Multifunctional intelligent walking aid system suitable for aging

Abstract

The invention relates to the technical field of health management, in particular to a multifunctional intelligent walking aid system suitable for aging, which comprises a fall sensing module, a life detection module, a health monitoring module, a voice interaction module and a positioning lighting module. According to the invention, by fusing attitude angle change and air bag linkage, the falling trigger judgment precision is enhanced, synchronous response of early warning and protection is realized, millimeter wave radar heart rate and breathing fluctuation are extracted, static life abnormality is dynamically identified, blood oxygen and electromyographic signals are synchronously acquired, low-amplitude abnormality is identified, muscle intervention is driven, tension is relieved, and activity is improved; voice instruction analysis and monitoring result linkage broadcasting, enhanced interaction feedback, night track recognition through combination of GPS and infrared, linkage lighting and remote reporting are achieved, a monitoring, analysis, intervention and broadcasting integrated link is integrally constructed, multi-dimensional fusion of health data and travel safety is achieved, and the intelligent sensing depth and response efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of health management technology, and in particular to a multifunctional aging-friendly intelligent walking assistance system. Background Art

[0002] The field of health management technology includes technologies related to real-time monitoring of individual physical conditions, collection and analysis of physiological parameters, health risk warning and chronic disease management. The core content of this technology field is to continuously monitor and record the user's vital signs such as heart rate, blood pressure, body temperature, activity status, etc. through sensor equipment, physiological signal acquisition devices and information processing systems, thereby providing technical support for personalized health management, rehabilitation guidance and emergency response. This field emphasizes the combination of wearable devices and information platforms, and systematically covers data collection, data transmission, cloud processing, user interaction and other links. It pays special attention to the continuous monitoring and intelligent assessment of daily health status in special groups such as the elderly and patients with chronic diseases.

[0003] Among them, the multifunctional aging-friendly intelligent walking aid system refers to an integrated wearable device system used to improve the safety, convenience and health management level of the elderly during their daily travel. The system targets the characteristics of the elderly due to physical function degeneration, memory loss and slow movement. It records the whereabouts through an integrated positioning tracking device, monitors walking status through a gait recognition component, confirms user information through an identity recognition module, and realizes emergency response presets through an emergency contact device. It also includes functional components such as heart rate monitors and body temperature collectors for obtaining basic health data, forming a composite intelligent walking aid system with wearables as the core carrier, integrating travel monitoring and health data collection.

[0004] In existing technology systems, vital sign monitoring often relies on adhesive or contact sensors. Prolonged wear can easily cause skin discomfort or user rejection, and continuous recognition and real-time identification are difficult to achieve at rest, posing risks of misjudgment and data interruption. Fall detection relies solely on a single acceleration threshold, making it difficult to distinguish between strenuous exercise and actual falls, easily leading to false triggering and impacting user experience and credibility. Analysis of gait or electromyography data is used solely for behavioral recognition, failing to further establish physiological intervention logic, resulting in delayed health responses. Voice interaction is mostly passive and command-based, lacking the ability to synchronize health status reports, limiting the proactive and complete nature of information feedback. Path tracking and nighttime travel protection primarily rely on positioning information units, failing to integrate multi-source sensing for dynamic scene identification. This results in reduced recognition accuracy in obstructed and interfering environments. These issues can easily lead to the omission of critical health information and delayed responses, especially among the elderly. This can delay emergency response and reduce the system's practicality and intelligence in high-risk scenarios. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a multifunctional aging-friendly intelligent walking aid system.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solutions: A multifunctional aging-friendly intelligent walking aid system includes:

[0007] The fall sensing module extracts the attitude angle change vector and the nitrogen cylinder activation status value based on the attitude angle change value of the three-axis acceleration sensor. By comparing the vector change amplitude with the release condition, it determines the trigger command, links the airbag release, and updates the sensor feedback signal to generate the fall protection linkage instruction stream.

[0008] Based on the fall protection linkage instruction stream, the life detection module extracts the heart rate and respiratory rate signals of the millimeter wave radar within three consecutive cycles, screens the fluctuation trend and the change amplitude between time windows, establishes feature labels through frequency interval analysis, and generates a static life anomaly label set;

[0009] The health monitoring module synchronously collects blood oxygen and leg electromyographic signals based on the static life anomaly label set, extracts saturation continuous values and electromyographic spectra, analyzes low-amplitude signal segments, and activates the vibration unit of the leg muscle relaxation device in combination with the muscle response frequency to generate a multi-source health perception action group;

[0010] The voice interaction module activates the command recognition engine of voice recognition according to the multi-source health perception action group, parses the commands and keyword content issued by the user, broadcasts the parsing results in conjunction with the health data, outputs semantically synthesized voice content, and outputs an interactive health voice stream.

[0011] As a further solution of the present invention, the fall protection linkage instruction stream includes the posture angle change threshold judgment result, the airbag release linkage mark, and the sensor status update information; the static life abnormality label set includes the heart rate fluctuation amplitude label, the respiratory rate abnormality interval label, and the physiological signal stability judgment result; the multi-source health perception action group includes the blood oxygen saturation fluctuation characteristics, the electromyography signal low amplitude response label, and the muscle vibration activation parameter; the interactive health voice stream includes the voice recognition keyword index, the health status broadcast content, and the semantically generated audio data.

[0012] As a further solution of the present invention, the fall sensing module includes:

[0013] The attitude change extraction submodule identifies the attitude angle change rate and direction of the X, Y, and Z axes based on the attitude angle change value of the three-axis acceleration sensor, selects the difference between the instantaneous angular velocity increment and the angle between the gravity vector, compares the difference with the set attitude mutation reference angle, and obtains the attitude offset interval value;

[0014] The activation status recognition sub-module calls the posture offset range value, extracts the activation feedback signal of the nitrogen cylinder, discriminates the release status of the gas cylinder in the current cycle, and obtains the gas cylinder release determination value;

[0015] The linkage command generation sub-module, according to the gas cylinder release determination value, combines the triaxial sudden change amount of the attitude angle change, the mutation duration, and the sensor vector amplitude, and uses the formula:

[0016]

[0017] Calculates the linkage determination response value, makes a difference judgment between the response value and the release command setting threshold, and superimposes the gas cylinder status determination signal and the feedback trigger information to obtain the fall prevention linkage command stream;

[0018] Among them, A1 represents the X-axis angle change value, A2 represents the Y-axis angle change value, A3 represents the Z-axis angle change value, T represents the attitude mutation maintenance time, S represents the current vector amplitude, P represents the gas cylinder release determination value, and D represents the linkage determination response value.

[0019] As a further solution of the present invention, the life detection module includes:

[0020] The radar acquisition sub-module, based on the fall prevention linkage command stream, acquires the heartbeat and respiration signal data within the millimeter-wave radar period, identifies the peak difference and valley distance in each period, and obtains the body steady-state fluctuation value;

[0021] The frequency band analysis sub-module calls the body steady-state fluctuation value, screens the jump segments of the respiration and heartbeat frequency bands, compares the difference and covariance ratio between the two frequency bands, analyzes the section difference amplitude and cooperation level, and obtains the respiration and heartbeat frequency coordination amount;

[0022] The attitude calibration sub-module, according to the respiration and heartbeat frequency coordination amount, judges the frequency band jump trend, extracts the amplitude difference sequence and boundary value difference, and uses the formula:

[0023]

[0024] Calculates the frequency band linkage amplitude, combines the coordination index to judge the attitude state, and generates a static life anomaly label set;

[0025] Among them, b i is the respiration main frequency value of the heartbeat abnormal amplitude change interval i, c i is the same-section heartbeat main frequency value of the heartbeat abnormal amplitude change interval i, h i is the synchronous offset amplitude of the heartbeat abnormal amplitude change interval i, m i is the frequency jump starting value of the heartbeat abnormal amplitude change interval i, r i is the jump ending value of the heartbeat abnormal amplitude change interval i, n is the number of heartbeat abnormal amplitude change intervals, and R represents the frequency band linkage amplitude.

[0026] As a further solution of the present invention, the health monitoring module includes:

[0027] The blood oxygen capture sub-module collects blood oxygen signals based on the static life anomaly tag set, reads the amplitude and reconstructs the time series, calls the saturation critical threshold to screen the signals, and obtains the low saturation time segment interval;

[0028] The electromyogram extraction sub-module synchronously extracts electromyogram signals according to the low saturation time segment interval, analyzes the amplitude balance degree and the spectral distribution trend, screens the fluctuation frequency range and separates the low value segment of the potential change, and obtains the low amplitude electromyogram spectrum interval;

[0029] The vibration control sub-module calls the low amplitude electromyogram spectrum interval, judges the matching degree between the electromyogram response frequency and the stimulation frequency, identifies the activation intensity of the corresponding frequency band, and uses the formula:

[0030]

[0031] Calculate the response vibration activation value to obtain the multi-source health perception action group;

[0032] Wherein, Q represents the response vibration activation value, E1, E2, and E3 respectively represent the potential fluctuation values in the low frequency, medium frequency, and high frequency sections, V is the electromyogram response value under the stimulation frequency, U is the muscle sensitive response threshold in the target frequency band, F k represents the muscle contraction frequency in the three frequency bands of frequency band k, L k represents the signal duration length of frequency band k, k represents the index variable of the frequency band, and N represents the total number of frequency bands.

[0033] As a further solution of the present invention, the voice interaction module includes:

[0034] The recognition trigger sub-module screens the activation signals according to the recognition conditions according to the multi-source health perception action group, matches the command actions, extracts the associated segments and compares the trigger modes, analyzes the matching degree between the action state and the voice response, and obtains the action voice matching intensity;

[0035] The semantic analysis sub-module starts the voice recognition component according to the action voice matching intensity, collects the user voice command and extracts the audio features, constructs the phoneme sequence and the voice energy structure, analyzes the difference structure between the rhythm change and the keyword set, and uses the formula:

[0036]

[0037] Calculate the voice command recognition quantity, call the instruction tag set in the health perception action data to perform field association judgment, identify the semantic mapping relationship corresponding to the command, and obtain the linked semantic content;

[0038] Among them, W represents the voice command recognition quantity, θ represents the leading phoneme length, ξ represents the voice amplitude peak value, ψ represents the voice frequency feature value, ω represents the keyword reference frequency benchmark, and η represents the rhythm change amplitude;

[0039] The voice output sub-module extracts command keywords and health prompt segments according to the linked semantic content, combines the semantic structure and marks the word order relationship, converts it into voice generation content, and outputs an interactive health voice stream.

[0040] As a further solution of the present invention, the system further includes a positioning lighting module:

[0041] The positioning lighting module calls the interactive health voice stream, synchronously extracts the GPS coordinates and the infrared sensor status, compares the real-time coordinates with the electronic fence boundary, identifies the night movement trajectory, links the LED lighting and communication to trigger remote information upload, and obtains the position lighting control link data block;

[0042] The position lighting control link data block includes GPS positioning coordinate information, infrared sensor trigger records, and LED lighting control signals.

[0043] As a further solution of the present invention, the positioning lighting module includes:

[0044] The coordinate extraction sub-module calls the time identifier in the interactive health voice stream, collects the GPS coordinate values and the infrared sensor status values during the period, synchronously matches them according to the time sequence, analyzes the corresponding relationship between the coordinates and the perception status, and generates a walking node displacement set;

[0045] The trajectory recognition sub-module selects the coordinate segments activated by the night perception status according to the coordinate change difference of consecutive nodes in the walking node displacement set, combines the electronic fence boundary value to judge the trajectory deviation situation, and obtains the night deviation trend coefficient;

[0046] The lighting linkage sub-module calls the night deviation trend coefficient, sets the lighting trigger reference value, identifies the associated combination of the trigger time point and the LED lamp status value, records the relationship between the LED response time interval and the coordinate point change, and obtains the walking assistance lighting synchronization ratio;

[0047] The auxiliary evaluation sub-module judges the rhythm interruption and lighting feedback delay degree during the walking assistance process according to the walking assistance lighting synchronization ratio, combines the walking rhythm and the perception status change frequency, and integrates and evaluates the integrity of the node status to obtain the position lighting control link data block.

[0048] Compared with the prior art, the advantages and positive effects of the present invention are:

[0049] In the present invention, by coupling and identifying the attitude angle change value of the triaxial acceleration sensor with the activation state of the nitrogen cylinder, a linkage logic between the movement amplitude and the airbag release is established, effectively improving the accuracy of fall trigger determination, realizing the rapid response linkage of fall warning and physical protection. Based on the continuous periodic vital sign data of the millimeter-wave radar, the fluctuation trends of the heart rate and breathing frequency are extracted, and static vital abnormal signals are screened in combination with the frequency interval characteristics, realizing the dynamic discrimination of non-invasive vital detection under non-contact conditions. The blood oxygen and electromyogram signals are synchronously collected, focusing on low-amplitude and irregular segments, and active muscle intervention is carried out in combination with the frequency response law, achieving the effects of relieving muscle tension and improving lower limb mobility. The user's voice commands are parsed and bound to the health monitoring results in real time, and an information closed-loop is established by means of semantic synthesized voice feedback, effectively enhancing the interactive understanding and command execution feedback. By synergistically identifying the GPS positioning and infrared status, and combining the time period trajectory and electronic fence judgment, the synchronous joint control of night safety lighting and remote data reporting is guided, enhancing the path recognition accuracy and information connectivity in case of emergencies. The processing chain is guided by sensor input, integrating dynamic recognition, signal analysis, intervention trigger and interactive broadcast, constructing a closed-loop link from monitoring to response, realizing the collaborative processing of multi-dimensional health data, behavior information and travel safety factors, breaking through the limitations of static collection and single-point analysis, and enhancing the depth of intelligent perception and the health management response ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is the system flow chart of the present invention;

[0051] Figure 2 is the flow chart of the fall perception module in the present invention;

[0052] Figure 3 is the flow chart of the vital detection module in the present invention;

[0053] Figure 4 is the flow chart of the health monitoring module in the present invention;

[0054] Figure 5 is the flow chart of the voice interaction module in the present invention;

[0055] Figure 6 is the flow chart of the positioning lighting module in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0057] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0058] See also Figure 1 , a multifunctional aging-friendly intelligent walking assistance system includes:

[0059] The fall sensing module extracts the attitude angle change vector and the nitrogen cylinder activation status value based on the attitude angle change value of the three-axis acceleration sensor. By comparing the vector change amplitude with the release condition, it determines the trigger command, links the airbag release, and updates the sensor feedback signal to generate the fall protection linkage instruction stream.

[0060] Based on the fall protection linkage command stream, the life detection module extracts the heart rate and respiratory rate signals of the millimeter-wave radar within three consecutive cycles, screens the fluctuation trend and the change amplitude between time windows, establishes feature labels through frequency interval analysis, and generates a static life anomaly label set;

[0061] The health monitoring module, based on a static life anomaly label set, synchronously collects blood oxygen and leg electromyographic signals, extracts saturation continuous values and electromyographic spectra, analyzes low-amplitude signal fragments, and activates the vibration unit of the leg muscle relaxation device based on the muscle response frequency to generate a multi-source health perception action set.

[0062] The voice interaction module activates the voice recognition command recognition engine based on the multi-source health perception action group, analyzes the user's commands and keyword content, links the analysis results with health data, and outputs semantically synthesized voice content and interactive health voice streams;

[0063] The positioning lighting module calls the interactive health voice stream, synchronously extracts GPS coordinates and infrared sensor status, compares real-time coordinates with electronic fence boundaries, identifies nighttime movement trajectories, links LED lighting with communications to trigger remote information upload, and obtains the position lighting joint control data block.

[0064] The fall protection linkage instruction stream includes the attitude angle change threshold determination result, the airbag release linkage identifier, and the sensor status update information. The static life anomaly label set includes the heart rate fluctuation amplitude label, the respiratory rate abnormal range label, and the physiological signal stability determination result. The multi-source health perception action group includes the blood oxygen saturation fluctuation characteristics, the EMG signal low-amplitude response label, and the muscle vibration activation parameter. The interactive health voice stream includes the speech recognition keyword index, the health status broadcast content, and the semantic generation audio data. The position lighting control data block includes the GPS positioning coordinate information, the infrared sensor trigger record, and the LED lighting control signal.

[0065] Please refer to Figure 2 , the fall perception module includes:

[0066] The attitude change extraction sub-module, based on the attitude angle change value of the three-axis acceleration sensor, identifies the attitude angle change rate and direction of the X, Y, and Z axes, selects the difference between the instantaneous increment of the angular velocity and the gravity vector, compares the difference with the set attitude mutation reference angle, and obtains the attitude offset interval value.

[0067] First, calculate the attitude angle change rate and direction of the X, Y, and Z axes. The process involves the preliminary processing of the acceleration sensor data. Noise is removed through digital filtering technology to accurately extract the instantaneous acceleration change. For example, in a smart belt, when the wearer accidentally falls, the belt can real-time monitor the abnormal acceleration change. Select the difference between the instantaneous increment of the angular velocity and the gravity vector, and compare it with the real-time data through a set threshold. This threshold is dynamically calculated and adjusted based on the user's activity data in the past period. Compare this difference with the set attitude mutation reference angle. If the current value exceeds the preset safety range, it is considered that an attitude mutation has occurred. Through further data processing, determine the amplitude and direction of the mutation to obtain the attitude offset interval value.

[0068] The activation status recognition sub-module calls the attitude offset interval value, extracts the activation feedback signal of the nitrogen cylinder, and discriminates the release status of the cylinder in the current cycle to obtain the cylinder release determination value.

[0069] Perform a secondary analysis on the transmitted data to determine whether it meets the nitrogen cylinder activation conditions. This includes comparing the attitude offset with the preset nitrogen cylinder activation threshold. For example, when integrated into a belt to protect workers from falling from heights, the sensors inside the belt will real-time monitor the waist attitude. When a waist attitude mutation is detected, through comparison with the threshold, immediately determine whether it is necessary to activate the nitrogen cylinder to release the airbag. Perform screening calculations through the mutation threshold difference function. Here, the difference function is constructed based on historical accident data experience to accurately judge whether the current attitude change is sufficient to trigger a safety response, and discriminate the release status of the cylinder in the current cycle to obtain the cylinder release determination value.

[0070] The linkage instruction generation sub-module, based on the gas cylinder release determination value, combines the triaxial sudden changes in attitude angle, the duration of the sudden change, and the sensor vector amplitude, and uses the formula:

[0071]

[0072] Calculate the linkage determination response value, make a difference judgment between the response value and the set threshold of the release instruction, and superimpose the gas cylinder status determination signal and the feedback trigger information to obtain the fall protection linkage instruction stream;

[0073] Among them, A1 represents the X-axis angle change value, A2 represents the Y-axis angle change value, AЗ represents the Z-axis angle change value, T represents the duration of the attitude mutation, S represents the current vector amplitude, P represents the gas cylinder release determination value, and D represents the linkage determination response value;

[0074] Combine the triaxial sudden changes in attitude angle, the duration of the sudden change, and the sensor vector amplitude, perform joint calculations and construct a judgment function. Among them, the triaxial sudden changes A1, A2, and AЗ are the instantaneous change values of the attitude angle, and the unit is uniformly degrees. They are obtained by sampling and calculating the angle difference between the two adjacent time sampling points. For example, with a sampling period of 0.1 seconds, if the X-axis angle changes from 10° to 34° within 0.1 seconds, then A1 = |34 - 10| = 24. Similarly, if the Y-axis angle changes from 5° to 27°, then A2 = 22, and if the Z-axis angle changes from 15° to 30°, then AЗ = 15. The duration of the sudden change T represents the duration of the above-mentioned triaxial angle mutation, and the unit is uniformly seconds (s). It is obtained by multiplying the number of sampling points in the continuous mutation interval by the sampling period. If the above sudden change lasts for 3 sampling periods (i.e., 0.3 seconds), then T = 0.3;

[0075] The total amplitude S of the attitude vector represents the modulus of the triaxial acceleration vector at the current moment, and the unit is uniformly meters per second squared (m / s 2 ) and is obtained by calculation. If the sampling values are a x = 3.0m / s 2 , a y = 2.0m / s 2 , a z = 8.0m / s 2 , then

[0076] The gas cylinder release determination value P is the value generated in the previous step. According to the determination logic, its standard release critical value is set to 6.0 (the unit is also m / s 2 ), indicating that only when the acceleration modulus exceeds this value is it judged as releasable;

[0077] After substituting the above parameters, we get:

[0078] By integrating the dynamic difference analysis between the triaxial attitude change value, the duration, and the gas cylinder state threshold, a single numerical index D is formed to uniformly determine whether it is necessary to trigger the airbag release. This index can adapt to different human body attitude change modes and has sensitive response characteristics. The result shows that the linkage determination response value for the current cycle is 12.04. If the preset response threshold is 10, the protection action is triggered according to the judgment logic, and a fall protection linkage instruction stream is generated.

[0079] Please refer to Figure 3 , the life detection module includes:

[0080] Based on the fall protection linkage instruction stream, the radar acquisition sub-module acquires the heartbeat and respiration signal data within the millimeter-wave radar cycle, identifies the peak difference and valley spacing per cycle, and obtains the body steady-state fluctuation value.

[0081] Based on the indication of the fall protection linkage instruction stream, the millimeter-wave radar system is started, and the time span of three cycles is continuously monitored. The heartbeat and respiration signal data in the daily life of the elderly are collected in real time. This data is captured through the high-precision sensing function of the millimeter-wave radar and is used to analyze the correlation between the heartbeat frequency and the respiration frequency and their change rules. By calculating the peak difference and valley time interval of the heartbeat and respiration signals in each cycle, the data is processed into the total amplitude of fluctuations, that is, the difference between the maximum amplitude and the minimum amplitude of the signals in each cycle is statistically calculated. This process involves continuous sampling and peak detection of the amplitudes of the heartbeat and respiration signals. Among them, the peaks and valleys are calculated through the collected original data. For example, in a 10-second acquisition cycle, the highest point of the heartbeat signal is 100 beats per minute, and the lowest point is 60 beats per minute, and the difference is 40 beats per minute. For the respiration frequency, if the highest and lowest frequencies are 30 breaths per minute and 2 breaths per minute respectively, the difference is 10 breaths per minute. The differences are accumulated to obtain the total amplitude of fluctuations. The three-cycle fluctuation index obtained in this way can be directly used for further analysis and anomaly detection, providing instant health monitoring for the elderly and warning of potential health problems, and obtaining the body steady-state fluctuation value.

[0082] The frequency band analysis sub-module calls the body steady-state fluctuation value, screens the jump segments of the respiration and heartbeat frequency bands, compares the difference and covariance ratio between the two frequency bands, analyzes the difference amplitude and cooperation level of the segments, and obtains the respiration-heartbeat frequency cooperation quantity.

[0083] Analyze the frequency band jumps of the heartbeat and respiration signals in the time series. This analysis locates the jump events by identifying the rapidly changing parts of the signal frequencies, evaluates the synchronization between the heartbeat and respiration, compares the amplitude of the frequency jumps of the heartbeat and respiration signals, calculates the frequency difference and covariance ratio between the two in different time periods. For example, if the frequency of the heartbeat signal suddenly increases from 70 beats per minute to 100 beats per minute within a certain period, while the respiration frequency increases from 25 breaths per minute to 35 breaths per minute, this change indicates an increase or decrease in the coordination between the heartbeat and respiration. Quantify the difference in frequency bands, calculate its amplitude and cooperation level, and statistically analyze the synchronization rate between the heartbeat and respiration. For example, the synchronization rate can be obtained by dividing the number of rising and falling crossover points of the two signals by the total time. Based on this, obtain the frequency coordination index, which provides a basis for further judging the physiological state of the elderly, and obtain the respiration-heartbeat frequency coordination quantity.

[0084] The attitude calibrator sub-module judges the frequency band jump trend according to the respiration-heartbeat frequency coordination quantity, extracts the amplitude difference sequence and the boundary value difference, and uses the formula:

[0085]

[0086] Calculate the amplitude of the frequency band linkage, combine the coordination index to judge the attitude state, and generate a static life anomaly label set;

[0087] Among them, b i is the main respiration frequency value in the abnormal heartbeat amplitude change interval i, c i is the main heartbeat frequency value in the same section of the abnormal heartbeat amplitude change interval i, h i is the synchronous offset amplitude in the abnormal heartbeat amplitude change interval i, m i is the starting value of the frequency jump in the abnormal heartbeat amplitude change interval i, r i is the ending value of the jump in the abnormal heartbeat amplitude change interval i, n is the number of abnormal heartbeat amplitude change intervals, and R represents the amplitude of the frequency band linkage;

[0088] The "abnormal heartbeat amplitude change interval" refers to the time period during which the heartbeat frequency or respiration frequency undergoes continuous and sudden changes within a short time;

[0089] Analyze the synchronization trend of heartbeat and respiration signals according to the frequency coordination index, extract the cardiopulmonary frequency data in the jump frequency band, calculate the difference segment by segment by obtaining the jump boundary value and the dominant frequency value in each segment of the signal, construct the joint amplitude change amount, and select the segment where the heartbeat frequency change continuously exceeds the threshold from the original radar acquisition. Set the threshold to more than 15 times per minute. For example, in the first abnormal heartbeat amplitude change interval, the heartbeat rises from 72 to 95 beats per minute, and the respiration rises from 18 to 22 breaths per minute. Then calculate the main frequency difference in this segment b1 - c1 = 95 - 22 = 73, and the synchronization offset amplitude is calculated by the offset amount of the corresponding time points of the two signals, set as h1 = 4. The frequency start and end points of the abnormal heartbeat amplitude change interval are m1 = 72 and r1 = 95 respectively. By unifying the dimension, both the heartbeat and respiration frequencies are processed in the unit of "beats per minute" to ensure that the parameters can be directly involved in the calculation. Set the number of analysis segments as n = 3 and substitute the following values into the calculation:

[0090] Segment 1: b1 = 95, c1 = 22, h1 = 4, m1 = 72, r1 = 95;

[0091] Segment 2: b2 = 88, c2 = 26, h2 = 3, m2 = 78, r2 = 88;

[0092] Segment 3: b3 = 91, c3 = 24, h3 = 5, m3 = 80, r3 = 91;

[0093] Substitute into the original formula for multi-level operations:

[0094] Calculate the numerator

[0095] Segment 1:

[0096] Segment 2:

[0097] Segment 3:

[0098] The sum of the numerators is approximately: 163.23 + 124 + 164.15 = 451.38

[0099] Calculate the denominator

[0100] Segment 1: |72 - 95| = 23

[0101] Segment 2: |78 - 88| = 10

[0102] Segment 3: |80 - 91| = 11

[0103] The sum of the denominators is: 23 + 10 + 11 + 1 = 45

[0104] Finally, the calculation shows that:

[0105] The advantage of the formula is that, through the product form of the synchronous offset amplitude and the main frequency value, a non-linear weight effect is introduced to enhance the recognition sensitivity of the frequency hopping band, and normalization is performed using the start-stop frequency difference, making the high-variation trend segment have a greater influence. This result indicates that there is a continuous high-intensity frequency linkage phenomenon in the currently monitored object, which needs to be calibrated as an abnormal static posture state and recorded in the elderly static abnormal set for subsequent call by the linkage processing module.

[0106] Please refer to Figure 4 , the health monitoring module includes:

[0107] Based on the static life anomaly label set, the blood oxygen capture sub-module collects blood oxygen signals, reads the amplitude and reconstructs the time series, and calls the saturation critical threshold to screen the signals to obtain the low-saturation time segment interval;

[0108] Based on the static life anomaly label set, blood oxygen signals are collected. When the elderly are walking, a vital sign monitoring device is used to collect blood oxygen data. The device monitors the blood oxygen saturation in real time through a built-in sensor, and automatically identifies low-saturation time segments according to the saturation threshold defined in the label set. The time segments correspond to the moments of fatigue or dyspnea during the elderly's activities, so as to warn of health problems that need special attention. The device reads the amplitude of the collected data and reconstructs the time series, and then through comparative analysis, determines which data points are below the normal range. This process ensures that the data collected each time can accurately reflect the actual physiological state, so as to obtain the low-saturation time segment interval, and finally obtain the blood oxygen fluctuation data of the elderly during specific activities, providing a basis for medical professionals.

[0109] Based on the low-saturation time segment interval, the electromyogram extraction sub-module synchronously extracts electromyogram signals, analyzes the amplitude balance degree and spectral distribution trend, screens the fluctuation frequency range and separates the low-value segment of potential change to obtain the low-amplitude electromyogram spectrum interval;

[0110] Based on the low-saturation time segment interval, electromyogram signals are synchronously obtained, which is completed by an electromyogram device worn on the leg. The device synchronously records the electromyogram signals according to the preset time with reference to the low-saturation time segment. By analyzing the amplitude balance degree and spectral distribution change trend of the signals within the time window, medical experts can observe the correlation between muscle activity and low blood oxygen saturation, screen the amplitude fluctuation frequency range, and further separate the signal segments with smaller potential changes. The signal segments are common in cases of muscle fatigue or weakness, so as to obtain the low-amplitude electromyogram spectrum interval. This data helps to further study the interaction between muscles and blood oxygen in the elderly during activities such as walking.

[0111] The vibration control sub-module calls the low-amplitude electromyogram spectral range, judges the matching degree between the electromyogram response frequency and the stimulation frequency, identifies the activation intensity of the corresponding frequency band, and uses the formula:

[0112]

[0113] Calculate the response vibration activation value to obtain the multi-source health perception action group;

[0114] Among them, Q represents the response vibration activation value, E1, E2, and E3 respectively represent the potential fluctuation values in the low-frequency, medium-frequency, and high-frequency sections, V is the electromyogram response value at the stimulation frequency, U is the muscle sensitive response threshold in the target frequency band, F k represents the muscle contraction frequency in the three frequency bands of frequency band k, L k represents the signal duration length of frequency band k, k represents the index variable of the frequency band, and N represents the total number of frequency bands;

[0115] By reading the original potential data output by the leg electromyogram acquisition device, the electromyogram fluctuation values E1, E2, and E3 in the three frequency bands of low frequency (20 Hz), medium frequency (50 Hz), and high frequency (100 Hz) are calculated by frequency division, which respectively represent the average potential amplitudes of each frequency band, and the corresponding unit is millivolt (mV). The device performs a discrete integration operation on the signals in each frequency band to calculate the average effective amplitude within the time window length. For example, for an electromyogram signal with a duration of 3 seconds, the total amplitude of potential fluctuations in the low-frequency band is 25 mV, and the total number of samples is 1250 points, resulting in E1 = 25 / 1250 ≈ 0.02 mV. Similarly, E2 = 0.05 mV and E3 = 0.07 mV are calculated. At the same time, according to the target stimulation frequency (such as 70 Hz) set by the stimulation unit, the electromyogram response value V = 0.08 mV in the corresponding frequency band is measured, and the muscle sensitive response threshold U = 0.06 mV in the target frequency band for the elderly of this age group is set from previous clinical data. Then, the muscle contraction frequency per unit time in each frequency band is counted. For example:

[0116] In the low-frequency band, 5 times per second and the signal length is 2.5 seconds, then F1 = 5, L1 = 2.5;

[0117] Similarly, for the medium-frequency band: F2 = 4, L2 = 2.5;

[0118] For the high-frequency band: F3 = 3, L3 = 2.5;

[0119] Substitute into the formula:

[0120] The formula calculation process is as follows:

[0121] The numerator part:

[0123] Denominator part: (5 · 2.5) + (4 · 2.5) + (3 · 2.5) = 12.5 + 10 + 7.5 = 30;

[0124] Final result:

[0125] This result indicates that under the current sampling period and stimulation frequency conditions, the response vibration activation value is 0.00361, which will be transmitted to the muscle relaxation control as the driving input signal strength factor to adjust the driving mode of the vibrator and adapt to the muscle relaxation intensity required by the elderly's legs at present;

[0126] E1, E2, E3: Average potential fluctuation values of the electromyogram signal in the frequency bands of 20Hz, 50Hz, and 100Hz, with the unit of mV;

[0127] V: Electromyogram response value measured under the set stimulation frequency condition, with the unit of mV;

[0128] U: Muscle sensitivity response threshold in the target frequency band, set by empirical data, with the unit of mV;

[0129] F k : Muscle contraction frequency detected per unit time in each frequency band, with the unit of times / second;

[0130] L k : Signal duration corresponding to the frequency band, with the unit of seconds;

[0131] Represents the total time-frequency product of muscle activities in three frequency bands, with the dimension of times;

[0132] Through the synthesis of the three-frequency band fluctuation amplitudes, the calculation of the difference between the target and actual responses, and the introduction of the ratio of the signal frequency to the duration, the output activation value not only reflects the instantaneous electromyogram response intensity but also considers its persistence and sensitivity matching degree, enhancing the comprehensive regulation ability of the multi-frequency dynamic muscle state.

[0133] Please refer to Figure 5 , the voice interaction module includes:

[0134] The recognition trigger sub-module filters activation signals according to the multi-source health perception action group, matches command actions according to the recognition conditions, extracts associated segments and compares trigger patterns, analyzes the matching degree between the action state and the voice response, and obtains the action-voice matching intensity;

[0135] In the multi-source health perception action group, specific action recognition technology is used to screen action information related to voice commands. For example, when a user performs a preset action in a health monitoring device, such as raising a hand or nodding, the action will trigger the activation of a voice command. The process utilizes advanced action recognition algorithms, such as deep learning-based image processing technology, to ensure accurate action recognition. The action recognition evaluates the matching degree of the action with a preset model and generates an action recognition value, which characterizes the similarity between the action and a set threshold. If the action recognition value exceeds the set threshold, the action recognition is confirmed to be successful. Subsequently, the recognition result will activate the voice engine and prepare to receive voice commands, which ensures that the activation of voice commands is closely related to the user's actions, thereby improving device interactivity and user convenience and obtaining the action-voice matching intensity.

[0136] The semantic parsing sub-module starts the voice recognition component according to the action-voice matching intensity, collects the user's voice command, extracts audio features, constructs a phoneme sequence and a voice energy structure, analyzes the differential structure between the rhythm change and the keyword set, and uses the formula:

[0137]

[0138] Calculate the voice command recognition quantity, call the instruction label set in the health perception action data for field association judgment, identify the semantic mapping relationship corresponding to the command, and obtain the associated semantic content;

[0139] Among them, W represents the voice command recognition quantity, θ represents the length of the leading phoneme, ξ represents the peak value of the voice amplitude, ψ represents the voice frequency feature value, ω represents the keyword reference frequency benchmark, and η represents the rhythm change amplitude;

[0140] The voice command signal issued by the user is collected in real time through the microphone module, converted into time-domain audio waveform data, and divided into several short-time analysis windows, and the phoneme sequence, spectral features, and voice intensity values are extracted in turn;

[0141] When obtaining the phoneme sequence length value θ, the number of frames corresponding to the leading phoneme sequence detected in a 250ms analysis window is 25 frames, 10ms per frame, corresponding to θ = 0.25 seconds;

[0142] The maximum energy peak point is identified in the same audio segment, and the average value after energy normalization is ξ = 0.15 unit energy (the unit is the normalized linear unit);

[0143] The frequency feature value ψ is obtained by the main frequency extraction module of the voice signal, which represents the main frequency of the user's voice and is measured as ψ = 210Hz;

[0144] The standard main frequency of the keyword reference entry for comparison is ω = 200Hz, and the difference between the two reflects the keyword voice deviation degree;

[0145] The rhythm change amplitude η represents the root mean square fluctuation degree of the rhythm of the speech signal during the overall pronunciation process. By taking the square root of the average of the squared differences between the energies of the rhythm nodes, the result is η = 0.12 rhythm units;

[0146] Substitute into the formula:

[0147]

[0148] This value indicates that the comprehensive similarity matching degree of this speech sample with the set keyword in the word library is 10.0087. When the set recognition matching threshold is 9.5, the current value already meets the recognition standard, and the voice command can be determined as a valid input;

[0149] W: The recognition quantity of the voice command, which is a real value representing the comprehensive matching degree;

[0150] θ: The leading time length of the phoneme sequence, in seconds;

[0151] ξ: The normalized energy of the voice peak, in linear energy units;

[0152] ψ: The main frequency eigenvalue of the actual speech sample, in Hertz (Hz);

[0153] ω: The main frequency eigenvalue of the reference keyword in the dictionary, in Hertz (Hz);

[0154] η: The voice rhythm fluctuation amplitude, in rhythm units (which can be equivalently normalized to the mean difference of the rhythm);

[0155] By introducing three metrics to construct a dimensionless and unified dimension matching quantity from the perspectives of time, frequency, and rhythm respectively, it avoids the errors caused by relying solely on a single audio feature, enables the recognition to have a stronger comprehensive recognition ability for voice differences. The matching degree of the current voice command is higher than the set threshold, indicating that it has been effectively recognized, and the subsequent module can perform health information semantic linkage and feedback voice generation based on this.

[0156] The voice output sub-module extracts the command keyword and the health prompt segment according to the linked semantic content, combines the semantic structure and marks the word order relationship, converts it into the voice generation content, and outputs an interactive health voice stream;

[0157] After successfully identifying and parsing the instructions, corresponding voice feedback is generated. Based on the linkage semantic content obtained from the analysis, the voice output strategy is integrated, and the keywords in the user instructions are combined with the health monitoring data to form a complete voice feedback information. This information involves health parameters that the user cares about, such as heart rate or blood pressure values. The speech generation technology converts text information into natural language output to ensure that the speech is both clear and easy to understand. To this end, advanced speech synthesis technology is adopted, such as a speech synthesizer based on deep neural networks, which can adjust the tone, speed and pitch to adapt to different contexts and user preferences. The application of this technology ensures that the generated speech is not only accurate in information, but also sounds natural and fluent, which constitutes the final product output, namely the interactive health voice stream.

[0158] See also Figure 6 , positioning lighting module includes:

[0159] The coordinate extraction submodule calls the time stamp in the interactive health voice stream, collects the GPS coordinate values and infrared sensor status values within the time period, matches them synchronously according to the time sequence, analyzes the correspondence between the coordinates and the perception status, and generates a walking node displacement set;

[0160] Collecting GPS coordinates and infrared sensor status for aging-friendly smart walkers. For example, when an elderly person walks with a smart walker, the device uses its built-in GPS and infrared sensors to record their location and surrounding obstacle information in real time. Coordinate and sensor data are collected at set intervals, synchronized and matched chronologically, generating a detailed time-coordinate-state log that details the user's precise location and surrounding environment at a specific point in time. By performing aggregate operations on this data, such as determining the user's path from GPS data and correlating it with sensor status, it is possible to observe in detail the time and locations at which the user encountered potential obstacles. For example, if the infrared sensor detects an obstacle at night, this information is recorded and synchronized with the corresponding GPS location, forming a walking node displacement set. This walking node displacement set accurately reflects the user's walking status and environmental interaction along a specific road section.

[0161] The trajectory recognition submodule selects the coordinate segments activated by the nighttime perception state based on the coordinate change differences of consecutive nodes in the walking node displacement set, and judges the trajectory deviation in combination with the electronic fence boundary value to obtain the nighttime deviation trend coefficient.

[0162] Use data analysis to determine whether the walking trajectory of users at night deviates from a preset safe area, analyze the safety of users when using intelligent walking aids at night, filter data points corresponding to night time and the activation status of sensors, where the data points reflect situations where users encounter obstacles or deviate from the safe path, calculate the distance of each selected data point from the boundary of the electronic fence, and determine whether to trigger an alarm through a set safety distance threshold. For example, if an elderly user approaches the edge of the park at night, compare the distance between their GPS coordinates and the preset boundary. Once the distance is less than the threshold, it is considered that there is a risk of deviation. This not only involves basic distance calculations but also requires considering the accuracy of coordinates and the real-time update frequency to ensure the reliability of the analysis results. Finally, obtain the night shift trend coefficient, which helps guardians or the system to promptly identify and respond to potential deviation behaviors.

[0163] The lighting linkage sub-module calls the night shift trend coefficient, sets the lighting trigger reference value, identifies the associated combination of the trigger time point and the LED light status value, records the relationship between the LED response time interval and the coordinate point changes, and obtains the walking assistance lighting synchronization ratio;

[0164] Adjust the lighting of the age-friendly intelligent walking aid to provide necessary lighting support to ensure the safety of users in low-light environments. When it is identified that the user has a tendency to deviate from the safe area, immediately trigger the LED lighting to illuminate the path ahead of the user. The key execution actions include real-time monitoring of the deviation trend and calculation of the lighting response time. For example, if the user moves towards the periphery of the park at night, calculate the time difference from deviation detection to lighting activation to evaluate the timeliness of the lighting response. It is also necessary to consider the adjustment of lighting intensity to ensure sufficient lighting without causing interference under different ambient light conditions. In this way, visual assistance can be effectively provided to users, reducing the risk of walking at night, and thus obtaining the walking assistance lighting synchronization ratio, which reflects the timeliness of lighting response and environmental adaptability.

[0165] The auxiliary evaluation sub-module, based on the walking assistance lighting synchronization ratio, combines the walking rhythm and the change frequency of the perception state to judge the degree of rhythm interruption and lighting feedback delay during the walking assistance process, and integrates the integrity of the evaluation node status to obtain the position lighting control linkage data block;

[0166] Quantitatively evaluate the overall assistance effect of the age-friendly intelligent walking aid. By analyzing the lighting response duration and frequency, the user's walking rhythm, and the frequency of sensor activation, evaluate the coherence and rhythm maintenance during the walking assistance process. For example, if it is found that the lighting is frequently activated, indicating that the user encounters multiple obstacles in a specific area, in this case, the frequency of rhythm interruption will be evaluated, and the impact on the user's walking stability will be analyzed. Through analysis, targeted improvement suggestions can be provided, such as adjusting the route or increasing navigation prompts, to optimize the user's walking experience. An age-friendly lighting assistance metric value will be generated, which comprehensively considers the lighting effect, path safety, and user experience, providing a scientific basis for further improving the intelligent walking aid device.

[0167] The above are only the preferred embodiments of the present invention and do not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A multifunctional intelligent walking assistance system suitable for the elderly, characterized in that, The system includes: Based on the attitude angle change value of the triaxial acceleration sensor, the fall perception module extracts the attitude angle change vector and the nitrogen cylinder activation state value. By comparing the vector change amplitude with the release condition, it determines the trigger command, links to release the airbag, and updates the sensor feedback signal to generate a fall protection linkage instruction stream; Based on the fall protection linkage instruction stream, the life detection module extracts the heart rate and respiratory rate signals within three consecutive cycles of the millimeter-wave radar, screens the change amplitude between the fluctuation trend and the time window, and establishes a feature label through frequency interval analysis to generate a static life anomaly label set; Based on the static life anomaly label set, the health monitoring module synchronously collects blood oxygen and leg electromyography signals, extracts the continuous saturation value and the electromyogram spectrum, analyzes the low-amplitude signal segments, and activates the vibration unit of the leg muscle relaxation device in combination with the muscle response frequency to generate a multi-source health perception action group; According to the multi-source health perception action group, the voice interaction module activates the command recognition engine of voice recognition, analyzes the commands and keyword content issued by the user, links the analysis results with the health data for broadcast, outputs the semantic synthesis voice content, and outputs an interactive health voice stream.

2. The multifunctional age-friendly intelligent walking assistance system according to claim 1, wherein The fall protection linkage instruction stream includes the attitude angle change threshold determination result, the airbag release linkage identifier, and the sensor state update information. The static life anomaly label set includes the heart rate fluctuation amplitude label, the respiratory rate abnormal interval label, and the physiological signal stability determination result. The multi-source health perception action group includes the blood oxygen saturation fluctuation characteristics, the low-amplitude response label of the electromyography signal, and the muscle vibration activation parameter. The interactive health voice stream includes the voice recognition keyword index, the health status broadcast content, and the semantic generation audio data.

3. The multi-functional age-friendly intelligent walking assistance system according to claim 1, wherein The fall perception module includes: Based on the attitude angle change value of the triaxial acceleration sensor, the attitude change extraction sub-module identifies the attitude angle change rates and change directions of the X, Y, and Z axes, selects the difference between the instantaneous increment of the angular velocity and the included angle of the gravity vector, and compares the difference with the set attitude mutation reference angle to obtain the attitude offset interval value; The activation state recognition sub-module calls the attitude offset interval value, extracts the nitrogen cylinder activation feedback signal, and discriminates the cylinder release state in the current cycle to obtain the cylinder release determination value; According to the cylinder release determination value, the linkage instruction generation sub-module combines the triaxial sudden change amount, the mutation duration, and the sensor vector amplitude of the attitude angle change, and uses the formula: Calculate the linkage determination response value, perform a difference judgment on the response value and the release instruction set threshold, and superimpose the cylinder state determination signal and the feedback trigger information to obtain the fall protection linkage instruction stream; Among them, A1 represents the X-axis angle change value, A2 represents the Y-axis angle change value, A3 represents the Z-axis angle change value, T represents the attitude mutation maintenance time, S represents the current vector amplitude, P represents the cylinder release determination value, and D represents the linkage determination response value.

4. The multi-functional aging-friendly intelligent walking assistance system according to claim 3, characterized in that The life detection module includes: Based on the fall protection linkage instruction stream, the radar acquisition sub-module acquires the heartbeat and respiratory signal data within the cycle of the millimeter-wave radar, identifies the peak difference and valley spacing in each cycle, and obtains the body steady-state fluctuation value; The frequency band analysis sub-module calls the body's steady-state fluctuation value, screens the jumping segments of the breathing and heartbeat frequency bands, compares the difference and covariance ratio between the two frequency bands, analyzes the difference amplitude and cooperation level of the segments, and obtains the respiratory and heartbeat frequency cooperation quantity; The attitude calibration sub-module determines the frequency band jumping trend according to the respiratory and heartbeat frequency cooperation quantity, extracts the amplitude difference sequence and the boundary value difference, and uses the formula: Calculate the amplitude of the frequency band linkage, combine the cooperation index to judge the attitude state, and generate a static life anomaly label set; Among them, b i is the main respiratory frequency value of the abnormal heart rate amplitude range i, c i is the main heart rate frequency value of the same segment of the abnormal heart rate amplitude range i, h i is the synchronous offset amplitude of the abnormal heart rate amplitude range i, m i is the starting value of the frequency jump of the abnormal heart rate amplitude range i, r i is the ending value of the jump of the abnormal heart rate amplitude range i, n is the number of abnormal heart rate amplitude ranges, and R represents the amplitude of the frequency band linkage.

5. The multi-functional aging-friendly intelligent walking assistance system according to claim 4, wherein The health monitoring module includes: The blood oxygen capture sub-module collects blood oxygen signals based on the static life anomaly label set, reads the amplitude and reconstructs the time series, calls the saturation critical threshold to screen the signals, and obtains the low saturation time segment interval; The electromyogram extraction sub-module synchronously extracts electromyogram signals according to the low saturation time segment interval, analyzes the amplitude balance degree and the spectral distribution trend, screens the fluctuation frequency range and separates the low value segment of the potential change, and obtains the low amplitude electromyogram spectrum interval; The vibration control sub-module calls the low amplitude electromyogram spectrum interval, judges the matching degree between the electromyogram response frequency and the stimulation frequency, identifies the activation intensity of the corresponding frequency band, and uses the formula: Calculate the response vibration activation value to obtain a multi-source health perception action group; Among them, Q represents the response vibration activation value, E1, E2, and E3 respectively represent the potential fluctuation values in the low-frequency, medium-frequency, and high-frequency bands, V is the electromyogram response value at the stimulation frequency, U is the muscle sensitive response threshold in the target frequency band, F k represents the muscle contraction frequency in the three frequency bands of frequency band k, L k represents the signal duration of frequency band k, k represents the index variable of the frequency band, and N represents the total number of frequency bands.

6. The multi-functional aging-friendly intelligent walking assistance system according to claim 5, characterized in that, The voice interaction module includes: The recognition trigger sub-module screens the activation signals according to the recognition conditions according to the multi-source health perception action group, matches the command actions, extracts the associated segments and compares the trigger modes, analyzes the matching degree between the action state and the voice response, and obtains the action voice matching intensity; The semantic parsing sub-module starts the voice recognition component according to the action voice matching intensity, collects the user's voice command and extracts the audio features, constructs the phoneme sequence and the voice energy structure, analyzes the difference structure between the rhythm change and the keyword set, and uses the formula: Calculate the voice command recognition quantity, call the instruction label set in the health perception action data for field association judgment, identify the semantic mapping relationship corresponding to the command, and obtain the linked semantic content; Where, W represents the voice command recognition quantity, θ represents the leading phoneme length, ξ represents the voice amplitude peak value, ψ represents the voice frequency feature value, ω represents the keyword reference frequency benchmark, and η represents the rhythm change amplitude; The voice output sub-module extracts the command keywords and health prompt segments according to the linked semantic content, combines the semantic structure and marks the word order relationship, converts it into voice generation content, and outputs an interactive health voice stream.

7. The multifunctional age-friendly intelligent walking assistance system according to claim 1, wherein The system also includes a positioning and lighting module: The positioning and lighting module calls the interactive health voice stream, synchronously extracts the GPS coordinates and the infrared sensor status, compares the real-time coordinates with the electronic fence boundary, identifies the night movement trajectory, links the LED lighting and the communication to trigger the remote information upload, and obtains the position lighting control data block; The position lighting control data block includes GPS positioning coordinate information, infrared sensor trigger records, and LED lighting control signals.

8. The multifunctional aging-friendly intelligent walking assistance system according to claim 7, characterized in that, The positioning and lighting module includes: The coordinate extraction sub-module calls the time identifier in the interactive health voice stream, collects the GPS coordinate values and the infrared sensor status values within the time period, synchronously matches them according to the time sequence, analyzes the corresponding relationship between the coordinates and the perception status, and generates a walking node displacement set; The trajectory recognition sub-module screens the coordinate segments with the activated night perception state according to the coordinate change differences of consecutive nodes in the walking node displacement set, and judges the trajectory deviation situation in combination with the boundary value of the electronic fence to obtain the night deviation trend coefficient; The lighting linkage sub-module calls the night deviation trend coefficient, sets the lighting trigger reference value, identifies the associated combination of the trigger time point and the LED light state value, records the relationship between the LED response time interval and the coordinate point change, and obtains the walking assistance lighting synchronization ratio; The auxiliary evaluation sub-module judges the rhythm interruption and lighting feedback delay degree during the walking assistance process based on the walking assistance lighting synchronization ratio, combines the walking rhythm and the change frequency of the perception state, and integrates and evaluates the integrity of the node state to obtain the position lighting control linkage data block.

Images