Emergency patient communication auxiliary device based on eye signal identification

Abstract

The invention relates to an emergency patient communication auxiliary device based on eye signal recognition, and belongs to the technical field of medical instruments, the emergency patient communication auxiliary device comprises an eye signal capturing frame, and a subminiature high-definition flat camera and a pupil tracking and action recognition module are arranged in the eye signal capturing frame. Nanoscale infrared light emitting diode arrays are uniformly distributed around the inner side surface of the lens ring, and a miniature laser radar sensor is arranged above the outer side of the lens ring. The signal processing unit is connected with all the modules through wireless signals and receives and processes eye signals. The operation execution unit comprises a high-resolution display screen, a voice playing module and the like. The device can accurately collect eye signals and convert the eye signals into operation instructions, and intelligent recognition and response to communication requirements of patients are achieved. By capturing the eye movement of the patient and converting the eye movement into the clear communication instruction through processing, the patient can feed back the confirmation instruction, effective communication between the hand-constrained emergency patient and the medical staff is achieved, the requirements of the patient can be transmitted in time, and the communication efficiency and safety in emergency nursing are improved.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, specifically to an emergency patient communication assistance device based on eye signal recognition. Background Technology

[0002] In the emergency department, it is often necessary to treat various critically ill patients, such as those with cardiogenic shock, severe trauma leading to respiratory and circulatory failure, acute cerebrovascular accident coma, and drug or poisoning. Their lives are in danger, their physiological functions are severely impaired, and many patients have respiratory depression or disorders. At this time, emergency endotracheal intubation becomes a key treatment method.

[0003] While intubation is crucial for maintaining vital signs, it can cause discomfort and a feeling of a foreign body in patients. In this state, especially for patients with suppressed or disordered higher brain functions, their voluntary control of limbs decreases, and their hands are prone to unconscious movements. This can pull, shift, or even dislodge the tube, compromising airway seal and tube position, affecting respiratory support, and potentially leading to serious complications such as airway mucosal damage and bleeding, threatening their lives. Therefore, in emergency care, medical staff often restrain patients' hands to reduce risks and ensure safe intubation and the implementation of medical procedures. However, this raises a new problem: with hands immobilized, patients cannot express their needs through nonverbal means such as writing or gestures, and intubation further hinders verbal expression, effectively cutting off communication channels.

[0004] In actual nursing care, patients may experience localized pressure pain, thirst, difficulty breathing due to sputum accumulation, nausea and vomiting due to gastrointestinal discomfort, and muscle fatigue and numbness due to fixed posture caused by prolonged bed rest. If medical staff cannot promptly detect and understand these discomforts, it will affect patient comfort and treatment compliance, delaying the timing of treatment. For example, respiratory distress caused by sputum blockage, which cannot be expressed, may lead to a sharp drop in blood oxygen saturation and hypoxemia; prolonged immobility in the same pressure position can easily lead to pressure sores, increasing patient suffering, hospitalization time, medical costs, and infection risks, thus affecting the quality of prognosis and recovery. Therefore, establishing an effective communication method between patients and medical staff under the premise of hand restraint during emergency care and subsequent nursing is a key technical challenge that urgently needs to be overcome. It is also an inevitable requirement for improving the quality of emergency nursing, ensuring patient safety and comfort, and conforming to the trend of humanized and precise nursing in modern emergency medicine. This invention is based on this, aiming to fill the current gap in emergency nursing technology by capturing patients' eye signals and converting them into communication instructions, and has important clinical significance and application value. Summary of the Invention

[0005] In view of this, the purpose of this invention is to propose an emergency patient communication assistance device based on eye signal recognition. This device comprises an eye signal acquisition module consisting of a carbon fiber composite headband frame, various sensors and cameras, a signal processing unit with an image processing chip and a machine learning model storage chip, an operation execution unit with a display screen and a voice playback module, and a power supply module consisting of a rechargeable lithium battery pack. This addresses the problem of poor communication between emergency patients and medical staff when their hands are restrained, leading to difficulties in timely communication and impacting patient comfort, treatment effectiveness, and even delays in treatment. Furthermore, it improves the quality and efficiency of emergency care, reduces medical risks, aligns with the trend of modern emergency medicine towards precise and humanized care, fills a gap in current emergency nursing technology in this area, and provides emergency patients with higher quality and more efficient nursing care.

[0006] This invention is achieved through the following technical solution:

[0007] An emergency patient communication assistance device based on eye signal recognition includes an eye signal capture frame. Inside the eye signal capture frame are an ultra-miniature high-definition flat camera and a pupil tracking and motion recognition module. A miniature screen is fixedly mounted at the front end of the eye signal capture frame. A signal processing unit is fixedly mounted below the eye signal capture frame. An operation execution unit is mounted outside the eye signal capture frame. The signal processing unit communicates wirelessly with the ultra-miniature high-definition flat camera, the pupil tracking and motion recognition module, the miniature screen, and the operation execution unit.

[0008] Furthermore, the eye signal capturing frame includes a lens ring and an adjustable headband. The lens ring is connected by a connecting bridge. The ultra-small high-definition flat camera consists of a main camera and an auxiliary camera. The main camera is fixedly mounted inside the lens ring above the main camera via a precision adjustment bracket. The auxiliary camera is fixedly mounted below the main camera. A nano-scale infrared light-emitting diode array is evenly distributed around the inner surface of the lens ring. The pupil tracking and motion recognition module is fixedly mounted directly above the lens ring. An image sensor is fixedly mounted on the side wall of the lens ring. The image sensor is connected to the main camera and the auxiliary camera.

[0009] Furthermore, a miniature lidar sensor is fixedly installed above the outer side of the mirror ring. The miniature lidar sensor includes a left lidar sensor and a right lidar sensor installed above the outer side of the mirror ring near the edge.

[0010] Furthermore, the signal processing unit includes a metal housing, which is fixedly disposed on the lower right side of the outer edge of the lens ring. A multilayer printed circuit board is fixedly disposed inside the metal housing, and an image processing chip, a machine learning model storage chip, and a microcontroller are integrated and connected on the multilayer printed circuit board.

[0011] Furthermore, the operation execution unit is mounted on the head of the patient's bed via a fixed bracket. The execution unit includes a plastic shell, on the surface of which a high-resolution display screen is fixedly mounted. Inside the plastic shell, a voice playback module and an operation microcontroller are fixedly mounted, and the operation microcontroller is connected to the high-resolution display screen and the voice playback module.

[0012] Furthermore, a power supply module is fixedly installed on the rear side of the plastic casing.

[0013] Furthermore, the adjustable headband is made of highly elastic medical rubber material, and both ends of the adjustable headband are connected to the lens ring via adjustable plastic buckles.

[0014] Furthermore, the micro-screen displays the same content as the high-resolution display screen.

[0015] The beneficial effects of this invention are as follows:

[0016] This invention utilizes a communication aid device equipped with eye signal capture, intelligent processing, and two-way interaction capabilities. A suitable headband frame ensures stable wear, and a high-definition capture module and recognition algorithm translate simple eye movements into clear information about the patient's needs. This helps emergency patients with limited hand movement to communicate physical discomfort or basic needs. Two-way display and voice prompts allow patients to confirm the accuracy of their instructions, while enabling medical staff to obtain information promptly. This reduces delays caused by poor communication, minimizes information bias in nursing care, alleviates the burden on medical staff, and enhances the practicality and safety of communication in emergency care. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall device structure;

[0018] Figure 2 A schematic diagram of the eye signal capture frame structure;

[0019] Figure 3 Top view of the eye signal capture frame;

[0020] Figure 4 Remove the rear view of the adjustable headband from the eye signal capture frame;

[0021] Figure 5 This is a diagram of the internal structure of the signal processing unit.

[0022] Explanation of reference numerals in the attached figures:

[0023] 1. Eye signal capture frame; 101. Adjustable headband; 102. Lens ring; 103. Connecting bridge; 104. Miniature screen; 106. Ultra-small high-definition flat camera; 107. Nanoscale infrared light-emitting diode array; 109. Miniature LiDAR sensor; 110. Pupil tracking and motion recognition module; 111. Precision adjustment bracket; 112. Flexible circuit board; 113. Fixing bracket; 114. Shielded cable; 115. Adjustable plastic buckle; 2. Signal processing unit; 201. Image processing chip; 202. Machine learning model storage chip; 203. Microcontroller; 204. Metal casing; 205. Multilayer printed circuit board; 206. Wireless communication module; 3. Operation execution unit; 301. High-resolution display screen; 4. Power supply module. Detailed Implementation

[0024] like Figure 1-5 As shown, one embodiment of the present invention is an emergency patient communication assistance device based on eye signal recognition, which includes an eye signal capturing frame 1, a signal processing unit 2, an operation execution unit 3, and a power supply module 4. The signal processing unit 2 is fixedly disposed below the eye signal capturing frame 1, the operation execution unit 3 is located outside the eye signal capturing frame 1, and the power supply module 4 is fixed to the rear side of the operation execution unit 3.

[0025] The eye signal capturing frame 1 consists of lens rings 102 and adjustable headbands 101. The two lens rings 102 are connected by a connecting bridge 103, and the two ends of the connecting bridge 103 are welded and fixed to the middle of the outer side wall of the two lens rings 102 respectively. The adjustable headbands 101 are made of medical rubber material with good elasticity, with a thickness of 2mm and a width of 3cm. One end of each of the two adjustable headbands 101 is snapped to the rear end of the outer side wall of one lens ring 102 by an adjustable plastic buckle 115 (the adjustable plastic buckle 115 has an adjustment range of 0.5cm in each range), and the other end is connected by Velcro.

[0026] The lens ring 102 houses an ultra-miniature high-definition flat camera 106 and a pupil tracking and motion recognition module 110. The ultra-miniature high-definition flat camera 106 includes a main camera and an auxiliary camera: the main camera has a resolution of 1920×1080 and a frame rate of 30fps, and is fixed inside the lens ring 102 at the top via a precision adjustment bracket 111 (one end of the precision adjustment bracket 111 is fixed to the inner wall of the lens ring 102 with an M1.5 screw, and the other end is bonded to the main camera); the auxiliary camera has a resolution of 1280×720 and a frame rate of 30fps, is fixed 1cm directly below the main camera, and is soldered to the main camera via a 0.5cm wide flexible circuit board 112. An array of eight nanometer-scale infrared light-emitting diodes (LEDs) 107 (four on the inner side of each lens ring 102, arranged at a 90° angle) are evenly distributed around the inner surface of the lens ring 102. Each LED measures 1mm × 0.5mm, has a power of 0.01W, and a wavelength of 850nm. They are fixed to the inner wall of the lens ring 102 with conductive adhesive. The pupil tracking and motion recognition module 110 measures 10mm × 10mm × 3mm and is fixed directly above the lens ring 102 (0.5cm from the upper edge of the lens ring 102). It is fixed to the outer wall of the lens ring 102 with screws via an L-shaped bracket. The sampling frequency is 50Hz, and the pupil center coordinate measurement accuracy is ±0.5 pixels. An image sensor of model OV5640 is also fixed to the side wall of the lens ring 102. It is bonded to the inner side wall of the lens ring 102 with thermally conductive adhesive (thermal conductivity ≥1W / (m・K)) and is welded to the main camera and auxiliary camera through shielded cable 114 (cable diameter 1mm, shielding layer is aluminum foil + braided mesh).

[0027] A miniature screen 104, measuring 2.5 inches with a resolution of 800×480, is glued and fixed to the front end of the lens ring 102. A miniature lidar sensor 109 is fixed to the upper outer side of the lens ring 102, with a ranging range of 0.1-1m and an accuracy of ±2mm. It consists of a left lidar sensor and a right lidar sensor, which are fixed to the upper outer side of the two lens rings 102 near the edge (0.3cm from the upper edge of the lens ring 102) by clips.

[0028] The metal housing 204 of the signal processing unit 2 is made of aluminum alloy with a thickness of 1mm. It is fixed to the lower right side of the right lens ring 102 (1cm away from the outer edge of the lens ring 102) and is secured to the lens ring 102 with screws. The four-layer printed circuit board 205 (each layer is 0.2mm thick) inside the metal housing 204 is fixed to the inner side wall of the housing by four nylon pillars. The image processing chip 201, the machine learning model storage chip 202, the microcontroller 203 and the wireless communication module 206 are integrated on it. Specifically, the image processing chip 201 is a TMS320DM8168, the machine learning model storage chip 202 is an MX25L12805D, the microcontroller 203 is an STM32F407IGH6, and the wireless communication module 206 is an ESP8266. These components are all soldered to the multilayer printed circuit board 205. The wireless communication module 206 has a communication distance of 100m, a transmission rate of 150Mbps, and a communication frequency of 2.4GHz. It is used to realize the wireless connection between the signal processing unit 2 and the ultra-small high-definition flat camera 106, the pupil tracking and motion recognition module 110, the miniature screen 104, and the operation execution unit 3.

[0029] The machine learning model storage chip 202 stores a pre-trained simplified convolutional neural network model, whose three-layer core structure is as follows:

[0030] Input layer: Receives a 224×224 pixel grayscale image of the eye (1 channel). The pixel values ​​need to be normalized using the formula x'=x / 255 (where x is the original pixel value, ranging from 0 to 255; x' is the normalized value, ranging from 0 to 1).

[0031] Convolutional layer 1: Contains 32 3×3 convolutional kernels (K1) with a stride of 1; after processing by the ReLU activation function (f(x)=max(0,x), where x is the convolution result), it outputs a 112×112 feature map through 2×2 max pooling (taking the maximum value within the window).

[0032] Convolutional layer 2: Contains 64 3×3 convolutional kernels (K2) with a stride of 1; after processing by the ReLU activation function, it outputs a 56×56 feature map through 2×2 max pooling.

[0033] Fully connected layer: contains 128 neurons, processed by the ReLU activation function, with a dropout rate of 0.2 (randomly masking 20% ​​of neurons to prevent overfitting).

[0034] Output layer: Contains 4 neurons. The probability value of outputting 4 instructions is calculated by the Softmax function, and the formula is P(i)=e^z_i / Σe^z_j (where z is the inactive output, and i and j are the neuron numbers, which take values ​​from 1 to 4).

[0035] The training data for this model came from 80 emergency room patients (60 for training and 20 for validation), covering different genders and eye characteristics aged 20-70. Data acquisition used the same ultra-compact high-definition flat camera 106 and pupil tracking and motion recognition module 110 as the device, under illumination conditions of 50-300 lux. Each action was captured 5 times, specifically including: 600 sets of single blinks (0.2-0.4 seconds), 600 sets of blinks with the left eye alone, 600 sets of blinks with the right eye alone, 600 sets of blinks with eyes closed for 2 seconds, and 800 sets of interference actions (such as natural blinks). The acquired images underwent the following preprocessing: cropping to a 224×224 eye region; grayscale conversion using the formula Gray=0.299R+0.587G+0.114B (where R, G, and B are the three-channel pixel values); and noise reduction using median filtering with a 3×3 window. The training set was expanded to 3200 sets through data augmentation methods such as brightness ±10%, horizontal flipping, and ±5° rotation.

[0036] The model training uses the cross-entropy loss function, with the formula L = -(y1 × logŷ1 + y2 × logŷ2 + y3 × logŷ3 + y4 × logŷ4) (where y is the true label, with a value of 1 for a specific instruction and 0 for the rest; ŷ is the model's predicted probability), and uses the Adam optimizer with a learning rate of 0.001. The weight update formula is w = w - η × m / ( +ε) (where w is the weight, η is the learning rate, m is the first moment estimate, v is the second moment estimate, and ε = 1e-8). The training termination condition is: the fluctuation of the validation set accuracy over 10 consecutive iterations is ≤0.5%, or the number of iterations reaches 300; finally, the model with the highest validation set accuracy is saved.

[0037] The contents stored in the machine learning model storage chip 202 include:

[0038] Structural parameters: 32 3×3×1 convolutional kernels K1, 64 3×3×32 convolutional kernels K2, fully connected layer weight matrix (128×(64×56×56)), output layer weight matrix (4×128), all stored in 16-bit floating-point format.

[0039] Command mapping table: Neuron 1 corresponds to command 001 (needs help), Neuron 2 corresponds to command 002 (feeling unwell), Neuron 3 corresponds to command 003 (physiological needs), Neuron 4 corresponds to command 004 (emergency call).

[0040] Judgment threshold: When the probability of a certain instruction is ≥0.8, it is judged as a valid instruction; otherwise, the re-entry mechanism of "quickly blinking twice" (interval ≤0.2 seconds) is triggered.

[0041] The operation execution unit 3 is mounted on the head of the patient's bed via a fixing bracket 113. The fixing bracket 113 is made of stainless steel and its length is adjustable within the range of 50-80cm. One end of the bracket is fixed to the bottom of the operation execution unit 3 with screws, and the other end is welded to the head of the bed. A high-resolution display screen 301 (10.1 inches, resolution 1920×1080) is fixed on the surface of the plastic shell (ABS engineering plastic) of the operation execution unit 3. Inside, a voice playback module (3W power, adjustable volume range 0-80 decibels) and an operation microcontroller (model STM32F103C8T6) are fixed. The operation microcontroller is welded to the high-resolution display screen 301 and the voice playback module via wires. The micro screen 104 displays the same content as the high-resolution display screen 301, and both have a refresh rate of 60Hz.

[0042] The power supply module 4 is a 5000mAh rechargeable lithium battery pack with a voltage of 3.7V. It is fixed to the plastic shell of the operation execution unit 3 with screws on the back side and is welded to the signal processing unit 2, the ultra-small high-definition flat camera 106, the pupil tracking and motion recognition module 110, the micro screen 104 and the operation execution unit 3 respectively through wires. The continuous power supply time is not less than 8 hours.

[0043] The specific implementation steps of this embodiment are as follows:

[0044] Medical staff place the eye signal capture frame 1 on the patient's head. By adjusting the adjustable plastic buckle 115 on the adjustable headband 101 (0.5cm increments), the distance between the lens ring 102 and the patient's eyes is maintained at 3-5cm. This ensures that the ultra-small high-definition flat camera 106 can clearly capture eye images, and the miniature screen 104 is positioned 3-5cm in front of the patient's eyes. Simultaneously, the operation execution unit 3 is fixed to the patient's bedside using the fixing bracket 113 (bracket length adjusted to 60cm). After connecting the power supply module 4, the power switch is turned on. After each component performs a 3-second self-test, the indicator light remains constantly lit, indicating normal startup. At this time, the miniature screen 104 automatically displays four basic command icons and corresponding action demonstration animations (each animation lasting 3 seconds): "blink once" corresponds to "need help," "left eye blink" corresponds to "discomfort," "right eye blink" corresponds to "physiological needs," and "close eyes for 2 seconds" corresponds to "emergency call."

[0045] A nanoscale infrared LED array 107 emits 850nm, 500lux infrared light to illuminate the patient's eyes. The main camera and auxiliary camera simultaneously capture images of the eyes (the main camera covers the entire eye area, while the auxiliary camera captures details of the eyelids). The images are transmitted to the image sensor via a shielded cable 114, and after noise reduction and enhancement processing, one frame is transmitted to the signal processing unit 2 every 0.1 seconds. The pupil tracking and motion recognition module 110 tracks pupil movement and eye actions in real time at a frequency of 50Hz, transmitting data such as pupil diameter changes, eyeball rotation angle, and blink duration to the signal processing unit 2 every 0.1 seconds. A miniature lidar sensor 109 detects the distance between the eyes and the camera in real time. When the distance deviates from the range of 3-5cm, a deviation signal is sent every 0.5 seconds, and a miniature screen 104 simultaneously displays a "Adjust headband" prompt.

[0046] The image processing chip 201 further processes the received image, extracting features such as pupil size (accuracy ±0.1mm), blink frequency (accuracy ±0.1 times / second), and eye movement direction (horizontal ±1°, vertical ±1°), and converting them into a 224×224 grayscale image. The microcontroller 203 reads model parameters from the machine learning model storage chip 202 via the SPI interface (loading time ≤0.1 seconds), and calls the model to analyze and recognize the grayscale image. The specific process is as follows:

[0047] Convolutional layer: Calculates the cross-correlation between the input image and the convolution kernel (C=Σ(I×K), where C is the output feature, I is the input image, K is the convolution kernel, and Σ is the summation), and then performs ReLU activation and pooling.

[0048] Fully connected layer: Performs matrix multiplication (y=W×x+b, where y is the output, W is the weight matrix, x is the input vector, and b is the bias).

[0049] Output layer: Calculate the probability of the four instructions according to the Softmax formula.

[0050] The microcontroller 203 uses 16-bit fixed-point arithmetic and has a main frequency of 168MHz, ensuring a total response time of ≤0.3 seconds.

[0051] The recognized instruction is transmitted to the operation execution unit 3 via the wireless communication module 206. The operation microcontroller controls the high-resolution display screen 301 to display the corresponding text (such as "Need Help", in bold red font, size 40), and simultaneously controls the voice playback module to broadcast the instruction at a rate of 120 words per minute and 65 decibels. The patient can confirm the instruction through the miniature screen 104. If the instruction is incorrectly recognized, the patient only needs to blink twice quickly (with an interval of ≤0.2 seconds), and the device will clear the current instruction and redisplay the basic instruction icon.

[0052] The four sets of core commands used by the patient were all single-step actions:

[0053] A single blink (lasting 0.2-0.4 seconds) → Instruction 001 (needs help), and medical staff will proactively inquire about the specific needs.

[0054] Blink the left eye once (keep the right eye open) → Instruction 002 (discomfort), the screen will simultaneously display the sub-options "pain / numbness / itching", and the patient can further select by "selecting the left option when blinking the left eye, and selecting the right option when blinking the right eye".

[0055] Blink your right eye once (keep your left eye open) → Command 003 (physiological needs), and the screen will simultaneously display the sub-options "drink water / turn over / defecate", and the selection method is the same as above.

[0056] Close your eyes for 2 seconds and then open them → Command 004 (Emergency Call) will trigger an audible and visual alarm, and the screen will flash red.

[0057] Next, medical staff can click the "Parameter Settings" icon on the operation execution unit 3 and enter the password "123456" to enter the settings interface. For example, when adjusting the shooting angle of the ultra-small high-definition flat camera 106, the shooting angle can be increased by extending the precision adjustment bracket 111 (range 0-2cm, accuracy 0.1cm) by 0.5cm, and decreased by 0.5cm by 5°. When the power supply module 4's battery level is below 20%, the high-resolution display screen 301 displays a red battery icon and the text "Low battery, please charge," and the voice playback module provides a reminder every 60 seconds at a volume of 70 decibels. Charging uses a 5V, 2A voltage and current, taking approximately 2 hours to fully charge, and the screen displays the charging progress from 0-100% in real time.

[0058] The beneficial effects of this embodiment are as follows:

[0059] The instructions are simplified into four single steps, with on-screen icon prompts, which greatly reduces the patient's memory burden. It is especially suitable for emergency patients who are confused or physically weak. The single-step action design reduces the rate of operational errors, shortens the recognition response time, and improves communication efficiency.

[0060] The lens ring 102 and adjustable headband 101 of the eye signal capture frame 1 maintain stability, the ultra-small high-definition flat camera 106 and pupil tracking and motion recognition module 110 optimize the algorithm for single eye movements, and the nanoscale infrared light-emitting diode array 107 provides stable light to ensure higher accuracy of single eye motion recognition.

[0061] The signal processing unit 2 adopts a simplified machine learning model to balance recognition speed and false judgment rate. The microcontroller 203 has a simple process and the wireless communication module 206 has lower transmission latency.

[0062] The high-resolution display 301 of the operation execution unit 3 features a large font and sub-option design, coupled with slow voice broadcast, which makes it easy for medical staff to quickly understand the needs, and the instant feedback of the micro screen 104 enhances the patient's confidence in using it.

[0063] The low battery reminder interval of power supply module 4 is reasonable, and the continuous working time is ≥8 hours, which is suitable for emergency nursing cycle.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An emergency patient communication assistance device based on eye signal recognition, characterized in that: The device includes an eye signal capture frame, inside which is installed an ultra-small high-definition flat camera and a pupil tracking and motion recognition module. A miniature screen is fixedly installed at the front end of the eye signal capture frame, and a signal processing unit is fixedly installed below the eye signal capture frame. An operation execution unit is installed outside the eye signal capture frame. The signal processing unit communicates wirelessly with the ultra-small high-definition flat camera, the pupil tracking and motion recognition module, the miniature screen, and the operation execution unit.

2. The emergency patient communication assistance device based on eye signal recognition according to claim 1, characterized in that: The eye signal capturing frame includes a lens ring and an adjustable headband. The lens ring is connected by a connecting bridge. The ultra-small high-definition flat camera consists of a main camera and an auxiliary camera. The main camera is fixedly mounted inside the lens ring at the top via a precision adjustment bracket. The auxiliary camera is fixedly mounted below the main camera. A nano-scale infrared light-emitting diode array is evenly distributed around the inner surface of the lens ring. The pupil tracking and motion recognition module is fixedly mounted directly above the lens ring. An image sensor is fixedly mounted on the side wall of the lens ring. The image sensor is connected to the main camera and the auxiliary camera.

3. The emergency patient communication assistance device based on eye signal recognition according to claim 2, characterized in that: A miniature lidar sensor is fixedly installed on the upper outer side of the mirror ring. The miniature lidar sensor includes a left lidar sensor and a right lidar sensor installed on the upper outer side of the mirror ring near the edge.

4. The emergency patient communication assistance device based on eye signal recognition according to claim 3, characterized in that: The signal processing unit includes a metal housing, which is fixedly disposed on the lower right side of the outer edge of the lens ring. A multi-layer printed circuit board is fixedly disposed inside the metal housing, and an image processing chip, a machine learning model storage chip, and a microcontroller are integrated and connected on the multi-layer printed circuit board.

5. The emergency patient communication assistance device based on eye signal recognition according to claim 4, characterized in that: The operation execution unit is mounted on the head of the patient's bed via a fixed bracket. The execution unit includes a plastic shell, on the surface of which a high-resolution display screen is fixedly mounted. Inside the plastic shell, a voice playback module and an operation microcontroller are fixedly mounted. The operation microcontroller is connected to the high-resolution display screen and the voice playback module.

6. The emergency patient communication assistance device based on eye signal recognition according to claim 5, characterized in that: A power supply module is fixedly installed on the rear side of the plastic shell.

7. The emergency patient communication assistance device based on eye signal recognition according to claim 2, characterized in that... The adjustable headband is made of highly elastic medical rubber material, and both ends of the adjustable headband are connected to the lens ring via adjustable plastic buckles.

8. The emergency patient communication assistance device based on eye signal recognition according to claim 6, characterized in that... The miniature screen displays the same content as the high-resolution display screen.

Images