Medical care shoes with emergency alarm function

Abstract

This invention relates to the field of medical safety protection equipment and discloses a medical shoe with an emergency alarm function, addressing the problems of existing shoe-mounted alarm systems being prone to false alarms under complex gait and missed alarms under weak emergency movements. The shoe includes a shoe body and a built-in processing module. By collecting foot pressure and movement state data, it determines whether the user is stationary. If not stationary, it calculates the ground hardness index and performs impact compensation on the pressure center trajectory. Subsequently, it extracts the action time window, calculates the trajectory fractal dimension, action consistency coupling characteristics, and macroscopic motion intensity characteristics. These characteristics are then normalized and weighted to form a discrimination index, which is compared with a dynamic decision threshold range. If the index exceeds the threshold, it is determined to be a valid alarm action, and an alarm command is output simultaneously. This invention improves alarm accuracy and concealment, making it suitable for medical personnel to quickly and discreetly seek help when facing threats.

Description

Technical Field

[0001] This invention relates to the field of medical safety protection equipment technology, and more specifically, this application relates to a medical shoe with an emergency alarm function. Background Technology

[0002] In the fields of medical and health monitoring and occupational safety, wearable concealed alarm technology is of great significance for protecting the personal safety of medical staff, especially in responding to violence or sudden threats within hospitals, enabling rapid and unnoticed calls for help. Existing technical solutions mostly use sensors integrated into the foot or insole to monitor changes in pressure or acceleration of movement, and preset fixed thresholds or standard action templates to identify actively triggered alarm actions.

[0003] However, healthcare workers, especially nurses, have frequent and varied gaits during their daily work, such as quickly moving around patient rooms, tiptoeing while preparing medications, and pacing. These unconscious, high-frequency small movements are easily captured by traditional sensors and exceed static thresholds, leading to frequent false alarms that may mask real dangers. On the other hand, in a real, controlled emergency situation, to avoid disturbing the perpetrator, victims often move their feet with smaller amplitude and slower speed. These gentle movements may not exceed static thresholds, thus preventing the system from triggering an alarm. This means that analysis methods based on ideal "standard movement" models are prone to misinterpreting real alarms as invalid movements in reality, resulting in missed dangers. Therefore, this paper proposes a medical shoe with an emergency alarm function to address these issues. Summary of the Invention

[0004] To solve the above-mentioned technical problems, a medical shoe with an emergency alarm function is provided. This technical solution solves the problems mentioned in the background technology.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] This application provides a medical shoe with an emergency alarm function, used to sense the postural safety of medical personnel, including:

[0007] The shoe body includes an upper, a sole located at the bottom of the upper, and a functional module compartment located within the sole;

[0008] The functional modules located within the functional module repository include:

[0009] The data acquisition and status determination unit is used to acquire plantar pressure distribution data and foot movement status data, and determine whether the user is stationary based on the foot movement status data.

[0010] The event detection and trajectory adjustment unit is used to extract the pressure time series of plantar pressure distribution data within a continuous gait cycle when the user is determined to be non-stationary, calculate the mean and variance of the rising slope and peak value of the series, and generate a weighted ground hardness index; at the same time, it identifies candidate foot movement events that meet preset trigger conditions based on foot movement state data.

[0011] It is also used to obtain the original coordinate sequence of the pressure center of the plantar pressure distribution data within the time window corresponding to the candidate foot action event, and to nonlinearly scale its vertical component using the ground hardness index to output the pressure center trajectory after impact compensation. The scaling factor of the nonlinear scaling is negatively correlated with the ground hardness index.

[0012] The coupling processing and feature fusion unit is used to process the pressure center trajectory to obtain the fractal dimension; it fuses plantar pressure distribution data and foot movement state data to calculate the coupling relationship features of movement consistency.

[0013] It is also used to normalize and weighted fuse the absolute values ​​of fractal dimension, coupling relationship features and macroscopic motion intensity features extracted from foot motion state data, and output identification index values.

[0014] The instruction generation unit is used to determine whether the identification indicator value continuously deviates from the preset decision threshold range for a preset time. If so, it is determined to be a valid alarm action and an alarm instruction is output.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0016] This application generates a pressure center trajectory compensated for impact force by synchronously collecting multi-dimensional data of plantar pressure and movement state and predicting the user's static state, combined with a ground hardness adaptive compensation mechanism. This solves the problem of trajectory distortion caused by the interference of impact transmission characteristics in plantar pressure distribution under different ground environments, and improves the environmental robustness of lower limb movement event detection.

[0017] This application solves the problem of insufficient discrimination of single motion or pressure features in complex action scenarios by multimodal fusion and weighted discrimination based on fractal dimension, action consistency coupling features and macroscopic motion intensity features. It achieves high-precision identification of emergency postures such as falls and slips and normal actions such as daily walking and running, effectively reducing the false alarm rate. Attached Figure Description

[0018] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Wherein:

[0019] Figure 1 This is a structural diagram of the medical shoe with emergency alarm function proposed in this invention;

[0020] Figure 2 This is a schematic diagram of the structure in operation of the functional modules in this invention;

[0021] Figure 3 A flowchart illustrating the method for operating the functional modules in this invention;

[0022] In the picture: 1. Upper; 2. Functional module compartment; 3. Sole. Detailed Implementation

[0023] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0024] In the fields of medical and health monitoring and occupational safety, existing wearable concealed alarm technologies are mostly based on foot sensors with preset fixed thresholds or standard action templates. These technologies are prone to false alarms due to high-frequency gait movements of medical staff in daily life (such as rapid movement, tiptoeing, etc.). At the same time, in emergency controlled states, foot movements are small in amplitude and slow in speed, which may lead to missed alarms because the static threshold is not exceeded. The reliability of the alarm is insufficient and may cover up the real danger.

[0025] To address the aforementioned issues, this application provides a medical shoe with an emergency alarm function. By acquiring data on plantar pressure distribution and foot movement, it extracts a continuous gait cycle pressure time series when the user is not stationary, calculates the ground hardness index, and nonlinearly scales the vertical component of the pressure center trajectory, outputting a trajectory compensated for impact force. Furthermore, it integrates the trajectory fractal dimension, motion consistency coupling characteristics, and macroscopic motion intensity characteristics to generate a discrimination index value. When this value exceeds a preset decision threshold, it is determined to be a valid alarm action, and an alarm command is output. This application, through dynamic trajectory adjustment adapted to ground hardness and multi-feature fusion, overcomes the limitations of static threshold or template methods in complex working environments. It can more accurately distinguish between daily actions and concealed alarm intentions, reducing the risk of false alarms and missed alarms, and improving the real-time and concealed nature of personal safety monitoring for medical personnel.

[0026] like Figure 1-2 As shown, this application introduces a medical shoe with an emergency alarm function for sensing the postural safety of medical personnel, including:

[0027] The shoe body includes an upper 1, a sole 3 located at the bottom of the upper, and a functional module compartment 2 within the sole 3;

[0028] The upper 1 is fixedly connected to the sole 3. The upper 1 has a connecting edge, which is combined with the periphery of the sole 3. The fixed connection includes, but is not limited to, sewing connection, adhesive connection, molding or injection molding connection.

[0029] Stitching connection refers to the upper 1 being fixed to the sole 3 by sewing. For example, the upper 1 and sole 3 are fixed to each other by sewing. Adhesive connection refers to the upper 1 and sole 3 being fixed by adhesive layer. For example, the upper 1 is glued to the upper surface of the sole 3 by adhesive process. Molding or injection molding connection (applicable to some sports shoes and work shoes) refers to the sole 3 covering and fixing the edge part of the upper 1 by injection molding process. For example, the upper 1 and sole 3 are integrally molded by vulcanization process.

[0030] The sole 3 and the functional module compartment 2 are connected by a detachable structure, which allows the functional module compartment 2 to be easily installed or removed as needed.

[0031] For example, the detachable connection structure includes complementary magnetic adsorption components and an electrical connector disposed on the sole 3 and the functional module compartment 2. The magnetic adsorption components achieve initial alignment and physical fixation of the functional module compartment 2, and the electrical connector enables electrical connection between the functional module compartment 2 and the main control circuit inside the sole 3.

[0032] Specifically, a first magnetic element and a first electrical connection terminal are embedded at a predetermined position on the upper surface of the sole 3. The first electrical connection terminal is preferably a spring pin (PogoPin) type interface and is electrically connected to the inside of the sole 3.

[0033] At the corresponding position on the bottom of the functional module compartment 2, a second magnetic component and a second electrical connection terminal are embedded. The polarity of the second magnetic component is opposite to that of the first magnetic component, so that when the two are brought close together, they can generate sufficient magnetic attraction to initially attract and align the functional module compartment 2 onto the sole 3. The second electrical connection terminal is a contact pad that matches the first electrical connection terminal (spring pin).

[0034] When the functional module compartment 2 is magnetically attracted to the correct position, the spring pin of the first electrical connection terminal and the contact pad of the second electrical connection terminal are pressed and form a stable electrical connection, thereby simultaneously completing physical fixation and circuit connection.

[0035] The detachable connection structure may also include auxiliary positioning mechanisms, such as guide grooves and guide protrusions, to ensure the precise alignment of the functional module compartment 2 during the adsorption process.

[0036] The functional module compartment 2 features a magnetically detachable design for easy maintenance, upgrades, and shoe cleaning and sterilization. The functional module compartment 2 is connected to the sole 3 via a precision PogoPin electrical connector and a powerful magnet array, ensuring reliable electrical connection and secure physical fixation. Disassembly time is less than or equal to 5 seconds. The shoe body (excluding the functional modules) supports 121℃ high-pressure steam sterilization, meeting the stringent hygiene requirements of medical environments.

[0037] The functional modules located in functional module repository 2 include:

[0038] The data acquisition and status determination unit is used to acquire plantar pressure distribution data and progressive motion status data, and to determine whether the user is stationary.

[0039] An array of flexible pressure sensors (e.g., arranged in an 8×6 matrix, covering the main weight-bearing areas of the forefoot, arch, and heel) integrated into the functional module compartment 2 of the sole collects pressure data at various points in real time at a sampling frequency of 100Hz, forming plantar pressure distribution data. Simultaneously, the inertial measurement unit (IMU) in the functional module compartment 2 collects raw data of the foot's triaxial acceleration and triaxial angular velocity at a frequency of 200Hz, forming foot motion state data.

[0040] Data synchronization and preprocessing: Align pressure data and IMU data by timestamp. For pressure data, calculate normalized pressure values ​​by forefoot, arch, and heel regions; for IMU data, calculate synthetic acceleration amplitude. With the magnitude of the combined angular velocity .

[0041] Feature extraction and threshold comparison: Calculate the rate of change of total plantar pressure within a set time window (e.g., 2 seconds). and calculate and window average and .Will Compared with the preset pressure change threshold (e.g., 0.5 N / s) Compare; With acceleration threshold (For example, 0.1g) for comparison, With angular velocity threshold (e.g., 5° / s) comparison.

[0042] State determination and output: If Less than ,and Less than and Less than If both conditions are met, the user is determined to be in a "stationary state"; otherwise, they are determined to be in a "non-stationary state". This determination result, along with the preprocessed data, is output to the subsequent unit.

[0043] Example: Suppose a nurse briefly stands at the nurses' station. During her standing period, pressure sensors detect a stable distribution of pressure on the sole of her foot. Approaching 0; the IMU detected only a slight amplitude of foot tremor or physiological shaking. Approximately 0.05g, The rate of change is approximately 3° / s. The system detects that the above data is below the set threshold for multiple consecutive time windows, and therefore continuously outputs the "stationary state" flag. When the nurse begins to walk, the rate of change of pressure and the amplitude of movement immediately and significantly exceed the threshold, and the system then switches to the "non-stationary state" flag.

[0044] Through the aforementioned data acquisition and status judgment unit, this solution solves the problem of inaccurate status judgment caused by isolated data in motion monitoring. By simultaneously acquiring plantar pressure distribution data and progressive motion status data, and comprehensively judging whether the user is stationary, it not only ensures the integrity of multi-dimensional motion information, but also improves the reliability of status recognition, thus providing a solid foundation for precise motion analysis and intervention.

[0045] Before acquiring plantar pressure distribution data, calibration processing is also included, specifically:

[0046] The system periodically acquires plantar pressure distribution data from users and partitions the plantar pressure distribution data within each gait cycle to obtain multiple pressure partitions.

[0047] Calculate the average pressure value of each pressure zone when the user walks, and form the zone average pressure;

[0048] Compare the average pressure of the current gait cycle partition with the average pressure of the standard partition in the storage, and calculate the attenuation ratio of the average pressure value of each pressure partition.

[0049] Based on the attenuation ratio and its partition identifier, gain compensation is performed on the acquired plantar pressure distribution data to generate compensated plantar pressure distribution data. The gain compensation is positively correlated with the attenuation ratio.

[0050] This calibration process aims to compensate for pressure signal attenuation caused by long-term sensor use, environmental changes, or individual differences, ensuring the accuracy of data used in subsequent analyses. The following steps and examples illustrate this process:

[0051] Continuously acquire raw plantar pressure distribution data generated by the pressure sensor array on the sole of the shoe.

[0052] Processing and Judgment: The system monitors the user's gait in real time. When a complete and stable walking gait cycle is identified, a calibration process is automatically triggered. The gait cycle can be determined by the periodic peaks of the plantar pressure sequence or by combining gait phase information from the IMU.

[0053] Partitioning: For the pressure data within the gait cycle that triggers calibration, the pressure value at each sampling time is assigned to the corresponding partition based on the predefined plantar region division (e.g., divided into 3 to 5 pressure partitions such as forefoot, arch, and heel).

[0054] Calculate the average pressure of each zone: Calculate the average pressure value of all sampling points in each pressure zone within the current gait cycle to obtain a set of average pressure values ​​for the current zone. , where n is the number of partitions.

[0055] Calculate the attenuation ratio: retrieve the pre-stored average pressure of the standard partition. This standard value can be derived from initial user calibration (e.g., measured while walking on a standard surface when new shoes are first used), preset models, or mass-produced calibration values. For each partition i, its attenuation ratio is calculated. . This reflects the relative attenuation of the sensor signal in that zone.

[0056] Application Compensation: During the subsequent data acquisition process, online compensation is performed on the raw plantar pressure distribution data acquired in real time. For each frame of pressure data, the corresponding gain coefficient is applied based on the partition identifier i to which each pressure sensing point belongs. Amplification is performed. Here, α is a preset gain adjustment factor (e.g., α=0.8), and the gain compensation and attenuation ratio are... Positive correlation.

[0057] Output the plantar pressure distribution data after partition gain compensation.

[0058] Example: Suppose that after several months of use, the sensor in the arch area (section 2) of a pair of medical shoes experiences a decrease in sensitivity due to material fatigue. During a calibration trigger, the system calculates the average pressure of the standard section: =100 (units omitted); Current partition average pressure: =82; then the attenuation ratio =(100-82) / 100=0.18;

[0059] If α=1, then the gain coefficient for this partition is... =1 + 1 × 0.18 = 1.18. After this, all raw pressure readings belonging to the subsequent zone are multiplied by 1.18 before output. For example, a pressure point with an original value of 50 will output 59 after compensation. This makes the output data closer to the expected value under the sensor's initial sensitivity, thus offsetting the attenuation effect and ensuring the reliability of subsequent analyses calculated based on pressure amplitude, pressure center, etc.

[0060] Through the above calibration scheme, this application solves the problem of decreased pressure measurement accuracy caused by the attenuation of sensor data with use. By establishing a dynamic zonal calibration mechanism, targeted gain compensation is performed based on the attenuation ratio of each pressure zone calculated in real time. This not only achieves dynamic correction of sensor performance degradation, but also maintains the relative accuracy and long-term reliability of pressure data in different areas, thereby ensuring that the plantar pressure distribution data on which subsequent status judgment and analysis are based is always accurate and effective.

[0061] The event detection and trajectory adjustment unit is used to extract the pressure time series of the plantar pressure distribution data in a continuous preset gait cycle if the user is not stationary, calculate the mean and variance of the rising slope and peak value of the series, and generate a weighted ground hardness index; at the same time, it identifies candidate foot action events that meet preset trigger conditions based on foot movement state data.

[0062] It is also used to obtain the original coordinate sequence of the pressure center of the plantar pressure distribution data within the time window corresponding to the candidate foot action event, and to nonlinearly scale its vertical component using the ground hardness index to output the pressure center trajectory after impact compensation. The scaling factor of the nonlinear scaling is negatively correlated with the ground hardness index.

[0063] Ground hardness index generation and candidate event identification:

[0064] When the user is determined to be in a "non-stationary" state, this unit extracts a sequence of total plantar pressure changes over time (pressure time series) from the calibrated plantar pressure distribution data over multiple consecutive preset gait cycles (e.g., 3 cycles). Simultaneously, it continuously acquires foot motion state data (IMU data).

[0065] Generating a ground hardness index: For the pressure time series, identify the rising edge of the pressure wave within each gait cycle and calculate its average slope; simultaneously extract the peak value of each pressure wave. Calculate the mean of these rising slopes and peak values ​​over consecutive cycles. , and variance , By pre-setting weights (such as...) , , , The ground hardness index is generated by weighted fusion. The higher the index value, the harder the ground.

[0066] Identify candidate foot motion events: Analyze foot motion data in real time. When a triaxial composite acceleration or angular velocity is detected to be greater than a preset trigger threshold (e.g., composite acceleration greater than 2.5g), mark that moment as the start point of a candidate foot motion event. When a triaxial composite acceleration or angular velocity is detected to be less than or equal to the preset trigger threshold (e.g., composite acceleration greater than 2.5g), mark that moment as the end point of a candidate foot motion event. Based on the timestamps of the start and end points, determine the time window of the candidate foot motion event.

[0067] Output ground hardness index Timestamp information for candidate foot movement events.

[0068] Example: A nurse walks on a tiled (hard) floor. The extracted pressure time series shows a rapid rise, high peak, and small fluctuations, and the calculated... The value is 85 (relative to the calibration value). At the same time, a rapid toe-touching motion causes the synthesized acceleration to reach 3.0g instantaneously, which is identified as the starting point of a candidate event. When the synthesized acceleration is less than or equal to 3.0g, it is identified as the ending point of a candidate event. Based on the timestamps of the starting point and the ending point, the time window of the candidate foot motion event is determined.

[0069] Impact force compensation and pressure center trajectory generation:

[0070] Within the time window of the candidate foot movement event (e.g., 0.5 seconds), obtain the original coordinate sequence of the center of pressure (CoP) in the plantar pressure distribution data. ,in The vertical component (usually representing the derived height related to the pressure amplitude).

[0071] Using ground hardness index For vertical components Nonlinear scaling is performed to obtain the compensated vertical component. Among them, the scaling factor , As the benchmark coefficient, This is the attenuation constant. Because... and Negative correlation, on hard surfaces ( When (large), for The magnification is reduced to compensate for the impact force transmission distortion caused by hard surfaces.

[0072] Output the pressure center trajectory coordinate sequence after impact compensation .

[0073] Example: Same as the previous example, the tile floor has H=85, assuming... =1.2, k=0.01, then Therefore, the original vertical component The output is significantly attenuated. The trajectory is smoother and closer to the shape when walking on soft ground, eliminating the high-frequency impact interference from hard ground.

[0074] Through the aforementioned event detection and trajectory adjustment unit, this application solves the problem of distortion in pressure center trajectory measurement caused by neglecting changes in ground hardness in existing motion analysis systems. For dynamic motion scenarios where users are not stationary, it not only dynamically generates a ground hardness index through pressure time series to quantify the influence of ground attributes, but also accurately identifies candidate action events based on foot movement data. At the same time, it uses the ground hardness index to perform nonlinear scaling compensation on the vertical component of the trajectory, ensuring that the pressure center trajectory remains accurate and reliable under different ground conditions, thereby providing a high-quality data foundation for subsequent motion analysis.

[0075] The coupling processing and feature fusion unit is used to process the pressure center trajectory to obtain the fractal dimension; it fuses plantar pressure distribution data and foot movement state data to calculate the coupling relationship features of movement consistency.

[0076] fractal dimension The calculation formula, obtained using the box counting method, is as follows:

[0077] ;

[0078] in, The required side length to cover the center of pressure trajectory is The number of grids.

[0079] The calculation process for coupling relationship characteristics specifically includes:

[0080] Based on foot movement data, calculate the user's body movement direction vector in the horizontal plane within the time window;

[0081] Based on plantar pressure distribution data, calculate the sequence of movement directions of the user's plantar pressure center within a time window;

[0082] Calculate the sliding window cross-correlation coefficient between the motion direction vector and the movement direction sequence as a coupling relationship feature.

[0083] Specifically, the trajectory of the pressure center after receiving the above compensation. Simultaneously, it acquires plantar pressure distribution data and foot movement status data synchronized within the candidate event time window.

[0084] Calculate fractal dimension Box counting is used to process the pressure center trajectory. The trajectory is covered with cubic meshes of different side lengths, and the required number of meshes is counted. With side length The relationship. By fitting The absolute value of the slope of a linear segment is the fractal dimension. It is used to quantify the complexity and irregularity of trajectories.

[0085] Calculate coupling relationship characteristics: Based on foot movement data, calculate the average motion direction vector of the human torso (estimated via IMU data) in the horizontal plane within a time window. Simultaneously, based on the plantar pressure center data, its position on the horizontal plane was calculated. movement direction sequence Then, calculate. and The cross-correlation sequence of sliding windows between the parameters is used, and the mean of the absolute values ​​of this sequence is taken as the coupling relationship feature of action consistency. The smaller the value, the worse the consistency between the foot movement and the overall direction of body movement, and the more likely it is to be an unconventional movement.

[0086] Output fractal dimension Features of coupling relationship .

[0087] Example: When threatened, the nurse physically tries to back away. Pointing backward), but the foot makes a subtle lateral movement ( (Primarily pointing to the lateral side). The calculated cross-correlation coefficients are low. Small values ​​(e.g., 0.3) show high consistency with normal walking. (Greater than 0.8) to form a contrast. At the same time, the trajectory of this movement is complex. The value is relatively high (e.g., 1.5).

[0088] Through the above-mentioned coupling processing and feature fusion scheme, this application solves the problem of incomplete motion pattern characterization caused by isolated features in existing gait analysis. For the processed pressure center trajectory and multi-source motion data, it calculates the fractal dimension of the trajectory by box counting to quantify the spatial complexity of the trajectory, and calculates the coupling relationship features of motion consistency by fusing plantar pressure and motion state data. In this way, a comprehensive feature set that can simultaneously reflect the geometric characteristics of motion and multimodal coordination is constructed, providing a deeper discriminative basis for accurate gait recognition and state assessment.

[0089] In the box counting method, the side length is dynamically adjusted based on foot movement data, and the recalculated fractal dimension after the side length adjustment replaces the original fractal dimension. Specifically, this includes:

[0090] Obtain foot movement state data within the time window corresponding to the candidate foot action event;

[0091] Frequency domain analysis was performed on foot movement data to obtain the frequency domain energy distribution;

[0092] Based on the proportion of energy above a preset high-frequency threshold in the frequency domain energy distribution, the preliminary type of the candidate foot movement event is determined to be either a high-frequency transient type or a low-frequency continuous type.

[0093] If the initial type is a high-frequency transient type, the side length of the grid in the box counting method is set to a first preset side length; if the initial type is a low-frequency continuous type, the side length is set to a second preset side length that is greater than the first preset side length.

[0094] The high-frequency threshold is dynamically adjusted based on the ground hardness index and macroscopic motion intensity characteristics, specifically including:

[0095] The ground hardness index is standardized to match its numerical range with the macroscopic motion intensity characteristics.

[0096] The standardized ground hardness index is weighted and fused with the macroscopic motion intensity characteristics to output a dynamic adjustment coefficient;

[0097] The dynamic adjustment coefficient is multiplied by the preset base threshold, and the product is used as the adjusted high-frequency threshold.

[0098] Specific implementation methods for dynamic adjustment of side length:

[0099] Dynamic adjustment of high-frequency threshold:

[0100] Get the current ground hardness index (Output from the event detection and trajectory adjustment unit) and macroscopic motion intensity features M (extracted from foot motion state data).

[0101] Standardization: The ground hardness index H is standardized to match its numerical range with the macroscopic motion intensity characteristic M, resulting in a standardized ground hardness index. ,in and They are respectively The historical mean and standard deviation.

[0102] Weighted fusion: The weighted fusion with M outputs a dynamic adjustment coefficient. ,in and The preset fusion weight coefficients, and .

[0103] Calculate the adjustment threshold: With the preset basic high-frequency threshold Multiply to obtain the adjusted high-frequency threshold. .

[0104] Output adjusted high-frequency threshold .

[0105] Example: Suppose at a certain moment H=85 (hard ground), M=0.1 (slight motion). After calculation, =1.2. Let... =0.6, =0.4, then =0.6×1.2+0.4×0.1=0.76. If =100, then =100×0.76=76. This threshold is lower than the baseline value, making it more suitable for detecting high-frequency transients on hard surfaces and under slight motion.

[0106] Dynamic adjustment of side length:

[0107] For each candidate foot movement event, obtain foot motion state data (IMU data, such as triaxial acceleration) within its corresponding time window (e.g., 0.5 seconds).

[0108] Frequency domain energy analysis: Perform a Fast Fourier Transform (FFT) on the acceleration data within the window to calculate its frequency domain energy distribution.

[0109] Preliminary event classification: Frequency distribution calculations show frequencies higher than [a certain value]. The proportion of energy in the total energy .like Greater than the preset ratio threshold If the percentage is 50%, the candidate event is determined to be "high-frequency transient type"; otherwise, it is determined to be "low-frequency continuous type".

[0110] Side length adjustment: If the event type is "high frequency instantaneous type", the side length for calculating the fractal dimension is set to the first preset side length (e.g., 0.03 meters); if it is "low frequency continuous type", it is set to the second preset side length (e.g., 0.08 meters), where the second preset side length is longer than the first preset side length.

[0111] The determined side length is output for subsequent coupling processing and feature fusion unit use.

[0112] Example: Continuing from the previous example, if the candidate event is measured to be "high-frequency transient type", the system will adjust the side length to 0.03 meters. If it is "low-frequency continuous type", then 0.08 meters will be used.

[0113] Through the above technical solution, this application solves the problem of insufficient adaptability caused by the fixed grid side length in the box counting method when analyzing foot movements. It dynamically adjusts the grid side length and recalculates the fractal dimension for different movement states, thus avoiding insufficient capture of details in high-frequency transient movements and enhancing the overall characterization ability of low-frequency continuous movements. Simultaneously, by dynamically adjusting the high-frequency threshold based on ground hardness and macroscopic movement intensity, the accuracy of frequency domain energy analysis is ensured, thereby improving the reliability of movement event classification and fractal feature calculation.

[0114] It is also used to normalize and weighted fuse the absolute values ​​of fractal dimension, coupling relationship features and macroscopic motion intensity features extracted from foot motion state data, and output identification index values.

[0115] Identification index value The calculation formula is:

[0116] ;

[0117] in, Denotes the fractal dimension. The absolute value representing the characteristics of the coupling relationship. Indicates the characteristics of macroscopic motion intensity. Represents the normalization function. , , Represents the weighting coefficient, and .

[0118] Extraction of macroscopic motion intensity characteristics:

[0119] Within the (dynamically adjusted) time window corresponding to the candidate foot motion event, obtain the foot motion state data within that window, mainly the original sequence of triaxial acceleration and triaxial angular velocity.

[0120] Calculate the composite magnitude: Calculate the composite acceleration amplitude at each sampling time. With the magnitude of the combined angular velocity .

[0121] Feature extraction: Calculate the features within this time window. and mean , with standard deviation , Macroscopic motion intensity characteristics Calculated using the following formula: ,in, , , , These are preset weighting coefficients used to comprehensively reflect the overall intensity and volatility of the action.

[0122] Output macroscopic motion intensity characteristics .

[0123] Example: The subtle swiping motion performed by nurse Xiao Zhang in an emergency, within its time window. =0.3g, =0.05g, =15° / s, =5° / s. Let... = =0.4, = =0.1, then M=0.4×0.3+0.1×0.05+0.4×15+0.1×5=6.145 (dimensional composite value, used for relative comparison).

[0124] Feature normalization and weighted fusion:

[0125] Receive the fractal dimension from the aforementioned unit The absolute value of the coupling relationship characteristics And the macroscopic motion intensity characteristics extracted in this unit. .

[0126] Normalization: A unified normalization function is used. Each feature is normalized to map it to the [0,1] interval. For example, a scaling method based on the maximum and minimum values ​​of historical data can be used:

[0127] ;

[0128] in, represent , Or M, and These are the maximum and minimum values ​​of the corresponding features in the historical dataset (or reasonable range boundaries preset based on experience).

[0129] Weighted fusion: The normalized feature values ​​are summed according to preset weight coefficients to calculate the identification index value. :

[0130] ;

[0131] in, , , Represents the weighting coefficient, and .

[0132] Output identification index value .

[0133] Example: Suppose the current measurement is =1.5, |C|=0.3, M=6.145. Based on historical data, normalization is performed, and the result is assumed to be: N( ) = (1.5-1.0) / (2.0-1.0) = 0.5; N(|C|) = (0.3-0.1) / (0.9-0.1) = 0.25; N(M) = (6.145-0) / (50-0) ≈ 0.1229; Let , , If the values ​​are 0.5, 0.3, and 0.2 respectively, then I = 0.5 × 0.5 + 0.3 × 0.25 + 0.2 × 0.1229 ≈ 0.375. This value will be used to compare with the decision threshold range to determine whether it is a valid alarm action.

[0134] Through the aforementioned feature fusion and calculation scheme, this application solves the problem of false alarms and missed alarms caused by existing action recognition systems relying on single features or lacking comprehensive evaluation. For concealed foot movements in emergency situations, it generates a comprehensive identification index value by normalizing and weighting the fractal dimension, the absolute value of coupling relationship features, and macroscopic motion intensity features. This comprehensively characterizes the spatial complexity and multimodal coordination of the action, quantifies the overall intensity and volatility of the action, and balances the contribution of each feature through weight coefficients to ensure that the identification index value can comprehensively and accurately reflect the emergency characteristics of the action, thereby improving the accuracy of alarm decision-making and the reliability of the system.

[0135] After outputting the identification index value, the process also includes smoothing it based on the current user's historical data, and replacing the original identification index value with the processed value. Specifically, this includes:

[0136] Obtain the current identification indicator value and the set of historical identification indicator values ​​associated with the current user, and calculate the historical mean and historical standard deviation based on the set of historical identification indicator values;

[0137] Based on the current identification index value, historical mean, and historical standard deviation, calculate the deviation degree, which characterizes the degree of difference between the current identification index value and the historical mean.

[0138] If the deviation is greater than the preset mutation threshold, the smoothing weight is determined based on the deviation, and the current identification index value and the historical mean are weighted and fused based on the smoothing weight to obtain the smoothed identification index value. The smoothing weight is negatively correlated with the deviation.

[0139] The specific implementation methods for smoothing the identification index values ​​include:

[0140] Obtain the current discrimination index value output by the coupling processing and feature fusion unit. .

[0141] Retrieve the set of historical identification indicator values ​​associated with the current user from local storage. (For example, the 100 most recent valid calculations).

[0142] Calculate historical statistics: Based on historical datasets, calculate their mean. with standard deviation .

[0143] Calculate deviation: Calculate the current value relative to historical average deviation The calculation formula is:

[0144] ;

[0145] This value represents the degree of difference between the current action characteristics and the user's historical norms.

[0146] Judgment and Smoothing: Deviance Compared with the preset mutation threshold Compare. If the deviation is... Less than or equal to the mutation threshold If the current action is considered to be within the normal fluctuation range, then the output will be directly applied. As the final identification index value = .

[0147] If deviation Greater than the mutation threshold If the current action is determined to be a "sudden change" that may be caused by emotions or sudden interference, then smoothing processing is required:

[0148] Based on deviation Calculate smoothing weights (For example ,in As the attenuation factor, and (Negative correlation).

[0149] For the current value Compared with historical average Weighted fusion is performed to obtain the smoothed final identification index value:

[0150] ;

[0151] Output the final discrimination index value after smoothing. This information is provided to the subsequent instruction generation unit for threshold comparison.

[0152] Example: Suppose that the historical average value of nurse Xiao Zhang's identification index is 1. =0.25, standard deviation is =0.08. This is the current identification index value calculated by the system during a specific emergency. =0.6. Therefore, the deviation is... =|0.6-0.25| / 0.08=4.375. Preset mutation threshold. =3.0, attenuation factor =0.5. Because Greater than ), calculate smoothing weights (Approximately equal to 0.50). The final discriminant value after smoothing is... It is approximately equal to 0.426. This value retains the signal of the alarm action (0.6) while being suppressed by the historical baseline (0.25), reducing the possibility of false alarms caused by a single intense emotional fluctuation.

[0153] Through the above smoothing processing scheme, this application solves the problem of misjudgment caused by the neglect of individual user habits and sudden abnormal fluctuations in existing identification indicators. For the current user's historical data and the current identification indicator value, it not only identifies significant abnormal changes by calculating the deviation, but also uses dynamic smoothing weights based on the deviation to weight and fuse the current value with the historical average. At the same time, it avoids unnecessary smoothing interference by setting a mutation threshold, ensuring that the output identification indicator value can not only reflect the user's current real-time status, but also maintain long-term stability and individual adaptability, thereby further improving the accuracy of emergency alarms.

[0154] The instruction generation unit is used to determine whether the identification indicator value continuously deviates from the preset decision threshold range for a preset time. If so, it is determined to be a valid alarm action and an alarm instruction is output.

[0155] Specifically, it receives the discrimination index value I from the output of the coupling processing and feature fusion unit, as well as a preset decision threshold range.

[0156] Threshold comparison: The identification index value I is compared with the decision threshold range. If the identification index value I deviates from the decision threshold range and the duration reaches the preset duration (e.g., 1 second), the current foot movement is determined to be a "valid alarm movement"; otherwise, it is determined to be a "non-alarm movement".

[0157] If the alarm action is deemed "valid," an alarm command is immediately generated and output. This command is a predefined digital or electrical signal used to trigger subsequent wireless communication modules (such as Bluetooth or NB-IoT) to send an alarm signal to the security center or designated terminal.

[0158] Example: Continuing from the previous example, the calculated identification index value I for nurse Xiao Zhang's subtle swiping motion is approximately 0.375. Since I (0.375) is greater than the preset decision threshold range (0.2877-0.3125) and the duration is greater than 1 second, the system determines this action to be a valid alarm action. The instruction generation unit then generates an alarm instruction and sends it to the monitoring center via the wireless module.

[0159] Through the above-mentioned instruction generation scheme, this application solves the problem of misjudgment response caused by the single identification mechanism or rigid decision threshold range of existing alarm systems. By comparing the identification index value after multi-feature fusion and smoothing with the preset decision threshold range, it ensures that effective actions in emergency situations can be identified and alarms triggered in a timely and accurate manner, while effectively filtering out false triggers caused by normal movements or non-emergency fluctuations. At the same time, by reasonably setting the decision threshold range, the sensitivity and specificity of the alarm are balanced, thereby realizing accurate and reliable emergency alarm instruction generation and output.

[0160] The decision threshold range is dynamically adjusted based on data from the user's normal working state, and the adjusted decision threshold range replaces the original decision threshold range. Specifically, this includes:

[0161] Obtain a continuous sequence of identification indicator values ​​when the user is in a normal working state within a preset learning period;

[0162] Statistical distribution modeling is performed on the identification index value sequence, and its mean and standard deviation are calculated to construct a decision threshold range that characterizes the fluctuation range of the user's normal behavior.

[0163] Within a preset learning period (e.g., the initial 24 hours), the system assumes the user is in normal working condition, continuously collecting and recording the sequence of discriminant index values ​​output by the coupling processing and feature fusion unit. During this period, the system can be in a "learning mode" that does not trigger alarms.

[0164] After the learning cycle ends, the sequence Perform statistical analysis:

[0165] Calculate the mean of the sequence with standard deviation :

[0166] ;

[0167] based on and Construct a personalized decision threshold range that characterizes the fluctuation range of the user's normal behavior. :

[0168] ;

[0169] in For the preset confidence coefficient (e.g.) =2.0).

[0170] The system can periodically (e.g., monthly) or when a significant shift in user behavior patterns is detected, re-execute the above learning process to update... , and This enables adaptive evolution of the threshold.

[0171] Example: Suppose a nurse, during a 24-hour learning cycle, has her daily work (walking, standing, small-scale movement) generated by the system into a sequence of discriminant index values, which are then calculated to obtain... =0.28, =0.05. Setting =2.0, then the individual dynamic decision-making threshold range is [0.18, 0.38]. In subsequent use, if the nurse makes a covert action due to a sudden threat, the system calculates the current identification index value as 0.45. Since 0.45 is greater than 0.38 and continues for more than the preset time, the instruction generation unit determines it as a valid alarm action and outputs an alarm instruction.

[0172] Through the above technical solution, this application establishes a user's own normal behavioral benchmark, transforming the alarm threshold from a fixed value into a personalized dynamic range. When a user performs an emergency action, the resulting identification index value is more likely to exceed the user's normal range. This effectively suppresses false alarms related to daily behavior while improving the sensitivity to identify hidden and subtle alarm actions, thus enabling precise safety monitoring tailored to individual needs.

[0173] The preset duration is dynamically adjusted based on the user's historical identification indicator value sequence, specifically including:

[0174] Obtain a continuous sequence of identification indicator values ​​when the user is in a normal working state within a preset learning period;

[0175] Statistical distribution modeling is performed on the sequence of identification index values, and its mean and standard deviation are calculated; based on the mean and standard deviation, the coefficient of variation characterizing the fluctuation range of normal user behavior is calculated;

[0176] Based on the coefficient of variation, a preset duration for dynamic adjustment is calculated through a preset mapping relationship, wherein the preset duration is positively correlated with the coefficient of variation;

[0177] Replace the original preset duration with the dynamically adjusted preset duration calculated.

[0178] The mapping relationship is as follows:

[0179] ;

[0180] in, This indicates the adjusted preset duration. Indicates the basic preset duration. Represents the coefficient of variation. This represents the preset positive adjustment coefficient.

[0181] Obtain a continuous sequence of discriminant indicator values ​​recorded within a preset learning period (e.g., the initial 24 hours) when the user is in a normal working state. .

[0182] For the sequence Perform statistical calculations to obtain its mean and standard deviation.

[0183] The coefficient of variation, which characterizes the relative fluctuation of normal user behavior, is calculated based on the ratio of the mean to the standard deviation.

[0184] Based on the coefficient of variation, through a pre-defined linear mapping relationship Calculate the dynamically adjusted preset duration. Among them, Based on preset duration, It is a positive adjustment coefficient to ensure and Positive correlation.

[0185] The calculated dynamic preset duration is updated and stored, replacing the original fixed preset duration, for use by the instruction generation unit in subsequent alarm judgment.

[0186] Through the above technical solution, when users' daily behavior fluctuates greatly, the system automatically extends the judgment duration to filter out occasional violent actions and reduce false alarms; when users' behavior is stable, a shorter judgment duration is used to ensure a rapid response in emergency situations, thereby achieving an adaptive balance between false alarm rate and response speed in individualized usage scenarios.

[0187] The functional module also includes a positioning processing unit, specifically including:

[0188] Acquire the trajectory of the center of pressure and gait temporal information from foot movement data;

[0189] The pressure center displacement vector for each step is calculated based on the pressure center trajectory, and the step frequency and gait cycle in the gait time sequence information are combined to calculate the estimated step length and movement direction for each step through the step length estimation model.

[0190] Based on the estimated step size and direction of movement, and combined with the user's position coordinates at the previous moment, the user's position coordinates at the current moment are calculated using the dead reckoning algorithm.

[0191] When the instruction generation unit outputs an alarm instruction, it simultaneously outputs the user's location coordinates as alarm location information.

[0192] The step length estimation model is as follows: based on the modulus of the pressure center displacement vector and the ground hardness index, the estimated step length is calculated through a preset mapping relationship. The estimated step length is positively correlated with the modulus of the pressure center displacement vector and negatively correlated with the ground hardness index.

[0193] The specific implementation methods of the positioning processing unit include:

[0194] Acquire the impact-compensated pressure center trajectory sequence output by the event detection and trajectory adjustment unit. .

[0195] Gait temporal information is extracted from foot movement data, including gait frequency F (steps / second) and gait cycles identified based on pressure or IMU data. (Unit: seconds).

[0196] Displacement vector and gait event detection: Within the pressure center trajectory, each complete "gait event" (such as from heel strike to second strike) is identified by detecting continuous pressure peaks or combining IMU data. For each step, the pressure center is calculated in the horizontal plane. displacement vector from the starting point to the ending point And calculate its modulus. .

[0197] Predicted step size calculation: Based on the step size estimation model, using the displacement vector magnitude With ground hardness index Calculate the estimated step size for each step. The model uses the following formula:

[0198] ;

[0199] in, and This is a preset positive coefficient. This formula reflects... and Positive correlation with Negative correlation (when walking on hard ground, the same pressure displacement may correspond to a shorter actual step length).

[0200] Direction of movement determined: Displacement vector The direction is used as the direction of movement for this step. .

[0201] Dead reckoning: Given the user's position coordinates at the previous moment. Then the current position coordinates Update using the following formula:

[0202] ;

[0203] Continuously update and output the current estimated user location coordinates .

[0204] When the instruction generation unit outputs an alarm instruction, the output of that moment is triggered synchronously. This information is used as alarm location information and is sent along with the alarm command.

[0205] Example: A nurse takes one step in a corridor (H=80). The magnitude of the horizontal displacement vector at the center of pressure during this step is measured. =0.15 meters (relative value in the shoe sole coordinate system). Let... =2.5, =0.008, then the estimated step size is... = equals 0.198 meters. If the previous position was (10.0, 20.0) and the movement direction was along the positive X-axis (1, 0), then the new position is updated to (10.198, 20.0). When this action is determined to be an alarm action, the coordinates (10.198, 20.0) are reported as the alarm position.

[0206] Through the aforementioned positioning processing unit, this application solves the problem of unreliable location information in existing positioning systems during emergencies due to reliance on external signals or lack of ground adaptability. For users in walking mode, it dynamically calculates step length and direction by combining pressure center trajectory and gait timing, optimizes the step length estimation model by combining ground hardness index, and continuously updates position coordinates through dead reckoning. When an alarm is triggered, it synchronously outputs high-precision location information to ensure timely and accurate user location in emergencies, effectively supporting rapid rescue response.

[0207] The positioning processing unit can also achieve high-precision indoor positioning via iBeacon, using a chip compliant with the Bluetooth Low Energy (BLE) 5.2 protocol, which features low power consumption and high communication stability. If several iBeacon base stations with known coordinates have been deployed in indoor environments such as hospitals and clinics, the iBeacon module on the medical shoes can periodically scan the signals broadcast by surrounding base stations and obtain the Received Signal Strength Indication (RSSI) value and base station identifier for each signal.

[0208] This unit can combine RSSI data from at least three base stations to calculate the current location using triangulation or a pre-established fingerprint database matching algorithm. To further suppress errors caused by signal fluctuations, algorithms such as Kalman filtering can be used to smooth and optimize the positioning results in real time, thereby achieving a real-time positioning accuracy of less than or equal to 0.8 meters in typical complex indoor environments.

[0209] Based on the real-time acquired base station information table, the base station identifier is converted into the corresponding base station coordinates to form a dataset {base station coordinates, RSSI value}.

[0210] Triangulation: Each RSSI value is converted into an estimated distance. Initial positioning coordinates are calculated using a trilateration algorithm, with the coordinates of at least three base stations as the center and the estimated distance as the radius.

[0211] The Kalman filter algorithm is applied to a continuous initial positioning coordinate sequence, and the motion model is used for prediction and correction to output smooth optimized positioning coordinates, achieving a real-time positioning accuracy of ≤0.8 meters.

[0212] The obtained location information can be uploaded to the hospital management system or cloud platform in real time through the wireless communication unit to form the personnel movement trajectory; it can also automatically attach the current high-precision location coordinates when the emergency alarm function is triggered to assist in rapid response and dispatch, and improve the traceability and safety monitoring level of medical staff.

[0213] Existing medical shoes often rely on limited built-in batteries, requiring frequent charging and affecting the continuity and convenience of use. To solve the problem of frequent charging, the smart medical shoe of this application also includes a self-powered unit, which includes a piezoelectric thin film array, a rectifier bridge, a boost circuit, and a capacitor. The piezoelectric thin film array is arranged in three easily deformable areas of the sole (such as the arch and forefoot) and is electrically connected to the rectifier bridge to convert mechanical energy (changes in foot pressure) during walking into electrical energy. The rectifier bridge is electrically connected to the boost circuit, and the boost circuit is electrically connected to the capacitor. The electrical energy is stored in the capacitor after passing through the rectifier bridge and the boost circuit, powering the entire functional module.

[0214] For example, the average output power per step for a user is greater than or equal to 0.6mW, and the system can support more than 90 days of continuous operation during 8 hours of daily walking.

[0215] To facilitate understanding of the above embodiments, a specific application scenario of the above embodiments will be used as an example for illustration below:

[0216] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A medical shoe with an emergency alarm function, characterized in that, Used to sense the postural safety of medical staff, including: The shoe body includes an upper, a sole located at the bottom of the upper, and a functional module compartment located within the sole; The functional modules located within the functional module repository include: The data acquisition and status determination unit is used to acquire plantar pressure distribution data and foot movement status data, and determine whether the user is stationary based on this data. The event detection and trajectory adjustment unit is used to extract the pressure time series of plantar pressure distribution data within a continuous gait cycle when the user is determined to be non-stationary, calculate the mean and variance of the rising slope and peak value of the series, and generate a weighted ground hardness index; at the same time, it identifies candidate foot movement events that meet preset trigger conditions based on foot movement state data. It is also used to obtain the original coordinate sequence of the pressure center of the plantar pressure distribution data within the time window corresponding to the candidate foot action event, and to nonlinearly scale its vertical component using the ground hardness index to output the pressure center trajectory after impact compensation. The scaling factor of the nonlinear scaling is negatively correlated with the ground hardness index. The coupling processing and feature fusion unit is used to process the pressure center trajectory to obtain the fractal dimension; it fuses plantar pressure distribution data and foot movement state data to calculate the coupling relationship features of movement consistency. It is also used to normalize and weighted fuse the absolute values ​​of fractal dimension, coupling relationship features and macroscopic motion intensity features extracted from foot motion state data, and output identification index values. The instruction generation unit is used to determine whether the identification indicator value continuously deviates from the preset decision threshold range for a preset time. If so, it is determined to be a valid alarm action and an alarm instruction is output.

2. The medical shoes with emergency alarm function according to claim 1, characterized in that, Before acquiring plantar pressure distribution data, calibration processing is also included, specifically: The system periodically acquires plantar pressure distribution data from users and partitions the plantar pressure distribution data within each period to obtain multiple pressure partitions. Calculate the average pressure value of each pressure zone when the user walks, and form the zone average pressure; Compare the average pressure of the current period's partitions with the average pressure of the standard partitions in the storage, and calculate the attenuation rate of the average pressure value of each pressure partition. Based on the attenuation ratio and its partition identifier, gain compensation is performed on the acquired plantar pressure distribution data to generate compensated plantar pressure distribution data. The gain compensation is positively correlated with the attenuation ratio.

3. The medical shoes with emergency alarm function according to claim 1, characterized in that, The fractal dimension The calculation formula, obtained using the box counting method, is as follows: ； in, The required side length to cover the trajectory of the pressure center is... The number of grids.

4. The medical shoes with emergency alarm function according to claim 3, characterized in that, The side length of the box counting method is dynamically adjusted based on foot movement data, and the recalculated fractal dimension after the side length adjustment replaces the original fractal dimension. Specifically, this includes: Obtain foot movement state data within the time window corresponding to the candidate foot action event; Frequency domain analysis was performed on foot movement data to obtain the frequency domain energy distribution; Based on the proportion of energy above a preset high-frequency threshold in the frequency domain energy distribution, the preliminary type of the candidate foot movement event is determined to be either a high-frequency transient type or a low-frequency continuous type. If the initial type is a high-frequency transient type, the side length of the grid in the box counting method is set to a first preset side length; if the initial type is a low-frequency continuous type, the side length is set to a second preset side length that is greater than the first preset side length. Recalculate the fractal dimension after adjusting the side lengths and replace the original fractal dimension.

5. The medical shoes with emergency alarm function according to claim 4, characterized in that, The high-frequency threshold is dynamically adjusted based on the ground hardness index and macroscopic motion intensity characteristics, specifically including: The ground hardness index is standardized to match its numerical range with the macroscopic motion intensity characteristics. The standardized ground hardness index is weighted and fused with the macroscopic motion intensity characteristics to output a dynamic adjustment coefficient; The dynamic adjustment coefficient is multiplied by the preset base threshold, and the product is used as the adjusted high-frequency threshold.

6. The medical shoes with emergency alarm function according to claim 1, characterized in that, The calculation process of the coupling relationship characteristics specifically includes: Based on the foot movement data, calculate the user's body movement direction vector in the horizontal plane within the time window; Based on the plantar pressure distribution data, calculate the sequence of movement directions of the user's plantar pressure center within the time window; Calculate the sliding window cross-correlation coefficient between the motion direction vector and the movement direction sequence, and use it as the coupling relationship feature.

7. The medical shoes with emergency alarm function according to claim 1, characterized in that, The formula for calculating the identification index value is as follows: ； in, Indicates the identification index value, Denotes the fractal dimension. The absolute value representing the characteristics of the coupling relationship. Indicates the characteristics of macroscopic motion intensity. Represents the normalization function. , , Represents the weighting coefficient, and .

8. The medical shoes with emergency alarm function according to claim 7, characterized in that, After outputting the identification index value, the process also includes smoothing it based on the current user's historical data, and replacing the original identification index value with the processed value. Specifically, this includes: Obtain the current identification indicator value and the set of historical identification indicator values ​​associated with the current user, and calculate the historical mean and historical standard deviation based on the set of historical identification indicator values; Based on the current identification index value, the historical mean, and the historical standard deviation, calculate the deviation degree, which characterizes the degree of difference between the current identification index value and the historical mean. If the deviation is greater than a preset mutation threshold, a smoothing weight is determined based on the deviation, and the current identification index value and the historical mean are weighted and fused based on the smoothing weight to obtain a smoothed identification index value, wherein the smoothing weight is negatively correlated with the deviation. Replace the original identification index value with the smoothed identification index value.

9. The medical shoes with emergency alarm function according to claim 1, characterized in that, The decision threshold range is dynamically adjusted based on data from the user's normal working state, specifically including: Obtain a continuous sequence of identification indicator values ​​when the user is in a normal working state within a preset learning period; Statistical distribution modeling is performed on the identification index value sequence, and its mean and standard deviation are calculated to construct a decision threshold range that characterizes the fluctuation range of the user's normal behavior.

10. The medical shoes with emergency alarm function according to claim 1, characterized in that, The functional module also includes a positioning processing unit, specifically comprising: Obtain the trajectory of the pressure center and the gait timing information from the foot movement data; The pressure center displacement vector for each step is calculated based on the pressure center trajectory, and the step frequency and gait cycle in the gait time sequence information are combined to calculate the estimated step length and movement direction for each step through the step length estimation model. Based on the estimated step size and direction of movement, and combined with the user's position coordinates at the previous moment, the user's position coordinates at the current moment are calculated using the dead reckoning algorithm. When the instruction generation unit outputs an alarm instruction, it simultaneously outputs the user's location coordinates as alarm location information.

Images