms.

[0032] In the intelligent control program, the construction and training of the pre-trained AI model both employ conventional techniques. Specifically, the pre-trained AI model can utilize deep learning algorithms, and in this embodiment, it can employ the existing pre-trained AH-CNN-Res-LSTM (EEG / EMG + inertial sensing) model, the TDCARE nursing brain model, or other suitable models, such as the publicly available pre-trained model AH-CNN-LSTM-Weights (attached to Nature 2024), which is based on EEG / EMG + inertial data from 109 subjects and can be used for fine-tuning bedridden movement intentions; and the Mask R-CNN-Face (Pain) model, pre-trained on the UNBC pain expression database, which can directly extract facial pain levels. The AH-CNN-Res-LSTM model is primarily used for real-time prediction of bedridden individuals' movement intentions such as "getting up / turning over / drinking / calling," relying on sensor data including: piezoelectric films (micro-motion) under the mattress, and camera skeletal points. The main uses of the TDCARE nursing brain model include predicting "risk intentions" such as falls, pressure sores, VTE, pain, and extubation in bedridden patients, which will not be elaborated on further.

[0033] S5. Nursing Action Execution: Based on the prediction results, the intelligent control unit sends instructions to the PLC controller to control the nursing support unit to perform body position adjustment and airbag pressure adaptation, or to control the excretion support unit to perform excretion support actions. The specific process of body position adjustment and airbag pressure adaptation, S51, is as follows: S51-1, the user's current body position (supine, lateral, semi-reclining) is identified through the pressure sensor grid; S51-2, the initial airbag pressure threshold is determined based on the user's weight, and the target pressure distribution is calculated in combination with the body position data; S51-3, the intelligent control unit controls the air pump and electromagnetic switch valve to adjust the inflation volume of each airbag module and drive the hinge mechanism to realize the angle adjustment of the headboard, waist, and legs; S51-4, the pressure sensor array layer provides real-time feedback of pressure data and dynamically corrects the airbag pressure to ensure a body fit of ≥85%.

[0034] The specific process of the excretion support action S52 is as follows: S52-1, the intelligent control unit controls the electric telescopic push-pull rod to extend, driving the foot horizontal sliding bed panel and the rear sliding air-float mattress to move backward, revealing the smart toilet; S52-2, controls the electric lifting tray to descend and the electric telescopic pedal to extend, adjusting the height of the smart toilet to be level with the user's buttocks; S52-3, after the user finishes excretion, the intelligent control unit controls the smart toilet to complete self-cleaning, and then controls each mechanism to reset in the reverse order, with the rear sliding air-float mattress covering the receiving groove.

[0035] S6. Reset and Feedback: After the nursing action is completed, the PLC controller controls each actuator to reset, and the front-end computer synchronizes the execution results and real-time data to the remote network server and remote nursing terminal to update the model parameters.

[0036] The following describes several specific embodiments in detail.

[0037] Example 1 This embodiment, building upon the basic embodiment, specifically provides a pressure ulcer prevention and care device and method based on multimodal monitoring. Focusing on the core needs of pressure ulcer prevention, under the control of an intelligent control unit, it constructs a closed loop of "monitoring-prediction-control-execution" through the collaborative cooperation of the nursing bed, multimodal monitoring unit, and nursing support unit. Based on effective prediction, it automatically and proactively intervenes to address the pain points of high incidence of pressure ulcers in long-term bedridden patients and reliance on manual care. The specific solution is as follows: The main components of the age-friendly multifunctional nursing device based on multimodal monitoring provided in this embodiment include: 1. The main body of the nursing bed is made of high-strength aluminum alloy internal frame (load-bearing capacity ≥250kg). The horizontal bed surface is spliced ​​together by the head, waist and segmented footboards. The footboard has vertical mounting holes and some bedboards have receiving grooves. The bottom is equipped with 6 silent universal casters with a wheel diameter of 12cm (with brakes) for easy movement and stable fixation.

[0038] 2. Multimodal monitoring unit, including: Multimodal monitoring unit 5 includes electric telescopic rod 51, audio and video monitoring camera 53, and ultrasonic radar 52. The audio and video monitoring camera 53 and ultrasonic radar 52 are respectively installed on the upper section of electric telescopic rod 51. The lower section of electric telescopic rod 51 passes through vertical mounting hole 16 and is vertically fixed on the bed tailboard 15, so that the audio and video monitoring camera 53 and ultrasonic radar 52 face the horizontal bed surface 12. The electric telescopic rod 51 can extend and retract vertically. In the working state, its upper section extends upward, forming an angle of 30°~45° between the working center of the audio-visual monitoring camera 53 and the ultrasonic radar 52 and the plane where the horizontal bed surface 12 is located. In the non-working state, its upper section retracts downward and is housed in the lower section, so that the height of the upper surface of the ultrasonic radar 52 is level with or slightly lower than the height of the bed tailboard 15. The electric telescopic rod 51 has a rated thrust of 500N, an extension speed of 10mm / s, and is equipped with a travel limit switch and connected to the control unit. The ultrasonic radar 52 uses a 40kHz high-frequency probe, with a detection distance of 0.3-5m, a detection angle ≤15°, and an IP65 protection rating. The audio-visual monitoring camera 53 is equipped with a 2-megapixel CMOS sensor, a horizontal field of view ≥90°, and supports infrared night vision and 25fps video output.

[0039] The multimodal monitoring unit 5 is integrally installed at the foot of the nursing bed body 1. The height of the audio-visual monitoring camera 53 and the ultrasonic radar 52 are adjusted via an electric telescopic rod 51, positioning it above the nursing bed body 1. This forms an integrated structural layout with the nursing bed body 1, used to monitor the activities of the user or patient on the bed. It has a clearly defined spatial installation position associated with the nursing bed body 1. This position allows the sensors to comprehensively monitor the patient's activities on the bed, such as getting up and turning over. Data is acquired through the telescopic movement of the multimodal monitoring unit 5, providing timely feedback on the patient's status. This facilitates medical or nursing staff in understanding whether the patient is exhibiting abnormal activity. Furthermore, the data corresponds to the patient's position on the bed, ensuring that the collected data directly reflects the patient's condition. Relevant information within the horizontal bed surface area, such as the patient's positional shift on the horizontal bed surface and whether they have left the bed, provides an accurate basis for subsequent data analysis and nursing decisions. In addition, vertically setting the multimodal monitoring unit 5 at the foot of the nursing bed body 1 reduces interference caused by the movement of the nursing bed or accidental collisions with the patient's limbs. It is also far from the ground and other potential sources of interference, reducing the interference of ground vibration, environmental magnetic fields, and other factors on sensor data acquisition, allowing for more accurate monitoring of the patient's real activity status and ensuring the accuracy, reliability, and credibility of the data. This installation position does not hinder medical staff from performing normal nursing operations such as changing bed sheets and conducting physical examinations at the bedside, meeting the design requirements of the nursing bed to facilitate various nursing operations for medical staff.

[0040] Bio-radar array: MIMO architecture, deployed at the four corners of a horizontal bed, emitting 24GHz millimeter waves, with a detection range of 0.5-1.5m, spatiotemporal resolution of 0.5s / time, and positioning accuracy of ≤2cm, accurately collecting heart rate, respiratory rate, and micro-motion characteristics; Flexible pressure sensing mesh: A composite structure of PVDF piezoelectric film and conductive fabric, containing 512 sensing units (detection range 0-100kPa, resolution 0.1kPa), laid in a mesh pattern inside the air-floating mattress to generate a dynamic pressure-thermal map; Audio and video surveillance camera: equipped with a 2-megapixel CMOS sensor, horizontal field of view ≥90°, supports infrared night vision function, and is fixed by the upper section of an electric telescopic pole; Ultrasonic radar: 40kHz high-frequency probe, detection distance 0.3-5m, detection angle ≤15° (IP65 protection), installed side by side with camera on the upper section of electric telescopic pole; Noise-suppressed voice module: Dual microphone array + blind source separation algorithm, keyword recognition rate ≥92% under 60dB ambient noise, integrated inside the headboard panel; Electric telescopic rod: rated thrust 500N, telescopic speed 10mm / s, the lower section is fixed to the mounting hole of the bed end plate, and the upper section can be extended to make the working center of the camera and radar form an angle of 30°~45° with the horizontal bed surface. When not in use, it is retracted to be flush with the bed end plate.

[0041] During operation, the ultrasonic radar 52 transmits and receives ultrasonic signals to achieve real-time perception of the surrounding environment of the bed, the user's position and status. Combined with the audio-visual monitoring camera 53, it can detect the patient's situation on-site in a timely manner, assisting nursing staff in reducing care risks and alleviating workload. Falling out of bed is one of the most common safety hazards in nursing scenarios, especially for users with limited mobility, impaired consciousness, or who turn over at night. The ultrasonic radar 52 forms a protective closed loop from "prediction" to "intervention" through distance monitoring and dynamic early warning. The ultrasonic radar 52 sets a "safe distance threshold" of 30-50cm around the edge of the bed. When the user's body parts, such as hands or feet, extend beyond the edge of the bed or attempt to get up, and the distance between the body and the radar is less than the threshold, the control unit will immediately trigger an early warning and send a reminder to the nurses' station. Nursing staff can then monitor the situation through audio-visual monitoring. The high-frequency monitoring camera 53 provides immediate information about the patient's condition in bed, allowing nursing staff to intervene promptly and preventing falls due to loss of balance or lack of support. Compared to traditional pressure sensors (which can only detect whether the patient is in bed), the ultrasonic radar 52 can detect the dynamic process of "getting up or leaning forward" in advance, providing more timely warnings. The ultrasonic radar 52 can also detect environmental obstacles around the bed. When moving or significantly adjusting the height or angle of the bed, the ultrasonic radar 52 will scan for obstacles (such as bedside tables, chairs, and walls) within a 1-2 meter range around the bed. If an obstacle is detected, it will restrict the movement or adjustment range of the bed and issue an alarm to prompt nursing staff to clear the obstacle, preventing equipment damage from collisions or causing discomfort to the user. This also avoids collisions when adjusting or moving the nursing bed and optimizes the user's autonomous operation experience.

[0042] 3. Nursing support unit, including: multiple interconnected and independently structured air-floating mattresses, each air-floating mattress covering the horizontal bed surface 12 of the nursing bed body 1, each air-floating mattress including a base support layer 314, an inflatable airbag adjustment layer 315, an inflatable airbag adjustment layer 316, and a sealed skin-friendly surface layer 317 stacked sequentially from bottom to top, and horizontally set on the horizontal bed surface of the nursing bed body 1.

[0043] Each air-floating mattress has a layered structure: from bottom to top, it consists of a base support layer (upper layer 5-8cm memory foam + lower layer 10-15cm high-density foam, compression rebound rate ≥80%), an inflatable airbag adjustment layer, a pressure sensor array layer, and a sealed skin-friendly surface layer (medical-grade antibacterial fabric, breathability ≥1500g / (m²・24h)); airbag modules: 4×6 rows of independent airbags (size 15cm×20cm×5cm, burst pressure ≥30kPa), with built-in elastic airbag bags, connected to the air pump in the receiving slot through air intake and exhaust pipes and electromagnetic switch valves, and adjacent airbags are fixed by 1cm wide non-woven fabric connecting straps; pressure sensor array layer: with Shore hardness 60-70A. The pressure sensor is fixed by hot pressing with a polyurethane film as the substrate (adhesion strength ≥5N / 25mm), and an integrated optical heart rate sensor (accuracy ±1bpm), bioelectrical impedance sensor (respiratory rate accuracy ±0.2Hz) and temperature and humidity sensor (18-40℃, 30%-80% RH) are embedded between the breathable sponge and the sealed skin-friendly surface.

[0044] The nursing support unit 3 also includes an air inlet / outlet pipe 318 and an air pump 319 disposed in the receiving slot 144. One end of the air inlet / outlet pipe 318 is connected to the inflatable airbag adjustment layer 315 and the other end is connected to the air pump 319. The inflatable airbag adjustment layer 315 of the air-floating mattress is a sealed space formed by the base support layer 314 and the sealed skin-friendly surface layer 317, and the sealed space is filled with a semi-cylindrical breathable sponge. The pressure sensor array layer is mesh-like and includes multiple uniformly and spaced pressure sensors, which are disposed in the interlayer between the breathable sponge and the sealed skin-friendly surface layer 317.

[0045] The inflatable airbag adjustment layer 315 includes multiple independent inflatable airbag modules arranged in rows and columns (4x6). Each inflatable airbag module is connected to an air pump 319 through an air inlet / outlet pipe 318 and an electromagnetic switch valve. Each inflatable airbag module is provided with an elastic airbag bag 320, and each airbag bag is connected to the air pump 319 through an air inlet / outlet pipe 318 and an electromagnetic switch valve.

[0046] 4. Intelligent control unit, including: Hardware: PLC controller (Siemens S7-1200), front-end computer, remote network server, connected via wired connection, Internet connection or matching Wi-Fi (802.11b / g / n) + Bluetooth (BLE 5.0) dual-mode communication module (AES-128 encryption). The PLC controller has 128GB local storage and a 7-inch waterproof touch panel. Software: Built-in intelligent control software, including data preprocessing module (3σ criterion + Z-score normalization), STAP-Net spatiotemporal attention model and pre-trained AI model, etc., and remote server stores historical data and supports incremental training of models.

[0047] The electric telescopic pole 51, audio and video monitoring camera 53, ultrasonic radar 52, pressure sensor, electromagnetic switch valve, and air pump 319 are all wired or wirelessly connected to the PLC controller and controlled by it to work together.

[0048] The inflatable airbag adjustment layer 316 is equipped with 3-5 sets of detection units along the length of the bed board. Each set of detection units includes an optical heart rate sensor, a bioelectrical impedance sensor and a pressure sensing submodule. The detection units are electrically connected to the control unit and are used to collect the patient's heart rate, respiratory rate and body pressure distribution data in real time. Based on the pressure distribution data collected by the existing pressure sensing submodule, a three-level adjustment algorithm of "weight classification - body position recognition - pressure threshold matching" is used. For example, for patients weighing less than 50kg, the initial pressure threshold of each airbag unit is set to 0.08-0.12Mpa, for patients weighing 50-80kg, it is set to 0.12-0.16Mpa, and for patients weighing more than 80kg, it is set to 0.16-0.20Mpa. By identifying the pressure change trend of the pressure sensing submodule, the patient's body position is identified, such as supine, lateral, and semi-recumbent. When a change in body position is detected, the control unit completes the dynamic adjustment of the airbag pressure in the corresponding area within 3-5 seconds to ensure that the patient's body fits the mattress at a constant level of more than 85%, avoiding pressure concentration caused by local unsupported areas.

[0049] For critical areas prone to pressure ulcers in long-term bedridden patients, such as the sacrum, coccyx, and scapula, the corresponding airbag units are designed with a "center-periphery" pressure gradient structure. The airbag pressure in the central area is 0.02-0.03 MPa lower than that in the peripheral area, creating a gradient pressure field and promoting local blood circulation. For example, the central pressure of the airbag unit corresponding to the sacrum and coccyx is set at 0.10 MPa, and the peripheral pressure is set at 0.13 MPa. This pressure gradient reduces the continuous pressure time in critical areas, thus lowering the risk of pressure ulcers.

[0050] By combining heart rate variability data collected by an optical heart rate sensor and respiratory rate data collected by a bioelectrical impedance sensor, a sleep staging model is constructed. By analyzing the time-domain and frequency-domain indices of heart rate variability and the fluctuation patterns of respiratory rate, the system automatically identifies the patient's awake, light sleep, deep sleep, and REM sleep stages. The control unit uploads the sleep staging results, sleep duration, number of awakenings, and other data to the nursing terminal, providing a quantitative basis for medical staff to assess the patient's sleep quality. When sleep apnea is detected (respiratory rate lasting more than 10 seconds but less than 8 breaths / minute), the control unit automatically triggers a slight vibration of the mattress to wake the patient and restore normal breathing. If the patient does not recover within 1 minute, an alarm signal is sent to the nursing terminal.

[0051] A pressure monitoring sub-node is added to the air inlet of each airbag unit to monitor the internal pressure of the airbag in real time. When the pressure of an airbag unit drops by more than 0.02 MPa within 5 minutes, it is determined to be a "minor leak". The control unit automatically controls the inflation pump to replenish gas to the airbag unit until the pressure returns to the set threshold. When the pressure drops by more than 0.05 MPa within 1 minute, it is determined to be a "serious leak". The control unit immediately closes the air inlet and outlet valves of the airbag unit, sends a leak alarm signal to the nursing terminal, and activates the backup airbag unit (1-2 backup airbags are reserved in each area) to ensure that the overall support performance of the mattress is not affected.

[0052] The inflatable airbag adjustment layer 316 is equipped with a temperature and humidity sensor to collect real-time data on the mattress surface temperature (monitoring range 18-40℃) and humidity (monitoring range 30%-80% RH). When the detected temperature is higher than 30℃ and the humidity is higher than 60% RH, the control unit activates the miniature ventilation fan inside the mattress (wind speed 1.5-2.0m / s) and controls the exhaust frequency of the airbag unit to reduce the mattress surface temperature and humidity through gas circulation. When the detected temperature is lower than 20℃, the carbon fiber heating element inside the mattress (heating power 50-80W) is activated to maintain the mattress surface temperature at 22-25℃, improving patient comfort.

[0053] The aging-friendly multifunctional care method based on multimodal monitoring provided in this embodiment includes the following steps: A1. Equipment Deployment and Initialization: Fix the nursing bed in the ward of the nursing home, lock the casters at the bottom of the bed frame, and start the intelligent control unit after connecting the power. A2. Equipment Self-Check and Operation Preparation: The PLC controller (Siemens S7-1200) drives the multimodal monitoring unit and nursing support unit to complete self-checks: the upper section of the electric telescopic rod extends, so that the audio-visual monitoring camera and the ultrasonic radar (detection distance 0.3-5m) form a 30°~45° angle with the horizontal bed surface, and sets a 40cm bed edge safety threshold and a 1-2 meter surrounding obstacle detection range; the audio-visual monitoring camera activates the infrared night vision function and enters real-time monitoring mode; the pressure calibration of the 4×6 independent airbag modules of the nursing support unit is completed, the pressure sensor array layer (including optical heart rate sensor and temperature and humidity sensor) is activated, and the mattress surface temperature is monitored at 26℃ and the humidity at 55% RH; A3. Patient Fit and Mattress Adjustment: After the patient lies on the sealed, skin-friendly surface of the air-floating mattress, the pressure sensor array layer collects real-time data on the horizontal pressure distribution of the bed surface and transmits it to the intelligent control unit. The system calculates the pressure requirements of each area based on the patient's weight and position, and controls the air pump to precisely inflate the independent inflatable airbag modules through electromagnetic switch valves. It dynamically adjusts the inflation volume of each airbag to adapt to the body's support needs, thus achieving adaptive pressure distribution. A4. Multimodal and multi-source data acquisition: After the user (weighing 70kg, long-term bedridden) lies down, the multimodal monitoring unit comprehensively collects data: a) Millimeter-wave bio-radar array acquires heart rate, respiratory rate and micro-motion characteristics: Bio-radar array (bed four-corner MIMO architecture, 24GHz millimeter wave) acquires heart rate of 72bpm, respiratory rate of 18 breaths / minute, spatiotemporal resolution of 0.5s / time, and positioning accuracy ≤2cm; b) Flexible pressure sensing grid collects pressure distribution data and generates dynamic pressure heat map: The flexible pressure sensing grid (PVDF piezoelectric film + conductive fabric, 512 sensing units) generates dynamic pressure heat map and monitors that the pressure in the sacrococcygeal region is 38 kPa (exceeding the threshold of 30 kPa). c) Noise suppression voice module captures voice commands: The noise suppression voice module (dual microphone array) did not capture any abnormal commands; d) Audio and video surveillance cameras record visual information such as posture and facial expressions: audio and video surveillance cameras record users in a supine position, with the frequency of body movements reduced to 3 times / minute; e) The ultrasonic radar scans the distance between the patient and the edge of the bed, as well as obstacles around the bed, in real time; f) Data is transmitted to the intelligent control unit via wired or Wi-Fi dual-mode communication (AES-128 encryption); Pressure sensors at various locations continuously monitor changes in body pressure, audio and video monitoring camera 53 records visual information such as the patient's posture and expression, and ultrasonic radar 52 scans the distance between the patient and the edge of the bed and obstacles around the bed in real time. A5. Data Processing and Demand Forecasting: The front-end computer preprocesses the data (using the 3σ criterion to remove noise and Z-score normalization), and calls the STAP-Net model and pre-trained AI model on a local or remote network server. This model captures heart rate stability through a bidirectional LSTM, and the convolutional attention module focuses on the abnormal pressure area in the sacral and coccygeal region, according to weight formula 1. Calculate spatiotemporal attention, determine pressure ulcer risk level II, and predict the need for turning and care within 10 minutes; In formula 1: T: matrix / vector transpose; i: The current query position (token or frame index) where attention weights are to be calculated; j: The position of the currently focused "key" (within the same modality or across modalities); k: Summation traversal index, traversing all candidate key positions.

[0054] Q_i ∈ R ^d : The query vector is obtained by linear mapping from the i-th token / frame.

[0055] K_k, K_j ∈ R ^d The key vector is obtained by linear mapping from the k-th or j-th token / frame. When calculating "spatial attention", K comes from different spatial tokens within the same frame; When calculating "temporal attention", K comes from different frames of the same spatial token; P_i ∈ R^d′: Temporal Query vector, specifically designed to capture inter-frame dependencies; S_k, S_j ∈ R^d′: Temporal Key, obtained by mapping the same spatial tokens from different frames; N: Total number of candidate keys in the spatial branch (number of tokens in a single frame); M: Total number of candidate keys for time branches (total number of frames). Q / K is the temporal feature vector, and P / S is the spatial feature vector, achieving synergistic focusing on temporal dependence and spatial anomalies.

[0056] The main model architecture of the STAP-Net model for spatiotemporal attention prediction in this embodiment is as follows: Input layer: Bio-radar time-series signal (50Hz sampling) + pressure heatmap (updated every 5s) + speech feature vector (MFCC coefficients); Feature extraction includes: in the time dimension, bidirectional LSTM captures the long-term dependence of heart rate variability (HRV) on respiratory rhythm; in the spatial dimension, convolutional attention modules locate areas of abnormal pressure (such as persistent pressure on the ischial tuberosity).

[0057] Output layer: Predicts the type of nursing needs (turning over, toileting, first aid) and the urgency level (I-III) within the next 5-30 minutes; Training optimization: The loss function is cross-entropy loss + spatiotemporal consistency constraint (KL divergence penalty and spatial attention shift). A6. Abnormal Nursing Action Execution: When a patient's body part is found to extend beyond the bed edge to a distance less than the safety threshold, or when an obstacle is detected around the bed, the intelligent control unit immediately triggers an alarm and sends a reminder message to medical staff at the nurse station or family members, who will then handle the situation on-site. After receiving the alarm message, medical staff (or family members) can remotely view the patient's condition in real time using the audio-visual monitoring camera 53. If the mattress support needs to be adjusted, the control unit can individually control the inflation and deflation of the corresponding area's airbag module. If the bed needs to be moved or adjusted, the control unit can only perform bed movement or angle / height adjustment operations after scanning and confirming that there are no obstacles through the ultrasonic radar 52. A7. Automated Nursing Action Execution: The intelligent control unit sets the initial airbag pressure to 0.14 MPa (50-80 kg level) according to the "weight classification - body position recognition - pressure threshold matching" algorithm. The PLC controller drives the hinge mechanism of the nursing support unit to keep the head of the bed at 0°, the lumbar spine at 20°, and the legs at 0°, while simultaneously controlling the inflation of the left airbag and the deflation of the right airbag, completing a 35° leftward turn within 3 seconds; the pressure sensor array layer provides real-time feedback data and dynamically corrects the airbag pressure, reducing the pressure in the sacral and coccygeal region to below 20 kPa, forming a gradient pressure field; A8. After all nursing work is completed, the medical staff issues a shutdown command through the intelligent control unit. The air pump 319 releases air from each airbag module through the air intake and exhaust pipes 318. The upper section of the electric telescopic rod 51 retracts downward to a state that is flush with or slightly lower than the bed end plate 15. The control unit is turned off and the power is cut off. After unlocking the casters 15, the nursing bed is returned to its place and stored.

[0058] The brief process of frontal pressure ulcer prevention and proactive intervention nursing based on the STAP-Net model is as follows: Data acquisition: The pressure grid detected that the pressure value in the right hip remained at 38 kPa (>threshold 30 kPa) for 2.5 hours; the bioradar showed that the body movement frequency dropped to 3 times / minute (baseline value 15 times / minute).

[0059] Model inference: The STAP-Net spatiotemporal attention module focuses on the right hip area, determines the pressure ulcer risk level III, and predicts that the body position needs to be adjusted within 10 minutes.

[0060] Execution response: The main control unit triggers a 35° left side rotation and simultaneously starts the air cushion partition alternating inflation mode (inflation cycle 30s).

[0061] This embodiment can achieve full coverage of patient pressure, physiological and behavioral monitoring data. Compared with traditional manual turning (which relies on experience judgment and is done at 2-hour intervals), this embodiment combines the STAP-Net model to predict needs in advance, reducing the incidence of pressure ulcers from more than 10% to less than 2%, and reducing the workload of nursing staff in assisting with turning by 90%. The layered structure (base support layer + airbag adjustment layer) and dynamic pressure adjustment of the air-floating mattress solve the pain points of traditional mattresses, such as fixed support and concentrated local pressure, and can realize automatic and proactive intervention operations in a closed loop of "monitoring-prediction-control-execution".

[0062] Example 2 This embodiment, based on the basic embodiment and Embodiment 1, further provides an intelligent excretion care device and method based on multimodal monitoring. It focuses on the "bedtime excretion" needs of disabled / semi-disabled elderly individuals. Through the segmented structure of the nursing bed, the collaborative design of the excretion support unit and multimodal sensing, it addresses the pain points of traditional excretion care, such as reliance on manual assistance, cumbersome operation, and poor privacy, achieving automated and humanized excretion care. Its difference lies in: The intelligent excretion care device based on multimodal monitoring provided in this embodiment mainly includes: 1. Nursing bed body (optimized specifically for excretion): Segmented horizontal bed panel design: including a one-piece molded head horizontal bed panel, a lumbar horizontal bed panel, and a segmented foot bed panel consisting of a fixed foot horizontal bed panel and a sliding foot horizontal bed panel; Recessed groove: It is located at the junction of the bed panel and the footboard, and in the area enclosed by the bed frame. The size is adapted to the smart toilet to ensure that the top of the toilet is flush with the bed surface after installation. Horizontal sliding mechanism: includes 2 parallel horizontal sliding grooves (fixed to the bed support plate), electric telescopic push-pull rod (telescopic speed 10cm / s, maximum stroke 35cm) and 4 pulleys (embedded in the bottom of the sliding bed panel). The fixed end of the push-pull rod is welded to the bed support, and the moving end is hinged to the sliding bed panel to realize the smooth back and forth sliding of the bed panel; Foot panel: Features an inlet / outlet for the smart toilet (with a sealing strip on the inside), connecting to the receiving recess for easy removal of the entire toilet for cleaning.

[0063] 2. Excretion support unit: Smart toilet: Freestanding design (with built-in water tank and lithium battery pack, no need for external water pipes / power cords), opening upward and the axis is perpendicular to the axis of the bed panel, equipped with an automatic opening and closing lid (controlled by a PLC wireless controller), and the bottom is fixed to the bed frame by shock-absorbing pads; Electric lifting tray: It is located on the inside of the bed side panel in front of the smart toilet. It slides through the vertical guide rail. When not in use, the upper surface is flush with the bed side panel. When in use, it drops 10-15cm to form a hip support groove. Electric telescopic footboard: Installed under the tray, it is connected to the bed frame through a horizontal slide rail. When not in use, the front end is flush with the side panel of the bed. When in use, it extends forward 40-50cm and has an anti-slip texture on the surface.

[0064] 3. Air-floating mattress unit (optimized excretion and drainage): Segmented design: The corresponding bed panel is divided into a front air-floating mattress, a rear fixed air-floating mattress, and a rear sliding air-floating mattress. The rear sliding mattress is fixed to the footboard horizontally sliding bed panel by bottom buckles, so as to achieve synchronous sliding. Non-working state: The three mattress sections are joined together completely, with the sliding mattress covering the receiving groove and the edges tightly fitted to isolate toilet odors; Working state: The sliding mattress moves back with the bed panel to fully expose the toilet opening.

[0065] 4. Multimodal sensing unit (excretion trigger adapter): The core modules of Example 1, such as the bio-radar array, flexible pressure sensing grid, and audio-visual monitoring camera, are retained. The focus is on optimizing the noise-suppressed voice module (the keyword recognition rate of "toilet" is ≥93% under 60dB ambient noise) and the sacrococcygeal pressure capture accuracy of the pressure sensing grid to ensure accurate recognition of excretion intention.

[0066] Intelligent control unit and safety protection interlock: The hardware architecture of the PLC controller (Siemens S7-1200), front-end computer, and remote server is retained, while the software adds a new logic for the linkage of excretion actions. The built-in intelligent control software adds a mechanism interlock module, which includes interference recognition rules such as "disable turning over in the excretion state" and "disable mattress inflation if the sliding bed panel is not reset" to ensure safe and coordinated actions.

[0067] A horizontal sliding mechanism is provided at the lower part of the footrest sliding bed panel 1432. This mechanism includes horizontal sliding grooves and electrically telescopic push-pull rods arranged parallel to each other on the bed frame support (specifically, on the support plate), and pulleys (conventional technology, not shown in the figure) that are set on the bottom surface of the footrest sliding bed panel 1432, embedded in the horizontal sliding grooves, and capable of linear back-and-forth reciprocating motion. The fixed ends of each electrically telescopic push-pull rod are fixed to the bed frame support (support plate), and the moving ends are fixed (hinged) to the bottom surface of the footrest sliding bed panel 1432. The bed frame support adopts a conventional technology design, typically including four corner legs, multiple central support legs, and a grid-shaped internal support structure composed of multiple horizontal support rods, vertical connecting support rods, and the support plate, which serves to connect, support, accommodate, and fix the various parts, and will not be described in detail here.

[0068] In use, when the smart toilet is needed, the PLC controller 4 controls the electric telescopic push-pull rod to extend forward, causing the footrest sliding bed panel 1432 to slide backward, moving the rear sliding air-float mattress 313 backward as well, away from directly above the receiving groove 144, exposing the upper opening of the smart toilet 21 for patient use. After use, the PLC controller 4 controls the electric telescopic push-pull rod to retract, causing the footrest sliding bed panel 1432 to slide forward, moving the rear sliding air-float mattress 313 forward as well, covering directly above the receiving groove 144, closing the receiving groove 144 and the upper opening of the smart toilet 21, restoring the complete splicing of the air-float mattress 31.

[0069] The excretion support unit further includes an electrically adjustable tray, an electrically retractable pedal, and a smart toilet inlet / outlet. The smart toilet is embedded in a receiving groove, with its top opening flush with the upper surface of the horizontal bed panel and its bottom fixed to the bed frame via shock-absorbing pads. The electrically adjustable tray is located on the inner side of the bed side panel directly in front of the smart toilet and is slidably connected to the bed side panel via a vertical guide rail. When not in use, its upper surface is flush with the upper surface of the bed side panel. The electrically retractable pedal is installed below the electrically adjustable tray and is connected to the bed frame via a horizontal slide rail. When not in use, its front end is flush with the outer surface of the bed side panel. The smart toilet inlet / outlet is located on the footboard panel corresponding to the receiving groove, with an opening size adapted to the outer diameter of the smart toilet. A sealing strip is provided on the inner edge, communicating with the inside of the receiving groove.

[0070] The nursing support unit consists of multiple modular air-floating mattresses, which are divided into a front air-floating mattress, a rear fixed air-floating mattress, and a rear sliding air-floating mattress according to the horizontal bed panel. The bottom of the rear sliding air-floating mattress is fixed to the foot horizontal sliding bed panel by buckles, so as to achieve synchronous sliding.

[0071] The excretion support unit 2 also includes an electrically adjustable tray 22 that is mounted on the bedside panel 13 directly in front of the smart toilet 21 and moves up and down. When not in use, the upper surface of the electrically adjustable tray 22 is flush with the upper surface of the bedside panel 13, so as not to affect the patient's normal use. When in use, the upper surface of the tray is lower than the upper surface of the bedside panel 13, forming a groove, which is used to lower the patient's body position during use, so that the buttocks are flush with the upper opening of the smart toilet, and the patient's thighs are embedded in the groove, thereby improving comfort and safety.

[0072] The excretion support unit 2 also includes an electrically retractable pedal 23 that is installed on the bed side panel 13 directly in front of the smart toilet 21 and moves back and forth. When the electrically lifting pedal 22 is not in use, its front end is flush with the front end of the bed side panel 13 and does not change the normal shape of the nursing bed; when in use, its upper front end extends beyond the front end of the bed side panel 13 to provide lower support for the patient.

[0073] In use, when the smart toilet is needed, the PLC controller 4 first controls the electric lifting platform 22 to descend, so that its upper surface is lower than the upper surface of the bed side panel 13. Then, it controls the electric telescopic pedal 23 to extend, providing support for the patient. The patient can place both feet on the electric telescopic pedal 23, with their buttocks level with the upper surface of the smart toilet, lowering their center of gravity and improving safety and comfort. Simultaneously, when the patient gets in and out of bed, the electric telescopic pedal 23 can also be extended independently to assist the patient's movement and improve safety when getting in and out of bed.

[0074] The nursing support unit 3 includes multiple interlocking air-floating mattresses 31 mounted on a horizontal bed panel 14. Each air-floating mattress 31 comprises three independently mounted and interlocking parts: a front air-floating mattress 311, a rear fixed air-floating mattress 312, and a rear sliding air-floating mattress 313. The rear sliding air-floating mattress 313 is mounted on the footrest horizontal sliding bed panel 1432 and can slide horizontally back and forth along with the footrest horizontal sliding bed panel 1432. When the smart toilet 21 is in operation, the rear sliding air-floating mattress 313 can slide horizontally backward along with the footrest horizontal sliding bed panel 1432, opening the area above the corresponding receiving groove 144 of the smart toilet 21 and fully exposing the upper opening of the smart toilet. When the smart toilet 21 is not in operation, the rear sliding air-floating mattress 313 can slide horizontally forward in the horizontal direction along with the foot sliding bed panel 1432, sealing the position above the corresponding accommodating groove 144 of the smart toilet 21. At the same time, the rear sliding air-floating mattress 313 and the edge of its adjacent air-floating mattress are pressed together to prevent the odor of the smart toilet 21 from leaking upward.

[0075] On the bed foot panel 12 at the front side of the smart toilet 21, there is also a smart toilet inlet / outlet 121. The cavity inside the smart toilet inlet / outlet 121 is connected to the receiving groove 144, which is used to put the smart toilet 21 into or out of the nursing bed, facilitating the daily cleaning and maintenance of the smart toilet. Since the smart toilet 21 and the receiving groove 144 are both close to the bed foot panel 12, the entry and exit distances are short, and the internal cavities are relatively shallow. The smart toilet 21 can be placed on a sliding plate and moved in or out as a whole along the support plate and other flat parts of the bed frame. The smart toilet 21 can be designed as an independent structure using conventional technology, so that it has its own water tank and power supply (such as a lithium battery pack), without the need for external water pipes or power cords, so as to facilitate the whole-body entry and exit operation. The smart toilet 21 can also be further equipped with an automatically opening and closing lid using conventional technology. The operation of the smart toilet is controlled by a wireless controller that is matched with the PLC controller. When not in use, the upper opening is closed, and when in use, the opening is opened. All of these are implemented using conventional technology and will not be described in detail here.

[0076] The intelligent excretion care method based on multimodal monitoring provided in this embodiment further includes the following steps: B1. Triggering of Excretion Intent and Multi-Source Data Collection: Triggering conditions: The user issues the "toilet" command through the noise-suppressed voice module (confidence level 0.93), the flexible pressure sensor grid detects a 62% decrease in sacral and coccygeal pressure within 5 minutes, and the audio-visual monitoring camera captures the user's behavioral characteristics of raising their hips and slightly leaning forward. Environmental and physiological data verification: Ultrasonic radar scan showed no obstructions within a 1-2 meter radius around the bed; temperature and humidity sensors monitored mattress humidity at 58% RH (no abnormalities); and bio-radar array collected heart rate data at 75 bpm and respiratory rate data at 20 breaths / minute (stable physiological state). Data transmission: All data is transmitted to the intelligent control unit in real time via Bluetooth dual-mode communication (AES-128 encryption).

[0077] B2 Demand Forecasting and Order Issuance: The front-end computer calls its built-in STAP-Net nursing version model or a pre-trained AI model on a remote network server. Combining features such as "voice command + sudden drop in sacral and coccygeal pressure + hip lifting movement", it eliminates false triggers (such as turning over or adjusting body position) and predicts toilet needs at level III with an accuracy of 91.2%. The intelligent control unit sends a linkage command to the PLC controller, specifying the sequence of actions: the sliding bed panel moves backward → the toilet seat opens → the lifting tray descends → the telescopic pedal extends → the lumbar airbag pressure adjusts.

[0078] Taking bedridden patients' "nighttime toileting" as an example, the data flow and intent prediction process of the nursing version of STAP-Net model are as follows: 1. Sensing 30 seconds starting from 00:33:15: • Pressure matrix: High-frequency micro-vibration of 0.3-0.5 Hz in the hip area, energy increased by 230%; • HRV: LF / HF changed from 1.8 to 4.1; • Video: Leg flexion angle changes by 16° twice, facial AU12 (frowning) activated; • Voice: No calls, but two 260 ms groaning segments were heard.

[0079] 2. Tokenization & Embedding The previous four-mode signal was cut into 32 frames, each frame generated 128 tokens, and embedded in 256 dimensions → resulting in a 32×128×256 tensor.

[0080] 3. ST-SA Module Attention heatmap shows: • Frame-level: Frames 18-22 (corresponding to 18-22 seconds) have the highest weight; • Spatial: Stress token #2 (hip area), video token #27 (leg frame), and physiological token #0 (HRV) have the strongest responses.

[0081] The model automatically identifies the combination of features: "micro-movement of the hips + knee flexion + sympathetic excitation".

[0082] 4. WDRU Compression 32 frames → 4 keyframes, features 4×128×256.

[0083] 5. Intent Head Output: • Action intention "to use the toilet" μ=0.87, σ=0.04; • Risk intention "fall" μ=0.21, σ=0.08 (negligible).

[0084] 95% upper limit of 0.91 ≥ 0.8 threshold → Trigger bedside soft light + voice confirmation "Do you need to use the toilet?"; If the patient blinks twice (preset confirmation), the robot automatically moves the bedpan and notifies the nurses' station.

[0085] This embodiment uses the STAP-Net model for nursing care in conjunction with a nursing bed scenario. The experimental results of the direct application are as follows: Dataset: 112 long-term bedridden patients from a tertiary hospital between October 2024 and April 2025, totaling 1,800 bed-days, with 4,312 action intentions and 207 risk events annotated; Indicators: Action intention F1 = 94.1%, average lead time 1.9 s; Risk intention AUROC = 0.96, prediction interval coverage 94.3% (close to theoretical 95%).

[0086] Deployment: Edge box Jetson Orin Nano (20 W) (1) Inference latency 78 ms, (2) Memory usage < 2GB, (3) False alarms ≤ 2 times / bed per day.

[0087] B3. Execution of excretion care procedures: Step 1: Activation of the horizontal sliding mechanism. The PLC controller controls the electric telescopic push-pull rod to extend 35cm, which drives the horizontal sliding bed panel at the foot and the rear sliding air-floating mattress to move backward synchronously, fully exposing the smart toilet in the recessed area. The action takes 15 seconds. Step 2: The excretion support unit works in tandem. The electric lifting tray descends 12cm along the vertical guide rail, forming a hip support groove that matches the opening of the smart toilet; the electric telescopic pedal extends forward 45cm, providing foot support for the user, level with the user's hips, and lowering the body's center of gravity; Step 3: Fitting the air-supported mattress. The front and back sections of the fixed air-supported mattress maintain their original support pressure. The pressure sensor array layer monitors changes in lumbar pressure in real time and dynamically adjusts the inflation volume of the lumbar airbags to ensure 88% fit to the user's upper body, preventing the body from being suspended or shifting during excretion.

[0088] B4 Post-Expellement Repositioning and Maintenance: Self-cleaning and reset: After the user finishes excreting, the "end" signal is triggered by voice command or control panel, and the smart toilet automatically starts the flushing and deodorizing program (takes 30 seconds); the PLC controller drives each mechanism to reset in reverse order: the telescopic pedal retracts → the lifting tray rises to be flush with the side panel of the bed → the electric telescopic push-pull rod retracts → the foot sliding bed panel and the rear sliding air-floating mattress move forward to cover the receiving groove and restore the bed surface to flatness; Routine maintenance: Caregivers can place the freestanding toilet on the sliding plate through the smart toilet inlet / outlet on the foot panel of the bed, and move it out along the support plate inside the bed for cleaning, water replenishment, and lithium battery charging. After maintenance, simply push it back into the receiving groove without disassembling the pipes or adjusting the bed structure.

[0089] In this embodiment, the process of recognizing and executing toilet-use intentions based on the STAP-Net model prediction is briefly as follows: Data fusion: The voice module captured the keyword "need to get up" (confidence level 0.89); the pressure heatmap showed that the pressure value in the sacrococcygeal region decreased by 62% within 5 minutes; Feature association: The model identifies the combination of "sudden drop in pressure + voice command" features and predicts toilet needs with a confidence level of 91.2%.

[0090] Proactive service: The smart toilet hidden inside the nursing bed is automatically revealed, while the electric lifting tray descends along the vertical guide rail and the electric telescopic pedal extends forward for easy use by the patient.

[0091] This embodiment solves the problem of toilet-mattress interference in traditional nursing beds by using a segmented bed panel combined with a synchronously sliding mattress design. The bed surface is flat and free of protrusions when not in use, and the toilet quickly becomes visible when in use, balancing comfort and practicality. It improves the accuracy of multimodal triggering: compared to single voice or pressure triggering, this embodiment combines voice commands, pressure changes, and behavioral characteristics for triple judgment, resulting in a false trigger rate of less than 3%, ensuring accurate identification of elimination needs. It achieves both humanization and efficiency improvement: users can eliminate without leaving the bed, without the need for assistance from caregivers or manual bed adjustments, reducing operation time from 12.5 minutes in traditional care to 3.2 minutes. Within minutes, the toilet access success rate is ≥95.7%; the workload of caregivers is reduced by 82.3%, requiring only regular toilet maintenance, significantly reducing the burden of care; privacy and environmental friendliness are achieved: the sliding mattress fits snugly after repositioning, effectively isolating toilet odors; the independent toilet design avoids the problems of inconvenient cleaning and easy bacterial growth of traditional integrated toilets, improving hygiene; enhanced safety protection: the interlocking module ensures that each action is executed in sequence, avoiding mechanical interference; the support design of the lifting tray and telescopic pedal reduces the risk of falls during defecation for disabled users, with a safety protection rate of 99.1%. The embedded design and independent disassembly structure of the smart toilet in this embodiment solve the pain points of inconvenient cleaning and odor leakage of traditional nursing bed toilets, allowing this age-friendly and humanized design solution to be used for a long time and in multiple scenarios.

[0092] Example 3 This embodiment, based on the basic embodiment and embodiment 1, further provides a remote collaborative emergency nursing device and method based on multimodal monitoring. Its difference from the aforementioned embodiments lies in: The remote collaborative emergency nursing equipment based on multimodal monitoring further includes a remote network server communicating with the front-end computer network. This remote network server incorporates a spatiotemporal attention network (STAP-Net) model and a pre-trained deep learning-based AI model. By learning from existing data, the AI ​​model achieves better predictive and analytical accuracy, which is then directly invoked by the front-end computer. The computer then issues commands to the PLC controller, which in turn controls other units of the nursing equipment to execute commands. The intelligent control unit also includes a remote nursing terminal, which can be a mobile phone, tablet, or nursing workstation, connected wirelessly to the remote network server of the intelligent control unit. The remote nursing terminal receives user physiological data, nursing need prediction results, and alarm information in real time, and supports audio and video monitoring, nursing parameter adjustment, and sending remote emergency intervention commands.

[0093] The multimodal monitoring-based remote collaborative emergency care method provided in this embodiment includes the following steps: C1. Abnormal Data Acquisition and Local Response: During the nighttime period, the multimodal sensing unit suddenly generated abnormal data: the bio-radar array detected a sudden change in the user's heart rate from 75 bpm to 135 bpm. Δ =60 bpm), respiratory rate dropped to 10 breaths / minute, body movement disappeared; ultrasonic radar did not detect limbs extending beyond the edge of the bed, and the flexible pressure sensor grid showed no significant change in pressure distribution; the audio and video monitoring camera recorded the user's pale face through infrared night vision.

[0094] C2. Remote Collaborative Intervention: The intelligent control unit synchronizes SOS signals, real-time audio and video, and physiological data to the remote nursing terminal (nurse station workstation) via Wi-Fi. After logging into the terminal, medical staff check the data and determine that it is a Level I emergency need. They then remotely issue the instruction "adjust to semi-recumbent position + continuous monitoring." The instruction is forwarded to the front-end computer via the remote network server and then sent to the PLC controller.

[0095] C3. Emergency Care Execution: The PLC controller drives the hinge mechanism of the air-floating mattress unit, raising the headboard to 60°, the lumbar region to 20°, and keeping the legs at 0°. The airbag adjustment system dynamically adjusts the pressure in each area to ensure even pressure distribution on the user's upper body. The bio-radar array and optical heart rate sensor collect data every second and synchronize it to the remote terminal. Dual-mode communication maintains a stable connection with no data loss.

[0096] C4. Follow-up Processing and Feedback: On-site medical personnel arrived urgently and quickly implemented intervention based on data synchronized from the remote terminal. The user's heart rate gradually recovered to below 90 bpm. After the nursing care was completed, the PLC controller unlocked the actuator, and the front-end computer synchronized the intervention process and physiological data trends to the remote server, updated the STAP-Net model and AI model parameters, and optimized the emergency demand determination logic.

[0097] Compared to traditional emergency care (which relies on user calls or scheduled inspections, with response times exceeding 8 minutes), this embodiment achieves second-level emergency response through anomaly monitoring by multimodal sensing units and collaboration between local and remote terminals. This significantly shortens rescue arrival time and significantly improves the success rate of emergency treatment. The accurate determination of Level I emergency status based on the STAP-Net model solves the problems of false alarms and missed alarms in emergency needs of traditional equipment, demonstrating the efficiency and accuracy of the closed-loop operation of "monitoring-early warning-remote intervention".

[0098] It should be noted that in other embodiments of the present invention, other different solutions obtained by making specific selections within the scope of the structures, components, steps, instruments, algorithms, models, and process parameters and conditions described in the present invention can all achieve the technical effects described in the present invention. Therefore, the present invention will not list them one by one.

[0099] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention using the methods and techniques disclosed above, or modify them into equivalent embodiments with equivalent changes, without departing from the scope of the present invention. All equivalent changes made to the components, proportions, and processes of the present invention should be covered within the protection scope of the present invention.

Claims

1. A multi-modal monitoring-based age-friendly multifunctional nursing device, characterized in that, It includes the nursing bed body and a multimodal monitoring unit, nursing support unit, excretion support unit and intelligent control unit installed on the nursing bed body; The nursing bed body includes a headboard, a footboard, sideboards, a horizontal bed panel, and an internal support frame, which provide a physical carrier for the multimodal monitoring unit, nursing support unit, excretion support unit, and intelligent control unit. The multimodal monitoring unit includes an audio and video monitoring camera and an ultrasonic radar, which are vertically mounted on the nursing bed body via an electric telescopic rod. It is used to collect user physiological parameters, behavioral data and environmental data in real time, and transmit them to the intelligent control unit through dual-mode communication. The nursing support unit includes multiple air-floating mattresses that are horizontally set on the bed surface of the nursing bed body and spliced ​​together, used to perform nursing actions such as body position adjustment, airbag pressure adaptation, and assisting in excretion support. The excretion support unit includes a smart toilet, which enables users or patients to defecate without leaving the bed. The intelligent control unit includes a PLC controller and a front-end computer. The front-end computer has a built-in intelligent control program that is electrically or wirelessly connected to the multimodal monitoring unit, nursing support unit, and excretion support unit. It is used to receive and analyze multi-source data on user physiology, behavior, and environment collected by the multimodal monitoring unit, process the data, control the nursing support unit to complete adaptive nursing actions, control the excretion support unit to complete excretion support actions, and reset the system after the actions are completed, forming a hardware and software combined monitoring-prediction-control-execution collaborative architecture.

2. The age-friendly multifunctional nursing device based on multimodal monitoring according to claim 1, characterized in that, The horizontal bed panel of the nursing bed body includes a head horizontal bed panel, a lumbar horizontal bed panel, and a foot horizontal bed panel; the foot horizontal bed panel includes a foot horizontal fixed bed panel and a foot horizontal sliding bed panel; between the foot horizontal sliding bed panel, the foot horizontal fixed bed panel, and the lumbar horizontal bed panel, there is a receiving groove, in which a smart toilet is installed, the opening of the smart toilet facing upwards, and its axis perpendicular to the axis of the horizontal bed panel.

3. The age-friendly multifunctional nursing device based on multimodal monitoring according to claim 1, characterized in that, The multimodal monitoring unit also includes a bio-radar array, a flexible pressure sensing grid, and a noise-suppressed speech module. The bio-radar array adopts a MIMO architecture and is deployed at the four corners of the horizontal bed surface to collect heart rate, respiratory rate, and micro-motion characteristics. The flexible pressure sensing grid is used to generate dynamic pressure heatmaps. The noise-suppressed speech module uses a dual-microphone array and a blind source separation algorithm to suppress noise.

4. The age-friendly multifunctional nursing device based on multimodal monitoring according to claim 3, characterized in that, The electric telescopic rod includes a fixed lower section and a telescopic upper section. The lower section is vertically fixed in the vertical mounting hole of the bed foot panel, and the upper section extends and retracts vertically. In the working state, the upper section extends upward, so that the working center of the audio-visual monitoring camera and the ultrasonic radar forms an angle of 30°~45° with the horizontal bed surface. In the non-working state, the upper section retracts into the lower section, and the upper end face of the ultrasonic radar is flush with or slightly lower than the bed foot panel. The audio-visual monitoring camera supports infrared night vision function. The ultrasonic radar has a detection distance of 0.3-5m, and a safe distance threshold of 30-50cm from the bed edge is set.

5. The age-friendly multifunctional nursing device based on multimodal monitoring according to claim 1, characterized in that, The air-supported mattress of the nursing support unit further includes a base support layer, an inflatable airbag adjustment layer, a pressure sensor array layer, a sealed skin-friendly surface layer, and an air pump assembly. Each layer is stacked sequentially from bottom to top, and the edges are sealed and fixed using a heat-sealing process. The base support layer is a double-layer composite sponge structure, with a thick memory foam upper layer and a thick high-density sponge lower layer, covering the entire horizontal bed panel. The inflatable airbag adjustment layer consists of rows and columns of independent airbag modules, each with a built-in elastic airbag bag, connected one-to-one to the electromagnetic switch valve of the air pump assembly via branch inlet and outlet pipes. The pressure sensor array layer is distributed in a mesh pattern and evenly embedded between the inflatable airbag adjustment layer and the sealed skin-friendly surface layer. The nursing support unit is divided into three sections corresponding to the horizontal bed panel: a front air-supported mattress, a rear fixed air-supported mattress, and a rear sliding air-supported mattress. The bottom of the rear sliding air-supported mattress is fixed to the footrest horizontal sliding bed panel via buckles, enabling synchronous sliding.

6. The age-friendly multifunctional nursing device based on multimodal monitoring according to claim 5, characterized in that, The pressure sensor array layer integrates an optical heart rate sensor, a bioelectrical impedance sensor, and a temperature and humidity sensor; the optical heart rate sensor is used to monitor heart rate, the bioelectrical impedance sensor is used to monitor respiratory rate, and the temperature and humidity sensor monitors temperature and humidity; the air pump assembly includes an air pump, a main intake and exhaust pipe, and multiple electromagnetic switching valves, which are integrated and installed in the bed support accommodating area below the horizontal bed panel, and are connected to the branch intake and exhaust pipes of each airbag module through the main intake and exhaust pipe.

7. The age-friendly multifunctional nursing device based on multimodal monitoring according to claim 1, characterized in that, The excretion support unit further includes an electrically adjustable tray, an electrically retractable pedal, and a smart toilet inlet / outlet. The smart toilet is embedded in a receiving groove, with its top opening flush with the upper surface of the horizontal bed panel and its bottom fixed to the bed frame via shock-absorbing pads. The electrically adjustable tray is located on the inner side of the bed side panel directly in front of the smart toilet and is slidably connected to the bed side panel via a vertical guide rail. When not in use, its upper surface is flush with the upper surface of the bed side panel. The electrically retractable pedal is installed below the electrically adjustable tray and is connected to the bed frame via a horizontal slide rail. When not in use, its front end is flush with the outer surface of the bed side panel. The smart toilet inlet / outlet is located on the footboard panel corresponding to the receiving groove, with an opening size adapted to the outer diameter of the smart toilet. A sealing strip is provided on the inner edge, communicating with the inside of the receiving groove.

8. The age-friendly multifunctional nursing device based on multimodal monitoring according to claim 1, characterized in that, The intelligent control unit also includes a remote network server, a front-end computer, and a built-in intelligent control program on the remote network server, including a spatiotemporal attention network STAP-Net model and a pre-trained deep learning-based AI model. By learning from existing data in advance, the model has better predictive and analytical accuracy, which can be directly built into the front-end computer or called remotely, and then the computer issues instructions to the PLC controller, which controls other units on the nursing equipment to perform the operation.

9. A multi-modal monitoring-based, age-friendly, multifunctional nursing method, characterized in that, The device applied to the nursing equipment according to any one of claims 1-8 includes the following steps: S1. Equipment initialization: Move the nursing bed to the designated position and lock the silent casters. After powering on, the intelligent control unit drives the multimodal monitoring unit, nursing support unit, and excretion support unit to complete self-tests. The electric telescopic rod is adjusted to the working posture, and the ultrasonic radar sets the safety threshold of the bed edge. S2. Multi-source data acquisition: The multi-modal monitoring unit collects user physiological parameters, behavioral data and environmental data in real time, and transmits them to the intelligent control unit through dual-mode communication encryption. S3. Data Preprocessing: The intelligent control unit cleans, standardizes, and merges the collected data, removing noise and outliers to form a user status dataset in a unified format. S4. Nursing Demand Prediction: The front-end computer calls the STAP-Net model of its built-in or remote network server and the pre-trained AI model to analyze the dataset and output the nursing demand type and urgency level for the next 5-30 minutes. S5. Nursing Action Execution: Based on the prediction results, the intelligent control unit sends instructions to the PLC controller to control the nursing support unit to perform body position adjustment and airbag pressure adaptation, or to control the excretion support unit to perform excretion support actions. S6. Reset and Feedback: After the nursing action is completed, the PLC controller controls each actuator to reset, and the front-end computer synchronizes the execution results and real-time data to the remote network server and remote nursing terminal to update the model parameters.

10. The age-friendly multifunctional nursing method based on multimodal monitoring according to claim 9, characterized in that, In step S2, physiological parameters are collected collaboratively using a bio-radar array, an optical heart rate sensor, and a temperature and humidity sensor; behavioral data are collected using a flexible pressure sensing grid, an audio-visual monitoring camera, and an ultrasonic radar; and environmental data are collected using an ultrasonic radar and a temperature and humidity sensor. In step S3, data cleaning uses the 3σ criterion to remove outliers, data standardization uses the Z-score method, and data fusion uses the weighted average method to integrate multi-source sensor data.

11. The age-friendly multifunctional nursing method based on multimodal monitoring according to claim 9, characterized in that, In step S5, the specific process of step S51, which involves adjusting the body position and adapting the airbag pressure, is as follows: S51-1. Identify the user's current body position through a flexible pressure sensing grid; S51-2. Determine the initial pressure threshold of the airbag based on the user's weight, and calculate the target pressure distribution by combining body position data; The S51-3 and PLC controller control the air pump assembly and electromagnetic switch valve, adjust the inflation volume of each airbag module, and drive the hinge mechanism to achieve bed angle adjustment. S54: The pressure sensor array layer provides real-time feedback of pressure data and dynamically corrects the airbag pressure to ensure a body fit of ≥85%.

12. The age-friendly multifunctional nursing method based on multimodal monitoring according to claim 9, characterized in that, In step S5, the specific process of step S52 of the excretion support action is as follows: The S52-1 PLC controller controls the extension of the electric telescopic push-pull rod, which drives the horizontal sliding bed panel at the foot and the rear sliding air-floating mattress to move backward, revealing the smart toilet. S52-2, Control the electric lifting tray to descend and the electric telescopic pedal to extend, and adjust the height of the smart toilet to be level with the user's buttocks; S52-3 After the user finishes excreting, the smart toilet is controlled to complete self-cleaning. Then, the mechanisms are controlled to reset in the reverse order, and the rear sliding air-floating mattress covers the receiving groove.

Images