AI blood collection decompression handgrip system and application method

Abstract

The application provides an AI blood collection decompression grip ware system and an application method, and the system comprises a sensor matrix module, a main control processing module connected with the sensor matrix module, a voice interaction module connected with the main control processing module, and a feedback module connected with the main control processing module. The main control processing module constructs a three-dimensional grip curve surface model based on a pressure signal, calculates a blood vessel exposure index, dynamically adjusts a sensor weight according to patient classification, and forms a closed-loop control from signal collection to individualized intervention. The application reduces the workload of nurses through an automatic process, reduces the puncture failure rate through accurate grip evaluation, realizes accurate decompression guidance through clustering typing, and relieves psychological stress through multi-modal interaction. The standardized voice prompt and intelligent guidance reduce noise pollution in the blood collection area, improve the environmental comfort, and liberate nurses from repeated guidance, so that the nurses can pay more attention to core links such as sterile operation and improve the blood collection efficiency of the hospital.

Description

Technical Field

[0001] This invention relates to the field of blood collection grip strength device technology, and in particular to an AI blood collection decompression grip strength device system and its application method. Background Technology

[0002] I. Technical Pain Points in Existing Outpatient Blood Collection Scenarios

[0003] Large patient base and complex environment:

[0004] Outpatient blood collection is characterized by a high number of patients, a large number of fasting patients, and a high proportion of elderly patients and patients with underlying diseases, resulting in long waiting times during peak hours. Furthermore, hospital infection control requirements necessitate the installation of glass partitions between nurses and patients, leading to reduced volume in voice communication. Frequent personnel movement and noise in the blood collection area (such as conversations and the sound of equipment (phones, etc.)) further hinder the transmission of instructions, with some patients unable to hear nurses' instructions due to environmental noise.

[0005] Patient operational difficulties and psychological stress:

[0006] Patients commonly experience anxiety during blood collection, manifesting as sweaty palms and muscle stiffness, leading to improper fist clenching (such as using only the fingers without engaging the palm), and insufficient vein exposure. Traditionally, nurses manually tap the patient's veins to make them more visible, but this action exacerbates patient fear, and some patients experience increased anxiety due to the tapping motion.

[0007] Limitations of existing grip strength assist devices:

[0008] Traditional hand grip strengtheners are mostly simple mechanical structures that only provide grip strength training, and have the following drawbacks:

[0009] No intelligent interaction: It cannot automatically recognize grip strength status, requiring nurses to provide verbal guidance throughout the process, resulting in a large amount of repetitive work;

[0010] Lack of personalization: Uniform instructions cannot adapt to the different grip strength characteristics of different patients (such as weak type, finger strength type), and some patients cannot understand the "degree of force" which leads to ineffective fist clenching;

[0011] No stress-relief design: It does not take into account the psychological needs of patients and cannot alleviate their tension, especially for patients with hearing impairments or language barriers who lack support.

[0012] II. Shortcomings of the Existing Technology and the Motivation for Improvement in this Invention

[0013] Limitations of sensor technology:

[0014] Existing hand grippers mostly use single-point pressure detection, which cannot locate defects in grip force distribution (such as excessive finger force), resulting in inaccurate assessment of vascular exposure, reliance on nurses' experience to make judgments, and a high failure rate of puncture.

[0015] The disconnect between algorithms and clinical needs:

[0016] The lack of an algorithmic model that quantitatively correlates grip strength data with vascular exposure makes it impossible to provide nurses with an objective reference for blood collection timing; the absence of patient cluster analysis makes it difficult to achieve personalized guidance, resulting in limited effectiveness of a "one-size-fits-all" intervention.

[0017] The limited variety of interaction methods:

[0018] Traditional voice prompts follow a fixed procedure and do not support dynamic adjustments based on grip strength. Furthermore, they lack multilingual support and cannot resolve cross-language communication barriers, making it difficult for patients with dialects or foreign nationals to understand instructions. Summary of the Invention

[0019] To address the limitations of existing sensor technologies, the disconnect between algorithms and clinical needs, and the simplistic interaction methods, this invention provides an AI-powered blood collection decompression grip device method and system. The technical solution is as follows:

[0020] On the one hand, an AI-powered blood collection and decompression gripper system is provided, comprising:

[0021] The sensor matrix module is used to collect pressure signals and grip strength posture signals from the five fingers and palm area. The sensor matrix module includes five pressure sensors distributed on the fingertips, three pressure sensors in three areas of the palm, and a triaxial accelerometer. The main control processing module is connected to the sensor matrix module and is used to filter and extract features from the collected signals, identify grip strength state through a built-in LSTM neural network model, and classify patients using the K-means algorithm.

[0022] A voice interaction module, connected to the main control processing module, is used to output standardized voice prompts based on grip strength and patient classification results.

[0023] The feedback module, connected to the main control processing module, is used to provide tactile feedback through LED indicators and vibration motors. The main control processing module constructs a three-dimensional grip force surface model based on pressure signals, calculates the vascular exposure index, and dynamically adjusts the sensor weights according to patient classification, forming a closed-loop control from signal acquisition to personalized intervention.

[0024] Optionally, the pressure sensor is an ultra-thin flexible piezoresistive sensor; the triaxial accelerometer is used to detect the gripping action to trigger voice identity verification.

[0025] Optionally, the main control processing module includes a processor with an integrated communication unit for interacting with the nurse terminal to receive blood collection timing signals.

[0026] Optionally, the voice interaction module pre-stores Mandarin, dialect, and English voice clips, and the feedback module provides tactile feedback through a vibration motor when the fist is clenched effectively.

[0027] Optionally, the main control processing module adjusts the sensor weights based on the patient classification results, reducing the weight of the finger sensor for patients with strong finger strength and increasing the weight of the palm sensor for patients with weak finger strength.

[0028] On the other hand, a method for applying an AI blood collection decompression gripper is provided, characterized in that the method is applicable to any of the aforementioned AI blood collection decompression gripper systems, and the method includes:

[0029] S1. Grip force and posture signals are simultaneously collected by eight pressure sensors and accelerometers distributed on the five fingers and palm to generate raw data sequences;

[0030] S2. Perform filtering preprocessing on the original data sequence to extract time-domain and frequency-domain features and form a feature vector;

[0031] S3. Construct a three-dimensional grip force surface model based on the pressure signal in the feature vector, and calculate the vascular exposure index corresponding to the pressure in the middle of the palm.

[0032] S4. Input the feature vector into the LSTM neural network to identify the grip strength state, and classify the patients based on the grip strength statistics in the feature vector using the K-means algorithm;

[0033] S5. Based on the grip strength state, vascular exposure index and patient classification results, dynamically adjust the sensor weights and output personalized voice guidance and tactile feedback. When the grip strength state is effective and the vascular exposure index is ≥80, trigger a blood collection prompt.

[0034] Optionally, in the filtering preprocessing, the Kalman filter state equation for the acceleration signal is:

[0035]

[0036] Among them, z t Let G be the vector of observations at time t, and let w be the control matrix. t For process noise, x t Let v be the state vector. t To observe noise, the transition matrix F is a 6×6 identity matrix, and the observation matrix H is the first 3 rows of the identity matrix;

[0037] Perform a 7-point sliding median filter on the pressure signal:

[0038] The result is after 7-point moving median filtering, p iThe signal is a pressure signal, and median indicates the median operation.

[0039] Optionally, the process of constructing the three-dimensional grip force surface model is as follows:

[0040] Define the initial weight vector w 0 = [0.12, 0.12, 0.12, 0.12, 0.12, 0.1, 0.3, 0.1], where 0.3 is the sensor weight in the middle of the palm, the five 0.12 are the sensor weights of the five fingers respectively, and the two 0.1 are the sensor weights of the proximal end of the palm near the wrist and the distal end of the palm near the base of the fingers respectively;

[0041] The formula for calculating overall grip strength is:

[0042] Optionally, based on pressure in the center of the palm By fitting the quadratic relationship between the pressure and vessel diameter D(t), a pressure-vessel diameter model is obtained:

[0043]

[0044] Optionally, the blood vessel diameter is mapped to an exposure index E(t) ranging from 0 to 100:

[0045]

[0046] Among them, the mean value of not clenching the fist is D min =1.2mm, mean value D under optimal exposure conditions max =3.8mm, when E(t)≥80, it indicates to the nurse that blood can be drawn.

[0047] On the other hand, an AI blood collection and decompression grip device is provided, the AI ​​blood collection and decompression grip device comprising: a processor; a memory, the memory storing computer-readable instructions, which, when executed by the processor, implement any of the methods described above for the AI ​​blood collection and decompression grip device.

[0048] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored therein, the at least one instruction being loaded and executed by a processor to implement any of the methods in the above-described AI blood collection decompression gripper methods.

[0049] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0050] Automated processes reduce nurses' workload: By automating the entire process of picking up a hand gripper to automatically trigger identity verification, activating a tourniquet to automatically trigger a fist-clenching command, and automatically prompting the release of the fist upon completion of blood collection, the time nurses spend giving verbal instructions for each blood collection is shortened, increasing the number of patients that can be processed per day.

[0051] Precise grip strength assessment reduces puncture failure rate: An 8-channel pressure sensor matrix combined with LSTM state recognition accurately determines the "effective fist" state. Combined with quantitative assessment of vascular exposure index, it improves the recognition rate of the best blood collection time and reduces the puncture failure rate shown in clinical data.

[0052] Clustering and classification enable precise stress reduction guidance: Based on the K-means algorithm, patients are divided into 4 categories, and targeted voice and tactile feedback is provided (such as enhanced vibration prompts for patients with weak strength, and guidance on palm force for patients with weak finger strength), which improves the success rate of effective fist clenching and significantly improves the cooperation of elderly patients.

[0053] Multimodal interaction alleviates psychological stress: The voice module supports Mandarin, dialects, and English, covering the language needs of most patients. Hearing-impaired patients can understand instructions through vibration feedback (motor vibration when effectively clenching a fist), reducing the misunderstanding rate caused by communication barriers; it avoids nurses manually tapping blood vessels, reducing patient anxiety, and significantly reducing subjective stress during the blood collection process.

[0054] Synergistic advantages of sensor matrix and algorithm: The fusion perception of 8-channel pressure sensor and accelerometer constructs a "space-time" multi-dimensional grip force model. Compared with traditional single-point detection, it improves the accuracy of grip force distribution defect identification and can accurately locate problems such as "excessive finger force" and "insufficient palm force".

[0055] The closed-loop feedback mechanism enhances system adaptability: the patient clustering results adjust the sensor weights in reverse, forming an adaptive closed loop of "detection-identification-adjustment". The system's adaptability to different patients is significantly enhanced, and the effective fist-clenching learning time for first-time users is shortened.

[0056] Medical process optimization: Standardized voice prompts and intelligent guidance reduce noise pollution in the blood collection area, improving environmental comfort; nurses are freed from repetitive instructions and can devote more energy to core processes such as aseptic operation, thereby improving the efficiency of blood collection in the hospital.

[0057] Expanded application scenarios: The system supports multiple languages ​​and multiple patient types (elderly, children, cognitive impairment), making it especially suitable for resource-limited scenarios such as health check-up centers and community hospitals. It reduces reliance on nurses' experience and has broad clinical application value. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1This is a schematic diagram of the AI ​​blood collection decompression grip device system involved in this invention;

[0060] Figure 2 This is a schematic diagram illustrating the application method of the AI ​​blood collection decompression gripper involved in this invention;

[0061] Figure 3 This is a schematic diagram of a grip strengthener for the left hand involved in the present invention;

[0062] Figure 4 This is a schematic diagram of a right-handed grip strengthener involved in the present invention;

[0063] Figure 5 This is a schematic diagram showing the location of the buttons or touch keys involved in this invention;

[0064] Figure 6 This is a schematic diagram of the structure of an AI blood collection decompression grip device provided in an embodiment of the present invention.

[0065] Figure label:

[0066] 1. Grip strengthener body; 2. Finger groove; 3. Pressure sensor; 4. Recess; 5. Button or touch button; 6. Flexible area. Detailed Implementation

[0067] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0068] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0069] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0070] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0071] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0072] Figure 1 This is a block diagram of an AI blood collection decompression gripper system 100 provided in an embodiment of the present invention. This system is used for an AI blood collection decompression gripper method. (Refer to...) Figure 1 The system includes: a sensor matrix module 110, a main control processing module 120, a voice interaction module 130, and a feedback module 140, wherein:

[0073] The sensor matrix module 110 is used to collect pressure signals and grip strength sensor posture signals in the five fingers and palm areas. The sensor matrix module includes five pressure sensors 3 distributed on the fingertips, three pressure sensors 3 in the three areas of the palm, and a triaxial accelerometer.

[0074] The main control processing module 120 is connected to the sensor matrix module and is used to filter and extract features from the collected signals, identify grip strength state through the built-in LSTM neural network model, and classify patients through the K-means algorithm.

[0075] The voice interaction module 130 is connected to the main control processing module and is used to output standardized voice prompts based on grip strength and patient classification results.

[0076] Feedback module 140, connected to the main control processing module, is used to provide tactile feedback through LED indicator lights and vibration motors; the main control processing module constructs a three-dimensional grip force surface model based on pressure signals, calculates the vascular exposure index, and dynamically adjusts the sensor weights according to patient classification to form a closed-loop control from signal acquisition to personalized intervention.

[0077] In this specification, the pressure sensor 3 is an ultra-thin flexible piezoresistive sensor, and the triaxial accelerometer is used to detect the gripper picking up action to trigger voice identity verification.

[0078] In some embodiments, the main control processing module 120 includes a processor with an integrated communication unit for interacting with the nurse terminal to receive blood collection timing signals.

[0079] In some embodiments, the voice interaction module 130 pre-stores Mandarin, dialect and English voice segments, and the feedback module provides tactile feedback through a vibration motor when the fist is clenched effectively.

[0080] In some embodiments, the main control processing module 120 adjusts the sensor weights according to the patient classification results, reducing the weight of the finger sensor for patients with finger strength type and increasing the weight of the palm sensor for patients with weak strength type.

[0081] This invention provides an application method for an AI-powered blood collection and decompression gripper. This method can be implemented using an AI-powered blood collection and decompression gripper device, which can be a terminal or a server. Figure 1 The flowchart shown illustrates the application method of the AI ​​blood collection decompression gripper. The processing flow of this method may include the following steps:

[0082] S1. Grip force and posture signals are simultaneously collected by eight pressure sensors 3 distributed on the five fingers and palm and an accelerometer to generate raw data sequences;

[0083] S2. Perform filtering preprocessing on the original data sequence to extract time-domain and frequency-domain features and form a feature vector;

[0084] S3. Construct a three-dimensional grip force surface model based on the pressure signal in the feature vector, and calculate the vascular exposure index corresponding to the pressure in the middle of the palm.

[0085] S4. Input the feature vector into the LSTM neural network to identify the grip strength state, and classify the patients based on the grip strength statistics in the feature vector using the K-means algorithm;

[0086] S5. Based on the grip strength state, vascular exposure index and patient classification results, dynamically adjust the sensor weights and output personalized voice guidance and tactile feedback. When the grip strength state is effective and the vascular exposure index is ≥80, trigger a blood collection prompt.

[0087] In some embodiments, the Kalman filter state equation for the acceleration signal in the filtering preprocessing is:

[0088]

[0089] Where z t Let G be the vector of observations at time t, and let w be the control matrix. t For process noise, x t Let v be the state vector. t To observe noise, the transition matrix F is a 6×6 identity matrix, and the observation matrix H is the first 3 rows of the identity matrix;

[0090] Perform a 7-point sliding median filter on the pressure signal:

[0091] The result is after 7-point moving median filtering, p i The signal is a pressure signal, and median indicates the median operation.

[0092] In some embodiments, the process of constructing a three-dimensional grip force surface model is as follows:

[0093] Define the initial weight vector w 0 = [0.12, 0.12, 0.12, 0.12, 0.12, 0.1, 0.3, 0.1], where 0.3 is the sensor weight in the middle of the palm, the five 0.12 are the sensor weights of the five fingers respectively, and the two 0.1 are the sensor weights of the proximal end of the palm near the wrist and the distal end of the palm near the base of the fingers respectively;

[0094] The formula for calculating overall grip strength is:

[0095] In some embodiments, based on pressure in the center of the palm By fitting the quadratic relationship between the pressure and vessel diameter D(t), a pressure-vessel diameter model is obtained:

[0096]

[0097] In some embodiments, the blood vessel diameter is mapped to an exposure index E(t) ranging from 0 to 100:

[0098]

[0099] The mean value of not clenching a fist is D. min =1.2mm, mean value D under optimal exposure conditions max =3.8mm, when E(t)≥80, it indicates to the nurse that blood can be drawn.

[0100] The technical concept of this invention is as follows:

[0101] I. System Architecture and Hardware Implementation (I) Composition and Connection Relationship of Core Modules

[0102] 1. Sensor matrix module hardware composition: 5 Interlink FSR402 piezoresistive sensors (range 0~50N, accuracy ±0.5N), distributed on the fingertips of the thumb, index finger, middle finger, ring finger, and little finger;

[0103] Three Interlink FSR406 piezoresistive sensors (range 0~100N, accuracy ±0.8N) are distributed at the proximal, middle and distal ends of the palm;

[0104] Bosch BMI160 triaxial accelerometer (range ±16g, angular velocity ±2000° / s) is used to detect the posture of a hand gripper.

[0105] The Allegro A1324 Hall sensor is integrated into the tourniquet buckle to detect the tourniquet tightness.

[0106] Data output: The pressure signal (120Hz sampling) and acceleration signal (100Hz sampling) are transmitted to the main control processing module via the I2C / SPI bus.

[0107] 2. Main control processing module

[0108] Core components: ESP32-WROOM-32D processor (dual-core 240MHz, 4MB Flash), built-in BLE5.0 communication unit;

[0109] Data processing link: Receives raw signals from the sensor matrix module, performs analog-to-digital conversion using an ADS1115 16-bit ADC; performs sliding median filtering on the pressure signal and Kalman filtering on the acceleration signal;

[0110] Extract time-domain features (mean, variance, peak value) and frequency-domain features (FFT transform results) to construct an 11-dimensional feature vector; identify grip strength status using a built-in LSTM model and classify patients using the K-means algorithm;

[0111] The sensor weights are dynamically adjusted based on the classification results to calculate the vascular exposure index.

[0112] 3. Voice interaction module

[0113] Hardware connection: Connects to the main control processing module via SPI interface, including VS1053B decoding chip and dual-channel SPK-1612 speakers;

[0114] Storage content: The W25Q128 Flash pre-stores Mandarin, dialect (such as Cantonese, Sichuanese, etc.) and English audio clips, which can be dynamically recalled.

[0115] 4. Feedback Module

[0116] LED indicator: 0603RGB tri-color LED, green indicates a valid fist clench, red indicates an invalid grip;

[0117] Vibration motor: 3V eccentric wheel motor, connected to the main control processing module I / O port, triggers vibration feedback when the fist is clenched effectively.

[0118] 5. Power Management Module

[0119] Component configuration: TP4056 charging chip + 2200mAh 18650 lithium battery, supports USB-C interface 5V / 2A fast charging; Battery life: ≥8 hours of continuous operation (120Hz sampling + voice interaction mode).

[0120] (II) Data transmission relationship between modules

[0121] Sensor matrix → Main control processing: Pressure signals (8 channels) and acceleration signals (3 axes) are transmitted in real time via I2C bus, with a timestamp error of <50ms;

[0122] Main control processing → Voice interaction: Grip strength status and patient classification results trigger the playback of corresponding voice segments via the SPI bus;

[0123] Main control processing → Feedback module: LED control signals and vibration signals are output through GPIO ports;

[0124] Main control processing → Nurse terminal: Transmit grip strength heat map, vascular exposure index (E(t)) and blood collection permission signal (ASCII code "G01") via BLE.

[0125] II. Control Methods (Full Process)

[0126] (I) Step-by-step processing logic

[0127] S1. Synchronous acquisition of multi-source signals

[0128] Execution details: Eight pressure sensors simultaneously collect pressure data from the five fingers and palm at a frequency of 120Hz, generating the raw sequence p. i (t)(i=1~8);

[0129] The triaxial accelerometer acquires attitude signals at 100Hz. x (t), a y (t), a z (t)];

[0130] The Hall sensor detects the tourniquet status and outputs high and low level signals (high level when tightened).

[0131] Data output: The raw data sequence with timestamps, used as input for S2 preprocessing.

[0132] S2. Signal Preprocessing and Feature Extraction

[0133] Filtering:

[0134] Applying a 7-point sliding median filter to the pressure signal eliminates impulse noise, resulting in...

[0135] Applying Kalman filtering to the acceleration signal eliminates hand tremor noise, resulting in...

[0136] Feature calculation:

[0137] Time-domain characteristics: mean pressure μ in each channel i ,variance peak p i，max Rising slope s i ;

[0138] Frequency domain characteristics: Perform FFT transformation on the palm pressure signal to extract the energy concentration frequency bands f1, f2, and f3;

[0139] Output: 11-dimensional feature vector As input to the LSTM in S4.

[0140] S3. Grip Force Surface Modeling and Exposure Calculation

[0141] 3D surface construction:

[0142] Based on the initial weight vector w 0 =[0.12, 0.12, 0.12, 0.12, 0.12, 0.1, 0.3, 0.1], calculate the overall grip strength:

[0143]

[0144] Exposure quantification:

[0145] Based on the pressure in the middle of the palm Calculate vessel diameter using a clinical regression model:

[0146]

[0147] Standardized exposure index from 0 to 100:

[0148] Output: Combined grip strength F(t) and exposure index E(t) are used for S4 cluster feature calculation and S5 blood sampling decision.

[0149] S4. Grip strength recognition and patient clustering

[0150] LSTM state recognition:

[0151] Input: Feature vector X(t) generated by S2, with a time window of 100 time steps;

[0152] Model structure: Two-layer LSTM (128+64 neurons), outputting three state probabilities (not gripped, valid, over-gripped);

[0153] Training data: grip strength sequences from 2000 patients, optimized with cross-entropy loss function, accuracy ≥92%.

[0154] K-means clustering:

[0155] Input feature: average grip strength finger pressure percentage r f-p Exposure standard deviation σ E Effective fist clenching duration t eff ;

[0156] Clustering results: divided into four categories: weak type, finger strength type, tremor type, and standard type, for use in S5 personalized intervention.

[0157] S5. Dynamic Intervention and Blood Collection Decision

[0158] Weight adaptive adjustment:

[0159] For patients with finger strength issues: the weight of the finger sensor is reduced by 25%, forcibly focusing on palm pressure;

[0160] For patients with weak grip strength: The weight of the palm sensor is increased by 20% to 30%, enhancing the sensitivity of grip strength detection.

[0161] Personalized guidance:

[0162] Weak type: Voice prompt "Please clench your fist" + vibration intensity increases with E(t);

[0163] Finger-based: Voice prompt "Please use your palm to exert force" + dynamic decrease in the weight of the finger sensor;

[0164] Trembling type: Voice prompt "Please remain steady and apply force slowly" + reduce the sampling rate to 80Hz;

[0165] Standard version: Voice prompt "Please clench your fist".

[0166] Blood collection trigger conditions:

[0167] LSTM recognizes a "valid fist" state lasting ≥3 seconds;

[0168] E(t) ≥ 80;

[0169] A blood collection permission signal is sent to the nurse's terminal, triggering the indicator light to flash.

[0170] In one specific embodiment, the specific structure of the grip strengthener is as follows: Figure 3 and Figure 4 As shown, Figure 3 For left-handed grip strengtheners, Figure 4 For right-handed grip strengtheners, the main body 1 has finger grooves 2 that match the five fingers. The area containing the finger grooves 2 can be a flexible region 6 made of silicone material to accommodate different finger sizes, such as size and thickness. Corresponding pressure sensors 3 are installed within the finger grooves 2. On the surface of the main body 1, corresponding pressure sensors 3 are located at positions corresponding to the proximal end of the palm (near the wrist), the middle of the palm (area with dense radial arteries), and the distal end of the palm (near the base of the fingers). Figure 5 As shown, the top and / or bottom of the hand gripper body 1 has a recess 4, and a button or touch key 5 is provided in the recess 4. The nurse can actively trigger the "clench fist" or "release fist" voice prompt by pressing the button or touch key 5. Other system hardware (or modules) can be integrated into the hand gripper body 1. Other system hardware (or modules) are as follows: 8-channel flexible sensor array:

[0171] Finger area: 5 Interlink FSR402 piezoresistive sensors (range 0~50N, accuracy ±0.5N) are embedded in the fingertip contact area of ​​the thumb, index finger, middle finger, ring finger and little finger respectively. The thickness is only 0.2mm, which conforms to the curvature of the fingertip and can capture the changes in grip force of a single finger in real time.

[0172] Palm area: 3 Interlink FSR406 sensors (range 0~100N, accuracy ±0.8N) are distributed in a triangular pattern at the proximal end of the palm (near the wrist), the middle of the palm (the area with dense radial artery vessels), and the distal end of the palm (near the base of the fingers). The sensor in the middle of the palm has an area of ​​15mm×15mm and covers the projection area of ​​the main blood collection vessels.

[0173] Auxiliary sensors:

[0174] A Bosch BMI160 triaxial accelerometer (range ±16g, angular velocity ±2000° / s) is deployed at the center of gravity of the grip strengthener to detect the grip strengthener's picking-up action and posture changes, triggering voice interaction.

[0175] The Allegro A1324 Hall sensor, integrated inside the tourniquet buckle, determines whether the tourniquet is tight by detecting changes in magnetic signals, serving as one of the trigger conditions for a fist-clenching command.

[0176] Main control and data processing core: Employs an ESP32-WROOM-32D dual-core processor (240MHz clock speed, 4MB Flash, 8MB PSRAM), which integrates BLE 5.0 and WiFi modules to meet the following requirements:

[0177] Multi-channel synchronous data acquisition: The ADS1115 16-bit ADC (4 channels, up to 860 SPS sampling rate) is controlled via I2C bus to expand to 8 channels of pressure signal synchronous acquisition, with the sampling frequency set to 120Hz to ensure the capture of dynamic changes in grip force; Edge computing capability: Built-in neural network accelerator, supports TensorFlow Lite microcontroller version, directly runs LSTM grip force state recognition model, reducing cloud dependence;

[0178] Low power consumption design: Sleep mode power consumption <500μA, combined with a 2200mAh 18650 lithium battery (3.7V), it can support 8 hours of continuous high-intensity use (120Hz sampling + voice interaction).

[0179] Interaction and power supply module:

[0180] Voice interaction unit: The VS1053B MP3 decoding chip (signal-to-noise ratio >90dB) drives a dual-channel SPK-1612 speaker (16mm diameter, 0.5W power), supporting Mandarin, dialects (such as Cantonese and Sichuanese) and English voice segments stored in W25Q128128Mbit SPI Flash, and can dynamically call different prompt tones.

[0181] Status feedback system: The 0603RGB three-color LED indicator is integrated on the top of the gripper. Green indicates normal power, blue indicates Bluetooth connection, and red indicates grip strength is invalid. At the same time, a built-in vibration motor (3V, eccentric wheel type) provides tactile feedback when effectively clenching the fist.

[0182] Power management: The TP4056 charging chip supports 5V / 2A fast charging, has a USB-C interface, overcharge / over-discharge protection, charging efficiency >90%, and can fully charge a lithium battery in 2 hours.

[0183] The grip strength guidance process is as follows:

[0184] 1. Synchronous data acquisition and hardware-level preprocessing

[0185] Synchronous acquisition mechanism:

[0186] The main control unit configures the ADS1115 to 8-channel differential mode via the I2C protocol, with a sampling rate of 120Hz, simultaneously acquiring pressure signals from 5 fingers + 3 palms. i (t)(i=1~8); simultaneously, the BMI160 outputs triaxial acceleration [a] at 100Hz. x (t),a y (t),a z [(t)], all data with timestamps is cached in ESP32's 8MB PSRAM.

[0187] Key formula: The Kalman filter state equation for acceleration signals is: Where z t Let G be the vector of observations at time t, and let w be the control matrix. t For process noise, state vector v t To eliminate hand tremor noise, the transition matrix F is a 6×6 identity matrix, and the observation matrix H is the first 3 rows of the identity matrix.

[0188] Hardware filtering:

[0189] The ADS1115's built-in PGA amplifies the FSR402's mV-level output signal by 2 times, improving the signal-to-noise ratio; the ESP32 performs a 7-point sliding median filter on the pressure signal. Eliminate pulse interference to ensure the stability of subsequent calculations.

[0190] 2. Three-dimensional grip force surface modeling and dynamic weight allocation

[0191] Basic surface fitting:

[0192] Define the initial weight vector w 0 = [0.12, 0.12, 0.12, 0.12, 0.12, 0.1, 0.3, 0.1], where the sensor in the center of the palm has the highest weight (0.3), reflecting its strong correlation with blood vessel exposure. The comprehensive grip strength formula is:

[0193] Weight adaptive adjustment:

[0194] If the patient's history is clustered as "finger strength type" (τ=2), the finger sensor weight is reduced by 25%.

[0195]

[0196] If it is a "weak type" (τ=1), the palm weight increases by 20% to 30%.

[0197]

[0198] Adaptive weights are stored in W25Q128 and indexed by patient ID to enable personalized testing.

[0199] 3. Quantitative assessment and clinical mapping of vascular exposure

[0200] Pressure-vessel diameter model:

[0201] Based on 1500 ultrasound data, the pressure in the center of the palm was fitted. The quadratic relationship between the vessel diameter D(t) and the vessel diameter D(t):

[0202] (Unit: mm)

[0203] in The pressure value in the middle of the palm is filtered. This model has been clinically validated and has a correlation coefficient R with ultrasound imaging. 2 =0.85.

[0204] Exposure index standardization:

[0205] Mapping blood vessel diameter to an exposure index E(t) ranging from 0 to 100:

[0206] Where D min =1.2mm (average value when not clenching fist), D max =3.8mm (mean value of optimal exposure state), when E(t)≥80, it indicates to the nurse that blood can be drawn.

[0207] 4. LSTM grip strength recognition model (including the entire training process)

[0208] Network architecture design:

[0209] A two-layer LSTM neural network (implemented in TensorFlow Lite) is used, with the input being an 11-dimensional feature vector X(t) over 100 time steps (mean pressure from 8 channels + mean acceleration from 3 axes). The specific structure is as follows:

[0210] Layer 1: 128 LSTM neurons with Dropout (0.2) to prevent overfitting;

[0211] Layer 2: 64 LSTM neurons, with the output connected to a fully connected layer;

[0212] Output layer: 3 nodes (softmax activated), corresponding to the probabilities of the "not grasped", "valid", and "over-grasped" states.

[0213]

[0214] Model training process:

[0215] Dataset: Data from 2000 patients, including:

[0216] Unheld state: average pressure <10N, 2000 sets;

[0217] Effective fist clenching: 30-45 N with even distribution, 3000 sets;

[0218] Excessive fist clenching: >50N or single finger percentage >40%, 1500 sets.

[0219] Optimization objective: Cross-entropy loss function: Where y k Using the ground truth labels (one-hot encoded), and trained for 50 epochs with the Adam optimizer (learning rate 0.001), the validation set accuracy reached 92.3%.

[0220] Real-time inference:

[0221] The state probability is updated every 100 time steps, using the following formula:

[0222]

[0223] Where h t To be in a hidden state, c t In cellular state, W o and b o These are the output layer parameters.

[0224] 5. K-means patient clustering and personalized intervention strategies

[0225] Clustering feature engineering:

[0226] Construct a 6-dimensional feature vector Z, which includes:

[0227] Grip strength statistics: average grip strength Grip strength standard deviation σ F Finger pressure percentage

[0228] Exposure statistics: average exposure Exposure standard deviation σ E Effective fist clenching duration t eff .

[0229] Clustering optimization algorithm:

[0230] Four cluster centers were initialized using K-means++, and the optimization process was iterative:

[0231] Allocation phase: Calculate sample z j To the center c k Euclidean distance:

[0232] Update phase:

[0233] Until the central change is less than 10 -3 The process may terminate after 20 iterations, ultimately resulting in four categories.

[0234] Dynamic intervention logic:

[0235]

[0236] 6. Multi-source data fusion closed-loop control

[0237] Blood collection timing decision:

[0238] The blood collection permission signal is triggered when the following conditions are met:

[0239]

[0240] At the same time, the ASCII code "G01" is sent to the nurse terminal via BLE, triggering the indicator light to flash.

[0241] Real-time weight adjustment:

[0242] For patients with finger strength issues, the finger sensor weights are updated every 500ms:

[0243]

[0244] Where sgn is the sign function, when r f-p When the value is greater than 0.5, the weight decreases, forcibly guiding the palm to exert force.

[0245] The applications of this invention are as follows:

[0246] I. Typical Examples

[0247] (I) Scene of a hospital outpatient blood collection center

[0248] Application process:

[0249] 1. Scenario of an elderly patient (72 years old, mild hearing loss):

[0250] When the patient picks up the hand gripper, the system automatically triggers a voice prompt: "Hello, please tell me your name?", while the LED blue light flashes.

[0251] After the nurse tightens the tourniquet, she presses the hidden button on the side of the hand gripper, triggering the voice prompt: "Please clench your fist." At this time, the patient's palm pressure does not reach the threshold due to insufficient finger strength.

[0252] The system detected that the finger pressure accounted for 65% (finger force characteristic) and gave a voice prompt: "Please use your palm to exert force and reduce the force of your fingers." At the same time, the vibration motor provided feedback at an intensity of 50mV.

[0253] After the patient adjusted their grip strength, the pressure sensor in the middle of the palm detected 32N, the vascular exposure index E(t) = 82, the LSTM recognized it as a valid fist, the system triggered a green LED light and sent a blood collection signal to the nurse terminal;

[0254] After the blood collection is completed, the nurse presses another button, and a voice prompt says, "Please release your fist," while the vibrating motor vibrates briefly once.

[0255] 2. Child patient scenario (6 years old, nervous about first blood draw):

[0256] The system detected that the patient's grip strength fluctuation standard deviation reached 10N (tremor-type characteristics), automatically reduced the sampling rate to 80Hz, and played a children's song clip: "Little friend, slowly apply force like you are kneading clay~"

[0257] When the effective fist clench lasts for 4 seconds and E(t) = 85, the hand gripper plays a voice reward saying "You're great!" while the vibration motor provides feedback with a brisk rhythm to alleviate the child's fear.

[0258] (II) Community Health Service Center Scenario

[0259] Efficient blood collection process:

[0260] During peak hours (8:00-10:00 AM), nurses connect 10 hand grippers in batches via BLE, and the system automatically groups them by patient type.

[0261] Patients with weak strength (such as diabetic patients) are given priority to use high palm weight grip strengtheners, which increase the strength of voice prompts by 20%.

[0262] For patients with finger strength disorder, a hand gripper with a palm groove is provided to physically guide the force exerted by the palm.

[0263] The system provides real-time statistics on blood collection efficiency: the number of patients processed per hour has increased from 28 to 45 under the traditional method, and manual operation by nurses has been reduced by 60% (only puncture and triggering buttons are required).

[0264] II. Quantitative Data on Clinical Efficacy

[0265] (I) Patient Experience Improvement Data

[0266]

[0267] (II) Improvement of blood collection efficiency and quality

[0268]

[0269] (III) Targeted Effects on Special Populations

[0270] Elderly patients (≥65 years old):

[0271] The effective fist clenching rate increased from 62% to 91%, mainly due to:

[0272] The voice prompt volume is automatically increased (+15dB), and vibration feedback is used to compensate for hearing loss; the palm weight is increased by 30% for patients with weak hearing, reducing the strength required to clench their fist.

[0273] Pediatric patients (≤12 years old):

[0274] The rate of crying during blood collection decreased from 43% to 11%. Key measures included:

[0275] Cartoon-style voice interaction (e.g., "To defeat the little bacteria, you need to grip the magic ball tightly!");

[0276] After a tremor test, play soothing music to distract the patient.

[0277] III. Feedback and Efficiency Improvement from the Nurse Application

[0278] Simplified operation process:

[0279] In the traditional procedure, nurses need to complete four steps: "verify name → explain fist clenching → tap blood vessel → determine timing". This procedure reduces the number of steps to three core steps through system automation: "apply tourniquet → puncture → remove needle", shortening the operation time per patient by 55 seconds.

[0280] Decision support optimization:

[0281] The nurse terminal displays the grip strength heat map and E(t) curve in real time, avoiding the need to judge the degree of vascular exposure based on experience, and improving the accuracy of blood collection timing judgment from 78% to 96%.

[0282] In this invention, the "stress reduction" function is achieved through multi-dimensional collaboration involving hardware design, algorithm strategies, and interaction logic, specifically in the following aspects:

[0283] I. Pressure Reduction Mechanisms at the Hardware Design Level

[0284] 1. Ergonomic grip strengthener shape

[0285] The main body 1 of the hand grip strengthener can adopt a "C" or "U" shaped design (referring to the groove 2) that conforms to the curvature of the palm, and the surface is covered with medical-grade silicone, which is soft to the touch and non-slip. Compared with traditional hard hand grip balls, this shape reduces the pressure on the palm and avoids physiological tension caused by grip discomfort. For example, the groove design in the palm area can guide the patient to fit naturally and reduce muscle tension.

[0286] 2. Multi-sensor precise feedback replaces manual intervention

[0287] An 8-channel pressure sensor matrix detects grip force distribution in real time, replacing the traditional method of nurses manually tapping blood vessels. When the patient's grip force is insufficient or unevenly distributed, the system guides the correct action through voice and vibration feedback (such as "Please use your palm to exert force"), avoiding the psychological stress caused by physical contact with nurses.

[0288] II. Strategy for Reducing Algorithm and Interaction Logic

[0289] 1. Personalized grip strength guidance reduces operational anxiety.

[0290] Patient clustering and classification intervention: Using the K-means algorithm, patients are divided into four categories, including weak-fisted and finger-strengthed patients, and targeted voice prompts are provided. For example, weak-fisted patients will hear "Please clench your fist" accompanied by increasing vibration intensity, while finger-strengthed patients will receive precise guidance such as "Reduce the force of your fingers," avoiding frustration caused by operational errors.

[0291] Dynamic weight adjustment: For patients with finger strength deficiency, the weight of the finger sensor is automatically reduced (e.g., by 25%), and the algorithm is forced to focus on palm pressure, guiding the patient to naturally adjust the way of exerting force and reducing the anxiety of "insufficient grip strength".

[0292] 2. Standardized voice interaction alleviates communication pressure.

[0293] Pre-stored Mandarin, dialect, and English voice clips (such as "Please clench your fist" and "Please loosen your fist") resolve communication barriers caused by isolation glass and environmental noise. Hearing-impaired patients can receive vibration feedback (motor vibration when clenching their fist) to aid understanding and avoid tension caused by unclear instructions. Picking up the hand gripper automatically triggers a voice identity verification message, "What is your name?", replacing nurses' repetitive pronouncements, reducing noise pollution in the blood collection area, and improving environmental comfort.

[0294] III. The stress-reducing effect of process optimization and feedback mechanisms

[0295] 1. Objective judgment of blood collection timing

[0296] By using the vascular exposure index E(t) and LSTM status recognition, the system automatically determines the optimal time for blood collection when the patient's fist is effectively clenched and the exposure is ≥80%, avoiding the risk of repeated punctures due to nurses' subjective judgment. Clinical data shows that this mechanism improves the success rate of blood collection and directly reduces the psychological stress on patients caused by failed punctures.

[0297] 2. Closed-loop feedback reduces uncertainty.

[0298] The hand grip strengthener provides real-time tactile feedback through LED indicator lights (green for effective and red for ineffective) and a vibration motor, allowing patients to intuitively perceive whether their movements are correct and reducing their psychological doubts about whether they have done a good job.

[0299] The nurse's device displays a grip strength heatmap and exposure curve, reducing the need to repeatedly confirm whether the patient is cooperating, shortening blood collection preparation time, and reducing anxiety caused by waiting in line.

[0300] In summary, this solution addresses the pain point of "stress caused by manual intervention" in traditional blood collection by using a three-dimensional framework of "hardware comfort design → personalized algorithm guidance → automated process optimization" to reduce stress from three levels: physiological comfort, psychological safety, and operational clarity.

[0301] Figure 6 This is a schematic diagram of the structure of an AI blood collection decompression gripper device provided in an embodiment of the present invention, as shown below. Figure 6 As shown, the AI ​​blood collection decompression gripper device may include the above-mentioned Figure 2 The AI ​​blood collection decompression grip strengthener system is shown. Optionally, the AI ​​blood collection decompression grip strengthener device 410 may include a first processor 2001.

[0302] Optionally, the AI ​​blood collection decompression grip strength device 410 may also include a memory 2002 and a transceiver 2003.

[0303] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0304] The following is combined with Figure 6 The various components of the AI ​​blood collection and decompression grip strength device 410 are described in detail below: The first processor 2001 is the control center of the AI ​​blood collection and decompression grip strength device 410. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0305] Optionally, the first processor 2001 can perform various functions of the AI ​​blood collection decompression gripper device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0306] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 6 CPU0 and CPU1 are shown in the diagram.

[0307] In a specific implementation, as one example, the AI ​​blood collection decompression grip strength device 410 may also include multiple processors, such as... Figure 6 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). Here, "processor" can refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0308] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0309] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the AI ​​blood collection decompression grip device 410. Figure 6 (Not shown in the figure) is coupled to the first processor 2001, and the embodiments of the present invention do not specifically limit this.

[0310] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0311] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0312] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0313] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0314] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An AI-based blood collection and decompression grip strength device system, characterized in that, include: The sensor matrix module is used to collect pressure signals and grip strength sensor posture signals from the five fingers and palm area. The sensor matrix module includes five pressure sensors distributed on the fingertips, three pressure sensors in three areas of the palm, and a triaxial accelerometer. The main control processing module, connected to the sensor matrix module, is used to filter and extract features from the collected signals, identify grip strength status through the built-in LSTM neural network model, and classify patients through the K-means algorithm. A voice interaction module, connected to the main control processing module, is used to output standardized voice prompts based on grip strength and patient classification results. The feedback module, connected to the main control processing module, is used to provide tactile feedback through LED indicators and vibration motors; The main control processing module constructs a three-dimensional grip force surface model based on pressure signals, calculates the vascular exposure index, and dynamically adjusts the sensor weights according to patient classification, forming a closed-loop control from signal acquisition to personalized intervention. Mapping blood vessel diameter to an exposure index of 0-100. : ； Among them, the average value of not clenching the fist mean of optimal exposure ,when The nurse should be notified when blood can be drawn. The process of constructing the three-dimensional grip force surface model is as follows: Define the initial weight vector Where 0.3 represents the weight of the sensor in the center of the palm, and there are five... These are the weights of the five-finger sensor, two... The sensor weights are located at the proximal end of the palm near the wrist and the distal end of the palm near the base of the fingers, respectively. The formula for calculating overall grip strength is: ; Among them, based on the pressure in the middle of the palm With blood vessel diameter The quadratic relationship was used to fit the pressure-vessel diameter model: 。 2. The AI ​​blood collection decompression gripper system according to claim 1, characterized in that, The pressure sensor is an ultra-thin flexible piezoresistive sensor; the triaxial accelerometer is used to detect the grip force sensor picking up action to trigger voice identity verification.

3. The AI ​​blood collection decompression gripper system according to claim 1, characterized in that, The main control processing module includes a processor with an integrated communication unit, used to interact with the nurse terminal to receive blood collection timing signals.

4. The AI ​​blood collection decompression gripper system according to claim 3, characterized in that, The voice interaction module pre-stores Mandarin, dialect, and English voice clips, and the feedback module provides tactile feedback through a vibration motor when the fist is clenched effectively.

5. The AI ​​blood collection decompression gripper system according to claim 3, characterized in that, The main control processing module adjusts the sensor weights according to the patient classification results, reducing the weight of the finger sensor for patients with strong finger strength and increasing the weight of the palm sensor for patients with weak finger strength.

6. A method for applying an AI-powered blood collection and decompression gripper, characterized in that, The method is applicable to the AI ​​blood collection decompression gripper system according to any one of claims 1-5, and the method includes: S1. Grip force and posture signals are simultaneously collected by eight pressure sensors and accelerometers distributed on the five fingers and palm to generate raw data sequences; S2. Perform filtering preprocessing on the original data sequence to extract time-domain and frequency-domain features and form a feature vector; S3. Construct a three-dimensional grip force surface model based on the pressure signal in the feature vector, and calculate the vascular exposure index corresponding to the pressure in the middle of the palm. S4. Input the feature vector into the LSTM neural network to identify the grip strength state, and classify the patients based on the grip strength statistics in the feature vector using the K-means algorithm; S5. Based on the grip strength state, vascular exposure index and patient classification results, dynamically adjust the sensor weights and output personalized voice guidance and tactile feedback. When the grip strength state is effective and the vascular exposure index is ≥80, trigger a blood collection prompt.

7. The application method of the AI ​​blood collection decompression gripper according to claim 6, characterized in that, In the preprocessing of the filter, the Kalman filter state equation for the acceleration signal is: ； in, It is the vector of observations at time t. For the control matrix, For process noise, For state vectors, For observation noise, the transfer matrix The observation matrix is ​​a 6×6 identity matrix. Take the first 3 rows of the identity matrix; Perform a 7-point sliding median filter on the pressure signal: ; This is the result after 7-point moving median filtering. This is a pressure signal. This indicates the median operation.

Images