Remote monitoring nursing method, device and equipment for old people and storage medium

Abstract

The invention relates to the technical field of medical monitoring, and particularly discloses a remote monitoring nursing method, device and equipment for old people and a storage medium, and the method comprises the steps: obtaining wrist acceleration information and environment sound information; acquiring a wrist acceleration frequency spectrum and an environment sound frequency spectrum; calculating the ratio of the noise intensity of the environment sound information at each frequency to the environment noise reference intensity to obtain a noise intensity proportion; acquiring a weighted wrist acceleration frequency spectrum; splicing to form a fusion feature vector; and adopting a preset health risk identification model to carry out risk level evaluation so as to judge the current health state of the old people, and generating alarm information when the risk level exceeds a preset threshold value. According to the method, monitoring and abnormal early warning of the health state of the old people are realized, the motion information and the environment sound information are comprehensively utilized, the interference of sound noise of household appliances on the acceleration information can be effectively inhibited, and the accuracy of health risk identification is improved.

Description

Technical Field

[0001] The present application relates to the field of medical monitoring technology, and in particular to a method, device, equipment and storage medium for remote monitoring and care of the elderly. Background Art

[0002] As home-based elderly care becomes increasingly popular, the need for health monitoring for the elderly, especially those with tremor symptoms, is becoming more and more prominent. To meet this need, remote monitoring systems have emerged, aiming to monitor the physical condition of the elderly in real time and continuously, and to issue timely warnings in case of abnormalities, so as to provide safety protection for the elderly living at home. Currently, remote monitoring systems on the market mostly rely on wearable devices, such as smart watches and smart bracelets, which monitor the daily activity data of the elderly, such as the number of steps, activity duration, and sleep quality, through built-in motion sensors. More advanced systems also integrate environmental sensors, such as sound, temperature and humidity sensors, to comprehensively assess the health status of the elderly and the safety of the home environment.

[0003] However, when applied to elderly people with tremor symptoms, the existing technical solutions have obvious shortcomings. Hand tremors can interfere with the motion sensor data of the wrist wearable device, making it difficult for the system to accurately identify the elderly's true motion state, reducing the accuracy and reliability of motion data analysis. At the same time, complex sound interference in the home environment, such as the sound of household appliances such as televisions and stereos, can also affect the accuracy of the environmental sound sensor, making it impossible to accurately assess the current health status of the elderly.

[0004] There is currently no effective technical solution to the above problems. Summary of the invention

[0005] The purpose of this application is to provide a remote monitoring and care method, device, equipment and storage medium for the elderly, so as to improve the accuracy of the assessment of the current health status of the elderly in remote monitoring.

[0006] In a first aspect, the present application provides a remote monitoring method for the elderly, which is used to remotely monitor the elderly who carry a wrist wearable device. The steps of the method include: S1, obtaining wrist acceleration information based on an acceleration sensor in a wrist wearable device; S2, obtaining ambient sound information based on an ambient sound sensor in a wrist wearable device; S3, performing fast Fourier transform on the wrist acceleration information and the ambient sound information respectively to obtain a wrist acceleration spectrum and an ambient sound spectrum; S4, analyzing the ambient sound spectrum to identify the frequency range characteristics of household appliance sounds, calculating the ratio of the noise intensity of the ambient sound information at each frequency to the ambient noise baseline intensity, and obtaining a noise intensity ratio; S5. Calculate the weight of the wrist acceleration spectrum at the corresponding frequency using a weight calculation function according to the noise intensity ratio, and obtain a weighted wrist acceleration spectrum; S6, concatenating the weighted wrist acceleration spectrum with the ambient sound spectrum to form a fusion feature vector; S7. Based on the fused feature vector, a preset health risk identification model is used to perform risk level assessment to determine the current health status of the elderly, and an alarm message is generated when the risk level exceeds a preset threshold.

[0007] The remote monitoring method for the elderly of the present application realizes the monitoring of the health status of the elderly and abnormal warning. It synchronously collects wrist acceleration information and environmental sound information through acceleration sensors and environmental sound sensors to reflect the movement status and environmental sound conditions of the elderly. The frequency distribution characteristics of vibration noise and electrical noise can be effectively distinguished through frequency domain processing. The weighted wrist acceleration spectrum and the environmental sound spectrum are spliced ​​to form a fusion feature vector. The movement information and environmental sound information are comprehensively utilized to comprehensively reflect the health status of the elderly. It can effectively suppress the interference of household appliance sound noise on acceleration information, improve the accuracy of health risk identification, and provide more reliable technical means for remote monitoring of the elderly.

[0008] In the remote monitoring method for the elderly, there are multiple acceleration sensors and / or multiple ambient sound sensors, and step S5 includes: S51, obtaining a time synchronization error range of the wrist acceleration spectrum and the ambient sound spectrum; S52, determining a value interval of a noise intensity ratio of a corresponding ambient sound spectrum for each frequency in the wrist acceleration spectrum and the time synchronization error range; S53, calculating the mean value of the noise intensity ratio corresponding to each frequency within the value range; S54, using a weight calculation function to calculate the weight of the wrist acceleration spectrum at the corresponding frequency according to the mean value of the noise intensity ratio, and performing weighted processing on the wrist acceleration spectrum based on the weight to obtain a weighted wrist acceleration spectrum.

[0009] In this example, step S51 is used to determine the time synchronization error range of the wrist acceleration spectrum and the ambient sound spectrum. If there are multiple wearable devices or ambient sound sensors, the time synchronization error range of each combination of sensors is calculated. This process can be pre-set by the system, or an experimental calibration method can be used to pre-measure the time delay distribution of multiple acceleration sensors and multiple ambient sound sensors during data acquisition and transmission, thereby obtaining the statistical range of the time synchronization error.

[0010] The remote monitoring method for the elderly, wherein step S53 comprises: S531, for the value interval, using Gaussian filtering to smooth the noise intensity ratio; S532, calculating the arithmetic mean of the noise intensity ratio after smoothing to obtain a preliminary noise intensity ratio mean; S533. Determine whether the preliminary noise intensity ratio mean value ratio exceeds a preset noise threshold value. If so, correct the preliminary noise intensity ratio mean value to the noise threshold value and use it as the noise intensity ratio mean value of the corresponding frequency point. If not, directly use the preliminary noise intensity ratio mean value as the noise intensity ratio mean value of the corresponding frequency point.

[0011] The remote monitoring method for the elderly, wherein step S4 comprises: S41, extracting frequency characteristics of each frequency according to the ambient sound spectrum; S42, matching the frequency characteristics with reference frequency characteristics in a preset household appliance sound spectrum database to determine whether each frequency belongs to a household appliance sound frequency, wherein the household appliance sound spectrum database stores reference frequency characteristics of multiple household appliances in different working states; S43. Calculate the ratio of the ambient sound intensity to the ambient noise baseline intensity for each frequency. If the frequency is determined to be a household appliance sound frequency, use the ratio as the noise intensity ratio of the frequency. Otherwise, multiply the ratio by a preset attenuation factor as the noise intensity ratio of the frequency.

[0012] The remote monitoring method for the elderly, wherein step S6 comprises: S61, respectively calculating the energy entropy of the weighted wrist acceleration spectrum and the ambient sound spectrum to obtain the wrist acceleration energy entropy and the ambient sound energy entropy; S62, determining the splicing weight of the wrist acceleration spectrum and the ambient sound spectrum according to the wrist acceleration energy entropy and the ambient sound energy entropy, if the wrist acceleration energy entropy is greater than the ambient sound energy entropy, increasing the splicing weight of the wrist acceleration spectrum, otherwise, increasing the splicing weight of the ambient sound spectrum; S63. Perform weighted splicing on the weighted wrist acceleration spectrum and the ambient sound spectrum according to the splicing weight to form a fusion feature vector.

[0013] In the remote monitoring method for the elderly, the health risk identification model includes a plurality of pre-trained health risk identification sub-models, and step S7 includes: S71, inputting the fused feature vector into a plurality of health risk identification sub-models respectively to obtain a plurality of health risk level assessment results; S72, determining the fusion weight of multiple health risk level assessment results according to the matching rate between the historical health data of the elderly and the health risk level assessment results of different health risk identification sub-models; S73. Use fusion weights to perform weighted fusion on multiple health risk level assessment results, obtain the risk level to determine the current health status of the elderly, and generate an alarm message when the risk level exceeds a preset threshold.

[0014] The remote monitoring method for the elderly, wherein the method further comprises the following steps executed between step S2 and step S3: SA, fuzzy processing of environmental sound information, including: A1. After collecting the ambient sound information, extract multiple sound feature parameters of the ambient sound information, where the sound feature parameters include volume, pitch, and frequency distribution; A2. For each sound feature parameter, determine the corresponding fuzzy membership function, which is used to describe the membership degree of the sound feature parameter on different fuzzy sets; A3. According to the fuzzy membership of each sound feature parameter, fuzzy reasoning rules are used to perform reasoning to obtain the fuzzification result of the environmental sound information. The fuzzy reasoning rules are established based on expert knowledge or historical data and are used to determine the fuzzy set to which the current environmental sound information belongs; A4. Use the fuzzified result of the ambient sound information as new ambient sound information.

[0015] In a second aspect, the present application also provides a remote monitoring device for the elderly, which is used to remotely monitor the elderly who carry a wrist wearable device, and the device includes: A first acquisition module, configured to acquire wrist acceleration information based on an acceleration sensor in a wrist wearable device; A second acquisition module is used to acquire environmental sound information based on an environmental sound sensor in the wrist wearable device; A spectrum conversion module, used for performing fast Fourier transform on the wrist acceleration information and the ambient sound information respectively to obtain a wrist acceleration spectrum and an ambient sound spectrum; A noise analysis module, used to analyze the ambient sound spectrum to identify the frequency range characteristics of household appliance sounds, calculate the ratio of the noise intensity of the ambient sound information at each frequency to the ambient noise baseline intensity, and obtain a noise intensity ratio; A weighting module, used to calculate the weight of the wrist acceleration spectrum at the corresponding frequency using a weight calculation function according to the noise intensity ratio, and obtain a weighted wrist acceleration spectrum; A feature concatenation module is used to concatenate the weighted wrist acceleration spectrum with the ambient sound spectrum to form a fused feature vector; The risk assessment module is used to perform risk level assessment based on the fused feature vector and a preset health risk identification model to determine the current health status of the elderly, and generate an alarm message when the risk level exceeds a preset threshold.

[0016] The remote monitoring device for the elderly of the present application realizes the monitoring of the health status of the elderly and abnormal warning. It synchronously collects wrist acceleration information and environmental sound information through acceleration sensors and environmental sound sensors to reflect the movement status and environmental sound conditions of the elderly. It can effectively distinguish the frequency distribution characteristics of vibration noise and electrical noise through frequency domain processing. It forms a fusion feature vector by splicing the weighted wrist acceleration spectrum and the environmental sound spectrum, and comprehensively utilizes the movement information and environmental sound information to comprehensively reflect the health status of the elderly. It can effectively suppress the interference of household appliance sound noise on acceleration information, improve the accuracy of health risk identification, and provide more reliable technical means for remote monitoring of the elderly.

[0017] In a third aspect, the present application further provides an electronic device, comprising a processor and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps in the method provided in the first aspect are executed.

[0018] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps in the method provided in the first aspect are executed.

[0019] As can be seen from the above, the present application provides a remote monitoring and care method, device, equipment and storage medium for the elderly, wherein the remote monitoring method for the elderly realizes the monitoring of the health status of the elderly and abnormal warning. It synchronously collects wrist acceleration information and environmental sound information through acceleration sensors and environmental sound sensors to reflect the movement status and environmental sound conditions of the elderly. The frequency distribution characteristics of vibration noise and electrical noise can be effectively distinguished through frequency domain processing. The weighted wrist acceleration spectrum and the environmental sound spectrum are spliced ​​to form a fusion feature vector. The comprehensive use of motion information and environmental sound information can comprehensively reflect the health status of the elderly. It can effectively suppress the interference of household appliance sound noise on acceleration information, improve the accuracy of health risk identification, and provide more reliable technical means for remote monitoring of the elderly. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flowchart of a remote monitoring method for the elderly provided in an embodiment of the present application.

[0021] Figure 2 A schematic diagram of the structure of a remote monitoring device for the elderly provided in an embodiment of the present application.

[0022] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0023] Figure numerals: 201, first acquisition module; 202, second acquisition module; 203, spectrum conversion module; 204, noise analysis module; 205, weighting module; 206, feature splicing module; 207, risk assessment module; 301, processor; 302, memory; 303, communication bus. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0025] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0026] First, please refer to Figure 1 Some embodiments of the present application provide a remote monitoring method for the elderly, which is used to remotely monitor the elderly who carry a wrist wearable device. The steps of the method include: S1, obtaining wrist acceleration information based on an acceleration sensor in a wrist wearable device; S2, obtaining ambient sound information based on an ambient sound sensor in a wrist wearable device; S3, performing fast Fourier transform on the wrist acceleration information and the ambient sound information respectively to obtain a wrist acceleration spectrum and an ambient sound spectrum; S4, analyzing the ambient sound spectrum to identify the frequency range characteristics of household appliance sounds, calculating the ratio of the noise intensity of the ambient sound information at each frequency to the ambient noise baseline intensity, and obtaining a noise intensity ratio; S5. Calculate the weight of the wrist acceleration spectrum at the corresponding frequency using a weight calculation function according to the noise intensity ratio, and obtain a weighted wrist acceleration spectrum; S6, concatenating the weighted wrist acceleration spectrum with the ambient sound spectrum to form a fusion feature vector; S7. Based on the fused feature vector, a preset health risk identification model is used to perform risk level assessment to determine the current health status of the elderly, and an alarm message is generated when the risk level exceeds a preset threshold.

[0027] Specifically, the wrist wearable device is configured to be worn on the wrist of the elderly, and the accelerometer is integrated into the wearable device to collect acceleration information reflecting the movement state of the elderly's wrist, so as to sense the acceleration changes caused by the movement of the elderly's wrist; the ambient sound sensor is also integrated into the wrist wearable device, and works in conjunction with the accelerometer to collect sound information of the elderly's environment, so as to capture the sound of the elderly's environment. The ambient sound information can reflect the noise conditions in the home environment, such as the noise generated by household appliances.

[0028] More specifically, the purpose of step S3 is to convert the wrist acceleration information and environmental sound information in the time domain into the frequency domain to obtain the wrist acceleration spectrum and the environmental sound spectrum, so as to facilitate the analysis and processing of noise in the frequency domain. In the frequency domain, the frequency distribution characteristics of vibration noise and electrical noise are more easily identified and analyzed, which can effectively solve the problem that the interference of hand tremor on the motion sensor data of the wrist wearable device cannot be separated, thereby affecting the accuracy of subsequent evaluation results.

[0029] More specifically, in step S4, the ambient sound spectrum is analyzed to identify the frequency range characteristics of the sounds of household appliances in the area where the elderly are located. For example, the frequency components of the sounds of household appliances such as televisions and refrigerators in the ambient sound spectrum are determined through a pre-established household appliance sound spectrum database or a sound pattern recognition algorithm. After determining the frequency range characteristics of the sounds of these household appliances, the ratio of the noise intensity of the ambient sound information at each frequency to the ambient noise baseline intensity can be calculated to obtain the noise intensity ratio. The noise intensity ratio value can quantify the relative intensity of the noise at each frequency, that is, it reflects the intensity of the ambient noise at different frequencies. The ambient noise baseline intensity can be a pre-set fixed value or a value dynamically estimated based on the ambient sound spectrum.

[0030] More specifically, step S5 calculates the weight of the wrist acceleration spectrum at the corresponding frequency using a weight calculation function according to the noise intensity ratio obtained in step S4; the purpose of setting the weight calculation function for weight calculation is to reduce the weight of the frequency component interfered by noise in the wrist acceleration spectrum, and increase the weight of the non-noise frequency component, thereby improving the anti-noise performance, and the weighted wrist acceleration spectrum is obtained by multiplying the original wrist acceleration spectrum by the calculated weight.

[0031] More specifically, step S6 splices the weighted wrist acceleration spectrum with the ambient sound spectrum to form a fused feature vector. The splicing operation aims to integrate information from two different sources to form a comprehensive feature representation containing motion and ambient sound information, thereby providing a more comprehensive input for subsequent health risk identification. The splicing method can be a simple vector connection or a more complex feature fusion algorithm.

[0032] More specifically, step S7 inputs the fused feature vector into a preset health risk identification model to quickly and conveniently complete the assessment of the current risk level of the elderly. The health risk identification model can be a pre-trained machine learning model or a deep learning model, which can assess the health risk level of the elderly based on the input feature vector; step S7 also compares the risk level output by the model with a preset threshold. When the risk level exceeds the preset threshold, an alarm message is generated and the relevant guardians are notified through remote transmission so that timely intervention measures can be taken to monitor the health status of the elderly and provide abnormal warnings.

[0033] The remote monitoring method for the elderly in the embodiment of the present application realizes the monitoring of the health status of the elderly and abnormal warning. It synchronously collects wrist acceleration information and environmental sound information through acceleration sensors and environmental sound sensors to reflect the movement status and environmental sound conditions of the elderly. The frequency distribution characteristics of vibration noise and electrical noise can be effectively distinguished through frequency domain processing. A fusion feature vector is formed by splicing the weighted wrist acceleration spectrum and the environmental sound spectrum. The movement information and environmental sound information are comprehensively utilized to comprehensively reflect the health status of the elderly. It can effectively suppress the interference of household appliance sound noise on acceleration information, improve the accuracy of health risk identification, and provide more reliable technical means for remote monitoring of the elderly.

[0034] In some preferred embodiments, there are multiple acceleration sensors and / or ambient sound sensors, and step S5 includes: S51, obtaining a time synchronization error range of a wrist acceleration spectrum and an ambient sound spectrum; S52, determining a value interval of a noise intensity ratio of a corresponding ambient sound spectrum for each frequency in the wrist acceleration spectrum and a time synchronization error range; S53, calculating the mean value of the noise intensity ratio corresponding to each frequency within the value range; S54, using a weight calculation function to calculate the weight of the wrist acceleration spectrum at the corresponding frequency according to the mean value of the noise intensity ratio, and performing weighted processing on the wrist acceleration spectrum based on the weight to obtain a weighted wrist acceleration spectrum.

[0035] Specifically, when remotely monitoring elderly people with hand tremors, the elderly may wear multiple wearable devices or have multiple environmental sound sensors. There may be time synchronization errors between different devices or sensors, resulting in deviations in the weight calculation of the noise intensity ratio and the wrist acceleration spectrum, affecting the noise suppression effect and further affecting the accuracy of health risk identification. To address this problem, the remote monitoring method for the elderly in the embodiment of the present application introduces a time synchronization error correction mechanism and calculates the mean value of the noise intensity ratio to improve the accuracy of the weight calculation, thereby improving the noise suppression effect and the accuracy of health risk identification.

[0036] More specifically, step S51 is used to determine the time synchronization error range of the wrist acceleration spectrum and the ambient sound spectrum. If there are multiple wearable devices or ambient sound sensors, the time synchronization error range of each combination of sensors is calculated (the average error range or the maximum error range can be used). This process can be pre-set by the system, or an experimental calibration method can be used to pre-measure the time delay distribution of multiple acceleration sensors and multiple ambient sound sensors during data acquisition and transmission, thereby obtaining the statistical range of the time synchronization error.

[0037] It should be noted that, in other implementations, if there is only one acceleration sensor and one ambient sound sensor, step S51 is to calculate the synchronization error range of the acceleration sensor and the ambient sound sensor.

[0038] More specifically, in step S52, the value interval corresponding to each frequency is determined in combination with each frequency of the acceleration spectrum and the time synchronization error range. The frequency should be located in the middle of the value interval, and the value interval is a frequency interval.

[0039] More specifically, step S53 extracts the noise intensity ratios corresponding to all frequencies within the value corresponding to each frequency, and then calculates the average value of these noise intensity ratios as the mean noise intensity ratio; in the calculation of the mean noise intensity ratio, methods such as arithmetic mean, weighted mean or median filtering can be used to reduce random fluctuations introduced by time synchronization errors.

[0040] More specifically, in step S54, the weight calculation function can be a linear function, a nonlinear function or a piecewise function, etc. The design goal of the weight calculation function is to reasonably adjust the weight of the wrist acceleration spectrum according to the mean value of the noise intensity ratio, reduce the weight of the noise frequency component, and increase the weight of the effective motion signal frequency component; the above processing method fully considers the influence of the time synchronization error on the noise intensity ratio, and obtains a more stable and accurate noise intensity evaluation by calculating the mean value of the noise intensity ratio within the time error range. The noise intensity ratio mean value can more accurately reflect the degree of noise interference at the corresponding frequency, thereby more finely adjusting the weight of the wrist acceleration spectrum, making the subsequent weight calculation and wrist acceleration spectrum weighted processing more reasonable and effective, which can improve the noise elimination effect in the presence of time error or multiple sensors, and improve the accuracy and reliability of remote monitoring.

[0041] In some preferred embodiments, step S53 includes: S531, for the value interval, using Gaussian filtering to smooth the noise intensity ratio; S532, calculating the arithmetic mean of the noise intensity ratio after smoothing to obtain a preliminary noise intensity ratio mean; S533. Determine whether the preliminary noise intensity ratio mean value ratio exceeds the preset noise threshold. If so, correct the preliminary noise intensity ratio mean value to the noise threshold and use it as the noise intensity ratio mean value of the corresponding frequency point. If not, directly use the preliminary noise intensity ratio mean value as the noise intensity ratio mean value of the corresponding frequency point.

[0042] Specifically, in order to cope with the fluctuation of the noise intensity ratio within the value interval, in step S531, Gaussian filtering and smoothing processing is performed on the noise intensity ratio within the value interval to reduce volatility and provide a more reliable data basis for the subsequent calculation of the mean value of the noise intensity ratio. Gaussian filtering and smoothing processing can use a one-dimensional Gaussian filter, and the size of the filter can be adjusted according to the actual noise situation. For example, for stationary noise, a Gaussian filter with a smaller standard deviation can be used, and for impulse noise, a Gaussian filter with a larger standard deviation can be used.

[0043] More specifically, in step S532, the arithmetic mean of the noise intensity ratio after smoothing is calculated to obtain a preliminary noise intensity ratio mean. Since the arithmetic mean has been smoothed, it is less affected by abnormal values ​​and can better represent the average noise intensity level of the frequency.

[0044] More specifically, in step S533, a noise threshold is introduced, and the preliminary noise intensity ratio mean is compared with the noise threshold. If the noise threshold is exceeded, the preliminary noise intensity ratio mean is corrected to the noise threshold, thereby limiting the excessive impact of noise and ensuring the rationality of weight calculation. If the noise threshold is not exceeded, the preliminary noise intensity ratio mean is directly adopted. The noise threshold can be set according to empirical values ​​or experimental data, for example, it can be set to 1.5 times or 2 times the baseline intensity of the ambient noise.

[0045] More specifically, after executing the above steps, a more accurate and reasonable noise intensity ratio mean can be obtained, making the noise intensity ratio mean more stable and reliable, providing a more accurate noise reference for the subsequent wrist acceleration spectrum weighting, thereby providing more accurate parameters for the weight calculation function in the subsequent step S54 to calculate the weight of the wrist acceleration spectrum at the corresponding frequency, thereby improving the accuracy of the health risk identification model in risk level assessment.

[0046] In some preferred embodiments, step S531 includes: S5311, estimating the background noise intensity of the ambient sound to obtain a background noise intensity value; S5312, determining the standard deviation of the Gaussian filter according to the background noise intensity value, wherein the larger the background noise intensity value, the larger the standard deviation; S5313. Based on the standard deviation, a Gaussian filter is used to convolve the noise intensity ratio within the value interval to obtain a smoothed noise intensity ratio.

[0047] Specifically, the process of estimating the background noise intensity value in step S5311 adopts a variety of methods to estimate the background noise intensity. For example, the minimum intensity value in the ambient sound spectrum within a period of time can be taken as the background noise intensity value, or a statistical method can be used, such as calculating the percentile of the ambient sound intensity within a period of time, and taking the lower percentile, such as the 10th percentile, as the background noise intensity value to eliminate the interference of occasional noise; the purpose of this step is to obtain the noise floor level of the current ambient sound, aiming to solve the problem that the smoothing process cannot adapt to the background noise intensity of different ambient sounds.

[0048] More specifically, in step S5312, there is a preset functional relationship between the background noise intensity value and the Gaussian filter standard deviation, so the corresponding standard deviation can be determined according to the background noise intensity value, so that the degree of smoothing can be adaptively adjusted with the noise intensity. Among them, the functional relationship can be a linear function or a nonlinear function, such as a logarithmic function or an exponential function, to meet the filtering requirements under different noise intensity ranges. When the ambient noise intensity is high, a Gaussian filter with a larger standard deviation is used for smoothing, which can more effectively remove noise interference and retain valid signals. Conversely, when the ambient noise intensity is low, a Gaussian filter with a smaller standard deviation is used to avoid signal distortion caused by excessive smoothing.

[0049] More specifically, the Gaussian filter convolution process in step S5313 includes: using a discrete Gaussian kernel to convolve with the noise intensity ratio sequence, the size and shape of the discrete Gaussian kernel are determined by the standard deviation determined in step S5312, the larger the standard deviation, the wider the Gaussian kernel, and the stronger the smoothing effect, and the convolution operation is to slide the Gaussian filter on the value range of the noise intensity ratio and perform weighted averaging to achieve the purpose of smoothing the noise intensity ratio.

[0050] More specifically, through the above steps, the smoothing processing of the noise intensity ratio in the remote monitoring method for the elderly in the embodiment of the present application can better adapt to different noise environments and can effectively improve the accuracy of subsequent noise intensity ratio mean calculation.

[0051] In some preferred embodiments, step S4 comprises: S41, extracting frequency characteristics of each frequency according to the ambient sound spectrum; S42, matching the frequency characteristics with reference frequency characteristics in a preset household appliance sound spectrum database to determine whether each frequency belongs to the household appliance sound frequency, wherein the household appliance sound spectrum database stores reference frequency characteristics of various household appliances in different working states; S43. Calculate the ratio of the ambient sound intensity of each frequency to the ambient noise baseline intensity. If the frequency is determined to be a household appliance sound frequency, the ratio is used as the noise intensity ratio of the frequency. Otherwise, the ratio is multiplied by a preset attenuation factor and the result is used as the noise intensity ratio of the frequency.

[0052] Specifically, in step S41, the extraction of frequency features can be achieved through spectrum analysis, and parameters such as spectrum peak frequency, spectrum energy distribution, spectrum bandwidth, etc. that can effectively characterize the frequency characteristics of sound are extracted. These frequency characteristics can characterize the distribution characteristics of sound in frequency.

[0053] More specifically, in step S42, the frequency features of sound samples of various household appliances in different working states are stored in the household appliance sound spectrum database, and the matching process can be specifically: similarity calculation is performed between the frequency features extracted in step S41 and the reference frequency features in the database, for example, the Euclidean distance or cosine similarity between the extracted spectrum peak frequency and the reference spectrum peak frequency of various household appliance sounds stored in the database is calculated. A similarity threshold is set, and when the calculated similarity meets the threshold condition, it is determined that the current frequency belongs to the household appliance sound frequency, thereby achieving a preliminary judgment on the source of the ambient sound.

[0054] More specifically, in step S43, the baseline intensity of environmental noise can be pre-calibrated or estimated in real time, and the attenuation factor is preset to a value less than 1, such as 0.5 or 0.3, to moderately reduce the noise intensity ratio of non-household appliance sounds and highlight the weight of abnormal sounds. This step adopts a differentiated noise intensity ratio calculation method for sound frequencies from different sources, which can reduce the interference of household appliance sounds in subsequent health risk identification, increase the weight of abnormal sounds that truly reflect the health status of the elderly, and thus improve the accuracy of monitoring.

[0055] In some preferred embodiments, step S6 comprises: S61, respectively calculating the energy entropy of the weighted wrist acceleration spectrum and the ambient sound spectrum to obtain the wrist acceleration energy entropy and the ambient sound energy entropy; S62, determining the splicing weight of the wrist acceleration spectrum and the ambient sound spectrum according to the wrist acceleration energy entropy and the ambient sound energy entropy, if the wrist acceleration energy entropy is greater than the ambient sound energy entropy, increasing the splicing weight of the wrist acceleration spectrum, otherwise, increasing the splicing weight of the ambient sound spectrum; S63. Perform weighted splicing on the weighted wrist acceleration spectrum and the ambient sound spectrum according to the splicing weight to form a fusion feature vector.

[0056] Specifically, the energy entropy in step S61 is used as an indicator to measure the amount of signal information, which can be understood as the uniformity of the spectrum energy distribution. The more uneven the spectrum energy distribution is, the smaller the energy entropy is. Conversely, the more uniform the spectrum energy distribution is, the larger the energy entropy is. Generally speaking, the larger the energy entropy is, the more information may be contained in the spectrum. Therefore, by calculating the energy entropy of the wrist acceleration spectrum and the ambient sound spectrum, the amount of information contained in the two can be evaluated. There are many ways to calculate the energy entropy. For example, first calculate the energy of each frequency component in the spectrum, then normalize the energy to a probability distribution, and finally calculate the energy entropy according to the information entropy formula.

[0057] More specifically, step S62 determines the splicing weight based on the magnitude relationship of energy entropies. Specifically, if the wrist acceleration energy entropy is greater than the ambient sound energy entropy, it indicates that the wrist acceleration spectrum contains relatively more information. At this time, the splicing weight of the wrist acceleration spectrum needs to be increased. Conversely, if the ambient sound energy entropy is greater than the wrist acceleration energy entropy, the splicing weight of the ambient sound spectrum needs to be increased so that the fused feature vector can retain more features of the spectrum with a large amount of information.

[0058] More specifically, step S63 is used to splice the weighted wrist acceleration spectrum and the ambient sound spectrum according to the splicing weights, thereby forming a final fused feature vector. The splicing process can use a linear weighting method to linearly combine the weighted wrist acceleration spectrum and the ambient sound spectrum according to the determined weights to obtain the final fused feature vector. The size of the splicing weight directly affects the contribution of the corresponding spectrum in the fused feature vector, so that the fused feature vector can focus more on the spectrum containing more information, reduce the interference that may be caused by the spectrum with less information, and thus improve the accuracy of the subsequent health risk identification model in risk level assessment.

[0059] In some preferred embodiments, step S62 includes: S621, analyzing the historical health data of the elderly, obtaining the correlation between the daily activity types of the elderly and the corresponding environmental sound types, and constructing an activity-sound correlation model; S622, identifying the current activity type of the elderly based on the current wrist acceleration spectrum, and predicting the expected environmental sound type under the current activity type based on the activity-sound association model; S623. Perform sound type identification on the ambient sound spectrum to determine whether the actual ambient sound type is consistent with the expected ambient sound type. If they are consistent, determine the splicing weight of the wrist acceleration spectrum and the ambient sound spectrum based on the ratio of the wrist acceleration energy entropy to the ambient sound energy entropy. If they are inconsistent, reduce the weight of the frequency component corresponding to the sound type that does not match the expected ambient sound type in the ambient sound spectrum, and increase the splicing weight of the wrist acceleration spectrum.

[0060] Specifically, the construction of the activity sound association model is achieved by analyzing the historical health data of the elderly. The historical health data contains records of the types of daily activities of the elderly and records of the types of environmental sounds when the activities occur. Activity types include, for example, sitting, walking, sleeping, etc., and environmental sound types include, for example, TV sound, conversation sound, no sound, etc. The activity sound association model is used to characterize the types of environmental sounds expected to appear under specific activity types. Activity type recognition is completed based on the wrist acceleration spectrum. For example, a machine learning classification algorithm, such as a support vector machine or a random forest, can be used to train an activity type recognition model. The input of the activity type recognition model is the wrist acceleration spectrum feature, and the output is the activity type that the elderly are currently in. Sound type recognition is performed on the environmental sound spectrum. For example, sound event detection technology can be used to pre-establish a sound type database, and the sound type in the environmental sound spectrum can be identified through pattern matching or machine learning methods.

[0061] More specifically, the process of determining the splicing weight is as follows: when the actual ambient sound type is consistent with the expected ambient sound type, it can be directly calculated based on the ratio of wrist acceleration energy entropy to ambient sound energy entropy; when the actual ambient sound type is inconsistent with the expected ambient sound type, the weight of the frequency component corresponding to the inconsistent sound type in the ambient sound spectrum is reduced; for example, if the wrist acceleration energy entropy is twice the ambient sound energy entropy, the splicing weight of the wrist acceleration spectrum can be set to 0.67, and the splicing weight of the ambient sound spectrum can be set to 0.33.

[0062] More specifically, the above processing method introduces the activity sound association model, so that the determination of the splicing weight is no longer solely dependent on the ratio of energy entropy, but also takes into account the correlation between the activity type of the elderly and the environmental sound type. Therefore, in the case of complex and changeable environmental sounds, the determination of the splicing weight can be more accurate and reasonable, and the quality of the fused feature vector is improved, ultimately improving the accuracy and robustness of the health risk identification model.

[0063] In some preferred embodiments, the health risk identification model includes a plurality of pre-trained health risk identification sub-models, and step S7 includes: S71, inputting the fused feature vector into multiple health risk identification sub-models respectively to obtain multiple health risk level assessment results; S72, determining the fusion weight of multiple health risk level assessment results according to the matching rate between the historical health data of the elderly and the health risk level assessment results of different health risk identification sub-models; S73. Use fusion weights to perform weighted fusion on multiple health risk level assessment results, obtain the risk level to determine the current health status of the elderly, and generate an alarm message when the risk level exceeds a preset threshold.

[0064] Specifically, in response to the problem of insufficient accuracy of the evaluation using a single health risk identification model, the remote monitoring method for the elderly in the embodiment of the present application proposes a processing method using multiple pre-trained health risk identification sub-models. When evaluating the health risk level, step S71 simultaneously inputs the fused feature vector into the multiple sub-models, and each sub-model independently evaluates the health risk to obtain its own risk level evaluation result. Next, step S72 considers the accuracy of each sub-model in the past evaluation of the health data of the elderly, and assigns a suitable fusion weight to each sub-model by analyzing the degree of matching between the historical health data of the elderly and the evaluation results of different sub-models. The sub-model with a high matching rate indicates that the evaluation result is more in line with the health status of the elderly, so it is given a higher weight. Subsequently, step S73 uses these fusion weights to perform weighted fusion on the evaluation results of multiple sub-models, comprehensively considers the evaluation opinions of each sub-model, and obtains a final, more comprehensive and accurate risk level. The final risk level is used to judge the current health status of the elderly, and when the risk level exceeds the preset threshold, an alarm message is generated to achieve effective monitoring and risk warning of the health status of the elderly. The above processing method can effectively improve the accuracy and reliability of health risk identification by integrating the evaluation results of multiple sub-models and dynamically adjusting the weights based on historical data.

[0065] More specifically, the number of health risk identification sub-models can be expanded according to actual needs; the determination method of fusion weights can be more refined, for example, not only considering the historical matching rate, but also dynamically adjusting the weights in combination with the current physiological state of the elderly, environmental factors, etc. The weighted fusion method can also adopt a nonlinear fusion method to better adapt to the interaction between different sub-models. Health risk identification sub-models may include but are not limited to: a sub-model based on a support vector machine, a sub-model based on a neural network, and a sub-model based on a decision tree. Each sub-model uses different algorithms and model structures to capture different aspects of health risks. When determining the fusion weight, the calculation of the matching rate can adopt a cross-validation method to more accurately evaluate the generalization ability of each sub-model and its adaptability to the health data of the elderly. The preset threshold can be adjusted according to the actual application scenario and the tolerance for risk. For example, in a scenario that is more sensitive to risk, the preset threshold can be set lower so that alarm information can be generated earlier.

[0066] In some preferred embodiments, step S3 comprises: S31, preprocessing the wrist acceleration information and the ambient sound information respectively, wherein the preprocessing includes removing zero-frequency components and performing normalization processing to obtain preprocessed wrist acceleration information and ambient sound information; S32, performing windowing processing on the preprocessed wrist acceleration information and environmental sound information respectively, using a Hamming window to reduce spectrum leakage, and obtaining windowed wrist acceleration information and environmental sound information; S33, performing fast Fourier transform on the windowed wrist acceleration information and the ambient sound information respectively to obtain a wrist acceleration spectrum and an ambient sound spectrum.

[0067] Specifically, in step S31, the preprocessing operation is used to improve data quality, among which the step of removing the zero-frequency component can effectively eliminate the DC offset in the signal, avoid unnecessary interference of the DC component to the subsequent spectrum analysis, and ensure the accuracy of the spectrum analysis. The normalization processing can adjust the data amplitudes from different sensors to a uniform scale range, eliminate the influence caused by the difference in sensor sensitivity or signal amplitude, provide a consistent data basis for the construction of the subsequent fusion feature vector, and enhance the system stability and generalization ability.

[0068] More specifically, in step S32, windowing is used to optimize the spectrum analysis process. The Hamming window is selected because it can effectively smooth the truncation edge of the signal, thereby significantly reducing the occurrence of spectrum leakage and improving the accuracy of spectrum analysis, so that the spectrum energy can more concentratedly reflect the real frequency components of the signal. More specifically, in step S33, a fast Fourier transform is applied to the wrist acceleration information and the ambient sound information after the windowing process, in order to obtain the wrist acceleration spectrum and the ambient sound spectrum. After preprocessing and windowing, the spectrum quality is improved, and the frequency components of the wrist acceleration information and the ambient sound information can be more accurately reflected, providing a more reliable data basis for subsequent feature extraction and health risk identification.

[0069] In some preferred embodiments, the method further comprises the following steps performed between step S2 and step S3: SA, fuzzy processing of environmental sound information, including: A1. After collecting the ambient sound information, extract multiple sound feature parameters of the ambient sound information, where the sound feature parameters include volume, pitch, and frequency distribution; A2. For each sound feature parameter, determine the corresponding fuzzy membership function, which is used to describe the membership degree of the sound feature parameter on different fuzzy sets; A3. According to the fuzzy membership of each sound feature parameter, fuzzy reasoning rules are used to perform reasoning to obtain the fuzzification result of the environmental sound information. The fuzzy reasoning rules are established based on expert knowledge or historical data and are used to determine the fuzzy set to which the current environmental sound information belongs; A4. Use the fuzzified result of the ambient sound information as new ambient sound information.

[0070] Specifically, in the above processing method, the fuzzification of environmental sound information is introduced after the environmental sound information is collected and before the fast Fourier transform; the key to the above processing method is to fuzzify the environmental sound information so that it can express uncertainty and ambiguity, effectively filter out pseudo-anomalies caused by daily activities and sensor noise, greatly reduce the possibility of false alarms, and improve the practicality of the system. The fuzzy environmental sound information can more accurately reflect the real sound state of the environment in which the elderly are located, provide a more reliable data basis for subsequent health risk identification, and thus improve the accuracy and robustness of the monitoring system.

[0071] More specifically, the extraction of sound feature parameters can be implemented in the following ways: for volume, it can be obtained by calculating the root mean square value of the ambient sound information in the time domain; for pitch, a fundamental frequency detection algorithm can be used, such as the autocorrelation method or the average amplitude difference function method for estimation; for frequency distribution, the ambient sound information can first be subjected to a fast Fourier transform to obtain a spectrum, and then the energy distribution of the spectrum can be analyzed, such as calculating the spectrum centroid or extracting the Mel-frequency cepstrum coefficients.

[0072] More specifically, the fuzzy membership function can select the appropriate fuzzy set and membership function type according to the physical meaning and value range of the sound feature parameters. For example, for volume, three fuzzy sets of "low volume", "medium volume" and "high volume" can be set, and a triangular or Gaussian membership function can be used to quantify the membership of the volume value in each fuzzy set.

[0073] More specifically, the establishment of fuzzy inference rules can be based on the summary of experts' cognition and experience of environmental sounds, or by analyzing a large amount of historical environmental sound data, summarizing the distribution law of characteristic parameters of different types of environmental sounds, and thus formulating fuzzy inference rules. For example, a rule can be set: if the volume is "low volume" and the frequency distribution is concentrated in the low frequency band, then the possibility that the environmental sound information belongs to the "background noise" fuzzy set is high.

[0074] More specifically, the fuzzification result can be a representation of the fuzzy set to which the ambient sound belongs, which can be used as new ambient sound information. The original ambient sound information can be adjusted or weighted according to the fuzzy reasoning result. For example, if the fuzzy reasoning result indicates that the ambient sound information contains more noise components, the original ambient sound information can be denoised to obtain the denoised ambient sound information as new ambient sound information.

[0075] More specifically, by using the above fuzzification processing to take the fuzzification result as new ambient sound information for subsequent spectrum analysis, the interference of ambient noise on the subsequent spectrum analysis can be reduced, and the quality and accuracy of the ambient sound spectrum can be improved.

[0076] Second, please refer to Figure 2 Some embodiments of the present application also provide a remote monitoring device for the elderly, which is used to remotely monitor the elderly who carry a wrist wearable device, and the device includes: A first acquisition module 201 is used to acquire wrist acceleration information based on an acceleration sensor in a wrist wearable device; A second acquisition module 202 is used to acquire ambient sound information based on an ambient sound sensor in a wrist wearable device; The spectrum conversion module 203 is used to perform fast Fourier transformation on the wrist acceleration information and the ambient sound information to obtain the wrist acceleration spectrum and the ambient sound spectrum; The noise analysis module 204 is used to analyze the ambient sound spectrum to identify the frequency range characteristics of the household appliance sound, calculate the ratio of the noise intensity of the ambient sound information at each frequency to the ambient noise baseline intensity, and obtain the noise intensity ratio; A weighting module 205 is used to calculate the weight of the wrist acceleration spectrum at the corresponding frequency according to the noise intensity ratio using a weight calculation function to obtain a weighted wrist acceleration spectrum; A feature concatenation module 206 is used to concatenate the weighted wrist acceleration spectrum with the ambient sound spectrum to form a fused feature vector; The risk assessment module 207 is used to perform risk level assessment based on the fused feature vector and a preset health risk identification model to determine the current health status of the elderly, and to generate an alarm message when the risk level exceeds a preset threshold.

[0077] The remote monitoring device for the elderly in the embodiment of the present application realizes the monitoring of the health status of the elderly and abnormal warning. It synchronously collects wrist acceleration information and environmental sound information through acceleration sensors and environmental sound sensors to reflect the movement status and environmental sound conditions of the elderly. It can effectively distinguish the frequency distribution characteristics of vibration noise and electrical noise through frequency domain processing, and form a fusion feature vector by splicing the weighted wrist acceleration spectrum and the environmental sound spectrum. It comprehensively utilizes the movement information and environmental sound information to comprehensively reflect the health status of the elderly. It can effectively suppress the interference of household appliance sound noise on acceleration information, improve the accuracy of health risk identification, and provide more reliable technical means for remote monitoring of the elderly.

[0078] In some preferred implementations, the second acquisition module 202 is further configured to perform fuzzy processing on the ambient sound information.

[0079] In some preferred embodiments, the remote monitoring device for the elderly of the embodiments of the present application is used to execute the remote monitoring method for the elderly provided in the first aspect above.

[0080] Third, please refer to Figure 3 Some embodiments of the present application also provide a structural diagram of an electronic device. The present application provides an electronic device, including: a processor 301 and a memory 302. The processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanisms (not marked). The memory 302 stores computer-readable instructions executable by the processor 301. When the electronic device is running, the processor 301 executes the computer-readable instructions to execute the method in any optional implementation of the above embodiments.

[0081] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method in any optional implementation of the above embodiment is executed. The computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk.

[0082] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0083] In addition, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0084] Furthermore, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0085] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0086] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A remote monitoring method for the elderly, used for remote monitoring of the elderly who wear a wrist wearable device, characterized in that: The steps of the method include: S1, obtaining wrist acceleration information based on an acceleration sensor in a wrist wearable device; S2, obtaining ambient sound information based on an ambient sound sensor in a wrist wearable device; S3, performing fast Fourier transform on the wrist acceleration information and the ambient sound information respectively to obtain a wrist acceleration spectrum and an ambient sound spectrum; S4, analyzing the ambient sound spectrum to identify the frequency range characteristics of household appliance sounds, calculating the ratio of the noise intensity of the ambient sound information at each frequency to the ambient noise baseline intensity, and obtaining a noise intensity ratio; S5. Calculate the weight of the wrist acceleration spectrum at the corresponding frequency using a weight calculation function according to the noise intensity ratio, and obtain a weighted wrist acceleration spectrum; S6, concatenating the weighted wrist acceleration spectrum with the ambient sound spectrum to form a fusion feature vector; S7. Based on the fused feature vector, a preset health risk identification model is used to perform risk level assessment to determine the current health status of the elderly, and an alarm message is generated when the risk level exceeds a preset threshold.

2. A remote monitoring method for the elderly according to claim 1, characterized in that: There are multiple acceleration sensors and / or multiple ambient sound sensors, and step S5 includes: S51, obtaining a time synchronization error range of the wrist acceleration spectrum and the ambient sound spectrum; S52, determining a value interval of a noise intensity ratio of a corresponding ambient sound spectrum for each frequency in the wrist acceleration spectrum and the time synchronization error range; S53, calculating the mean value of the noise intensity ratio corresponding to each frequency within the value range; S54, using a weight calculation function to calculate the weight of the wrist acceleration spectrum at the corresponding frequency according to the mean value of the noise intensity ratio, and performing weighted processing on the wrist acceleration spectrum based on the weight to obtain a weighted wrist acceleration spectrum.

3. A remote monitoring method for the elderly according to claim 2, characterized in that: Step S53 includes: S531, for the value interval, using Gaussian filtering to smooth the noise intensity ratio; S532, calculating the arithmetic mean of the noise intensity ratio after smoothing to obtain a preliminary noise intensity ratio mean; S533. Determine whether the preliminary noise intensity ratio mean value ratio exceeds a preset noise threshold value. If so, correct the preliminary noise intensity ratio mean value to the noise threshold value and use it as the noise intensity ratio mean value of the corresponding frequency point. If not, directly use the preliminary noise intensity ratio mean value as the noise intensity ratio mean value of the corresponding frequency point.

4. A remote monitoring method for the elderly according to claim 1, characterized in that: Step S4 includes: S41, extracting frequency characteristics of each frequency according to the ambient sound spectrum; S42, matching the frequency characteristics with reference frequency characteristics in a preset household appliance sound spectrum database to determine whether each frequency belongs to a household appliance sound frequency, wherein the household appliance sound spectrum database stores reference frequency characteristics of multiple household appliances in different working states; S43. Calculate the ratio of the ambient sound intensity to the ambient noise baseline intensity for each frequency. If the frequency is determined to be a household appliance sound frequency, use the ratio as the noise intensity ratio of the frequency. Otherwise, multiply the ratio by a preset attenuation factor as the noise intensity ratio of the frequency.

5. A remote monitoring method for the elderly according to claim 1, characterized in that: Step S6 includes: S61, respectively calculating the energy entropy of the weighted wrist acceleration spectrum and the ambient sound spectrum to obtain the wrist acceleration energy entropy and the ambient sound energy entropy; S62, determining the splicing weight of the wrist acceleration spectrum and the ambient sound spectrum according to the wrist acceleration energy entropy and the ambient sound energy entropy, if the wrist acceleration energy entropy is greater than the ambient sound energy entropy, increasing the splicing weight of the wrist acceleration spectrum, otherwise, increasing the splicing weight of the ambient sound spectrum; S63. Perform weighted splicing on the weighted wrist acceleration spectrum and the ambient sound spectrum according to the splicing weight to form a fusion feature vector.

6. A remote monitoring method for the elderly according to claim 1, characterized in that: The health risk identification model includes a plurality of pre-trained health risk identification sub-models, and step S7 includes: S71, inputting the fused feature vector into a plurality of health risk identification sub-models respectively to obtain a plurality of health risk level assessment results; S72, determining the fusion weight of multiple health risk level assessment results according to the matching rate between the historical health data of the elderly and the health risk level assessment results of different health risk identification sub-models; S73. Use fusion weights to perform weighted fusion on multiple health risk level assessment results, obtain the risk level to determine the current health status of the elderly, and generate an alarm message when the risk level exceeds a preset threshold.

7. A remote monitoring method for the elderly according to claim 1, characterized in that: The method further comprises the steps performed between step S2 and step S3: SA, fuzzy processing of environmental sound information, including: A1. After collecting the ambient sound information, extract multiple sound feature parameters of the ambient sound information, where the sound feature parameters include volume, pitch, and frequency distribution; A2. For each sound feature parameter, determine the corresponding fuzzy membership function, which is used to describe the membership degree of the sound feature parameter on different fuzzy sets; A3. According to the fuzzy membership of each sound feature parameter, fuzzy reasoning rules are used to perform reasoning to obtain the fuzzification result of the environmental sound information. The fuzzy reasoning rules are established based on expert knowledge or historical data and are used to determine the fuzzy set to which the current environmental sound information belongs; A4. Use the fuzzified result of the ambient sound information as new ambient sound information.

8. A remote monitoring device for the elderly, used for remote monitoring of the elderly who wear a wrist wearable device, characterized in that: The device includes: A first acquisition module, configured to acquire wrist acceleration information based on an acceleration sensor in a wrist wearable device; A second acquisition module is used to acquire environmental sound information based on an environmental sound sensor in the wrist wearable device; A spectrum conversion module, used for performing fast Fourier transform on the wrist acceleration information and the ambient sound information respectively to obtain a wrist acceleration spectrum and an ambient sound spectrum; A noise analysis module, used to analyze the ambient sound spectrum to identify the frequency range characteristics of household appliance sounds, calculate the ratio of the noise intensity of the ambient sound information at each frequency to the ambient noise baseline intensity, and obtain a noise intensity ratio; A weighting module, used to calculate the weight of the wrist acceleration spectrum at the corresponding frequency using a weight calculation function according to the noise intensity ratio, and obtain a weighted wrist acceleration spectrum; A feature concatenation module is used to concatenate the weighted wrist acceleration spectrum with the ambient sound spectrum to form a fused feature vector; The risk assessment module is used to perform risk level assessment based on the fused feature vector and a preset health risk identification model to determine the current health status of the elderly, and generate an alarm message when the risk level exceeds a preset threshold.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the steps in the method according to any one of claims 1 to 7 are executed.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are executed.

Images