A smart clothing management method based on multi-sensing information fusion

Abstract

This invention discloses a smart clothing management method based on multi-sensor information fusion, specifically relating to the field of physiological signal monitoring. It addresses the problem that existing smart clothing suffers from sensor position shifts and signal distortion due to fabric deformation during movement, leading to inaccurate monitoring data and difficulty in reliable analysis. By incorporating a physiological signal acquisition unit and a flexible deformation sensing network within the clothing, the method acquires raw physiological signals and fabric deformation distribution data in real time. Local deformation features are extracted from the deformation distribution data, and the reliability weights of the physiological signals are calculated. The interference is estimated using a deformation-interference mapping model, and dynamic compensation is applied to the raw physiological signals to obtain correction signals. Furthermore, the correction signals from different regions are accumulated and filtered based on weights to extract the optimal estimated signal. When the optimal estimated signal exceeds a reasonable range, a wearing adjustment prompt is generated.

Description

Technical Field

[0001] This invention relates to the field of physiological signal monitoring technology, and more specifically, to a smart clothing management method based on multi-sensor information fusion. Background Technology

[0002] Existing smart clothing products are gradually expanding from simple wearing comfort to multifunctionality and health monitoring. By embedding various flexible sensors in clothing fabrics, they can collect physiological signals such as electrocardiogram, respiration, and pulse, providing new possibilities for sports and fitness, rehabilitation management, and telemedicine.

[0003] However, in real-world applications, smart clothing differs significantly from traditional wearable devices. The biggest challenge lies in the fact that clothing, as a carrier, undergoes complex deformations with human movement. For example, during running, bending over, or raising an arm, the fabric experiences longitudinal stretching, lateral twisting, and localized wrinkles. These dynamic deformations directly affect the fit between the sensor and the skin, leading to deviations and distortions in physiological signals. For instance, in electrocardiogram (ECG) monitoring, electrode misalignment relative to the heart's anatomical position can cause abnormal waveform amplitude or distorted shape, resulting in misjudgments in heart rate and rhythm analysis. Similarly, in respiratory monitoring, when the chest cavity's rise and fall point deviates from the optimal sensing area, the signal amplitude is underestimated, failing to accurately reflect respiratory rate and depth.

[0004] Current technologies largely rely on inertial sensors to identify the overall motion state, followed by simple filtering or labeling of the signals. However, this approach cannot capture the microscopic deformation of the fabric itself and its specific interference with individual sensor signals, resulting in insufficient data reliability. Especially in scenarios requiring continuous monitoring in long-term, dynamic environments, such as daily rehabilitation for patients with chronic diseases, home health management for the elderly, and tracking the training status of athletes, ensuring the accuracy of physiological signals under constantly changing clothing deformation has become a key technological bottleneck restricting the practical application of this type of smart clothing. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a smart clothing management method based on multi-sensor information fusion to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A smart clothing management method based on multi-sensor information fusion includes the following steps:

[0008] S1. Acquire raw physiological signals through the physiological signal acquisition unit and collect fabric deformation distribution data through the flexible deformation sensor network.

[0009] S2. Based on the fabric deformation distribution data, extract the local deformation features of the area where the physiological signal acquisition unit is located;

[0010] S3. Based on local deformation features, calculate the confidence weight of the original physiological signal, and estimate the amount of interference of deformation on the physiological signal through a pre-established deformation-interference mapping relationship model.

[0011] S4. Dynamically compensate the original physiological signal based on the amount of interference to obtain the corrected physiological signal sequence;

[0012] S5. Calculate the cumulative value of the confidence weight for the corrected physiological signal sequences in the regions where different physiological signal acquisition units are located, and extract the optimal estimated signal.

[0013] S6. When the parameters of the optimal estimated signal exceed the preset reasonable range, generate a clothing wearing status adjustment prompt message.

[0014] In a preferred embodiment, step S1, which involves acquiring raw physiological signals through a physiological signal acquisition unit and acquiring fabric deformation distribution data through a flexible deformation sensing network, specifically includes:

[0015] Raw physiological signals are collected based on the built-in fixed physiological signal acquisition unit of the clothing, with each physiological signal acquisition unit corresponding to a target anatomical location;

[0016] The original physiological signal includes amplitude, waveform morphology, and sampling timestamp;

[0017] Based on the sensor unit group arranged in matrix form in the flexible deformation sensor network, the fabric strain data of each sensor unit location is collected at a fixed sampling interval and then filtered.

[0018] The filtered fabric strain data is bound to the matrix position of the sensing unit to generate real-time deformation field data showing the relationship between strain magnitude and spatial distribution on the two-dimensional static plane of the garment.

[0019] In a preferred embodiment, step S2, extracting the local deformation features of the area where the physiological signal acquisition unit is located based on the fabric deformation distribution data, specifically includes:

[0020] Based on the initial static positional relationship between the physiological signal acquisition unit and the sensing unit matrix, a mapping association table between real-time deformation field data and physiological signal acquisition unit is established.

[0021] Acquire the strain data set of each physiological signal acquisition unit within a preset radius area, and use it as the fabric deformation distribution data corresponding to the area where the physiological signal acquisition unit is located;

[0022] Statistical analysis of strain data within the area where the physiological signal acquisition unit is located yields the average strain value and strain fluctuation degree of the corresponding area.

[0023] Principal component analysis was used to determine the main direction of strain distribution in the region where each physiological signal acquisition unit is located, and the main deformation direction was recorded.

[0024] The average strain value, strain fluctuation degree, and main deformation direction of the area where each physiological signal acquisition unit is located are integrated into local deformation characteristics.

[0025] In a preferred embodiment, in step S3, the confidence weight of the original physiological signal is calculated based on local deformation features, and the amount of interference of deformation on the physiological signal is estimated through a pre-established deformation-interference mapping model. Specifically, this includes:

[0026] The confidence weight value of the original physiological signal is set and adjusted according to the degree of strain fluctuation in the local deformation characteristics, and the confidence weight value is inversely proportional to the degree of strain fluctuation.

[0027] The average strain value, strain fluctuation degree and feature quantity corresponding to the main deformation direction are input into the pre-trained deformation-interference mapping relationship model, and the output is the original physiological signal amplitude deviation estimate and signal waveform distortion estimate corresponding to the current local deformation characteristics.

[0028] The amplitude deviation estimate and the waveform distortion estimate are combined to generate a complete signal interference description vector.

[0029] In a preferred embodiment, the process of establishing the pre-trained deformation-interference mapping model is as follows:

[0030] Three predetermined deformation modes are applied to the garment: longitudinal stretching, lateral twisting, and combined deformation.

[0031] Record the strain data set output by the flexible deformation sensing network under various deformation modes, and use a reference physiological sensor that is not built into the clothing to collect the reference signal sequence of physiological signals.

[0032] The collected strain data set is standardized, and the amplitude deviation and waveform distortion index between the reference signal sequence and the output signal sequence of the physiological signal acquisition unit built into the clothing are calculated as signal interference.

[0033] A piecewise regression method was used to establish a mathematical mapping relationship between standardized strain data and signal interference.

[0034] The mapping relationship data established under different deformation modes are integrated into a three-dimensional feature lookup table, which records the numerical relationship between three feature values—average strain, strain fluctuation degree, and main deformation direction angle—and the corresponding signal amplitude deviation estimate and waveform distortion estimate.

[0035] The feature lookup table is subjected to cubic spline interpolation to generate a continuous and smooth deformation-interference mapping model.

[0036] In a preferred embodiment, step S4, which involves dynamically compensating the original physiological signal based on the amount of interference to obtain the corrected physiological signal sequence, specifically includes:

[0037] The signal amplitude deviation estimate is extracted from the signal interference description vector, and amplitude compensation is performed on the original physiological signal. The compensation method includes subtracting the amplitude deviation or scaling proportionally.

[0038] The parameters of the compensation filter are constructed based on the waveform distortion estimate of the original physiological signal, and the compensation filter is used to perform convolution operation on the amplitude-compensated signal.

[0039] Phase correction is performed on the original physiological signal after convolution based on the phase of the baseline physiological signal.

[0040] The signal, after amplitude compensation, convolution operation, and phase correction, is output as a corrected physiological signal sequence.

[0041] In a preferred embodiment, step S5, which involves calculating the cumulative value of the confidence weight for the corrected physiological signal sequences located in different physiological signal acquisition units and extracting the optimal estimated signal, specifically includes:

[0042] Obtain the corrected physiological signal sequences and corresponding confidence weight values ​​for the regions where different physiological signal acquisition units are located;

[0043] Time alignment is performed on the corrected physiological signal sequences of all physiological signal acquisition units in the region, and the cumulative value of the confidence weight corresponding to each timestamp is calculated.

[0044] Within a preset sliding time window, the period with the largest cumulative average value of the confidence weight is taken as the optimal estimation period, and the corrected physiological signal sequence of each physiological signal acquisition unit within the corresponding period is extracted as the optimal estimation signal.

[0045] In a preferred embodiment, step S6, when the parameters of the optimal estimated signal exceed a preset reasonable range, specifically includes generating clothing wearing status adjustment prompt information including:

[0046] The amplitude range and waveform morphology of the optimal estimated signal of all physiological signal acquisition units are continuously monitored. When any parameter exceeds the preset reasonable range threshold, the area where the corresponding physiological signal acquisition unit is located is located and a prompt message is generated.

[0047] The type of adjustment prompt is determined based on the type of parameter exceeding the limit. The categories include clothing fit adjustment prompts and clothing wearing method correction prompts.

[0048] The technical effects and advantages of the intelligent clothing management method based on multi-sensor information fusion of this invention are as follows:

[0049] This method transforms the traditional passive filtering approach into an active method that utilizes fabric deformation feedback for dynamic calibration and compensation, significantly improving the stability and reliability of physiological signal acquisition. By constructing real-time deformation field data through a flexible deformation sensor network and combining local feature extraction with a deformation-interference mapping model, signal deviation and waveform distortion can be estimated in real time under different motion states or wearable scenarios, achieving multi-dimensional compensation for amplitude, waveform, and phase. This method avoids the signal quality degradation caused by sensor offset and contact impedance changes due to clothing stretching and twisting, ensuring that key indicators such as ECG, respiration, and pulse maintain clinically usable accuracy even in dynamic environments. Simultaneously, a reliability weighting mechanism is introduced to selectively fuse signals from multiple acquisition units, automatically eliminating low-reliability or severely interfered signal channels, improving the robustness of the final output.

[0050] In terms of user experience, this invention can also guide users to adjust the fit or wearing method of clothing through a prompting mechanism, further ensuring the quality of data collection. Overall, this method realizes a paradigm shift from "interference caused by clothing deformation" to "compensation through clothing deformation," and has significant anti-interference capabilities, individual adaptability, and long-term monitoring stability, which will help promote the large-scale application of smart clothing in medical health, sports rehabilitation, and daily health management. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of a smart clothing management method based on multi-sensor information fusion according to the present invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0053] Example 1

[0054] Figure 1 This invention presents a smart clothing management method based on multi-sensor information fusion, which includes the following steps:

[0055] S1. Acquire raw physiological signals through the physiological signal acquisition unit and collect fabric deformation distribution data through the flexible deformation sensor network.

[0056] S2. Based on the fabric deformation distribution data, extract the local deformation features of the area where the physiological signal acquisition unit is located;

[0057] S3. Based on local deformation features, calculate the confidence weight of the original physiological signal, and estimate the amount of interference of deformation on the physiological signal through a pre-established deformation-interference mapping relationship model.

[0058] S4. Dynamically compensate the original physiological signal based on the amount of interference to obtain the corrected physiological signal sequence;

[0059] S5. Calculate the cumulative value of the confidence weight for the corrected physiological signal sequences in the regions where different physiological signal acquisition units are located, and extract the optimal estimated signal.

[0060] S6. When the parameters of the optimal estimated signal exceed the preset reasonable range, generate a clothing wearing status adjustment prompt message.

[0061] In S1, raw physiological signals are acquired through a physiological signal acquisition unit, and fabric deformation distribution data are acquired through a flexible deformation sensing network.

[0062] Multiple fixed locations are pre-set within the garment structure as physiological signal acquisition points. Each acquisition point corresponds to a key anatomical location, including ECG electrodes in the midline of the chest, respiratory impedance sensing units in the shoulder and back area, and blood flow or pulse optical sensing modules in the arm or thigh area. To ensure the matching of fixed locations with anatomical features, during the production stage, the acquisition devices are embedded into the fabric via conductive fibers or flexible wiring paths through garment pattern design. During sewing, a strong bond is maintained between the acquisition units and the local fabric structure of the garment to prevent positional drift during wearing, removal, or washing. Each physiological signal acquisition unit outputs three types of data during operation: instantaneous amplitude, waveform morphology, and sampling timestamp. The amplitude data directly reflects the intensity of the electrical or optical signal at a specific moment at the acquisition site; the waveform morphology is a formatted curve structure of the amplitude sequence continuously acquired over a period of time; and the sampling timestamp marks the precise acquisition time of each sampling point, ensuring temporal consistency between different units. The raw data from each unit is sequentially cached by internal storage circuitry.

[0063] Embedded within the fabric layer of smart clothing are a matrix of flexible deformation sensing units, constructed from flexible strain gauges or conductive polymer films. These units are arranged in a two-dimensional grid, with each unit corresponding to a coordinate point on the garment's two-dimensional plane. For example, a 20x20 matrix structure is used, with the spacing between each sensing unit fixed at approximately 1 cm to ensure high-resolution strain monitoring coverage across most of the garment. During operation, the deformation sensing network acquires data at fixed sampling intervals, with a sampling period set to 10 milliseconds (100Hz), ensuring the capture of dynamic stretching and compression changes in the fabric even under intense movement. Each sensing unit outputs real-time strain change data at its location, expressed as a change in unit resistance or capacitance with fabric deformation. All unit data first undergoes filtering using a digital low-pass filter with a cutoff frequency set to 30Hz to effectively eliminate high-frequency noise interference and retain the main trend signals of garment deformation. After filtering, the strain data of each unit is bound to its position index in the matrix, establishing a correspondence between position and strain value. Based on this binding relationship, real-time deformation field data on the two-dimensional parameter domain of the garment is constructed. This deformation field records the distribution of strain on the entire garment plane at a certain sampling moment, including the boundaries of the tensile zone, compression zone, and relatively stable zone. Through continuous sampling and updating, a dynamic deformation field sequence that evolves over time can be formed.

[0064] In step S2, local deformation features of the area where the physiological signal acquisition unit is located are extracted based on the fabric deformation distribution data.

[0065] The relative static positional relationship between each physiological signal acquisition unit and the matrix units in the flexible deformation sensing network is clearly defined. Specifically, a unified coordinate reference frame is established on the parameter domain of the unfolded garment, and all sensing unit matrices are sequentially mapped to coordinate points according to their row and column numbers. Simultaneously, each physiological signal acquisition unit is fixed at its corresponding anatomical target location during garment fabrication, such as the midline of the chest cavity, the left lower intercostal space, or the right upper shoulder. These locations are also mapped to the same coordinate frame, obtaining the reference coordinates of each physiological signal acquisition unit on the two-dimensional plane of the garment. In this way, the spatial proximity relationship between the physiological signal acquisition unit and the surrounding sensing unit matrix positions can be established in the initial static state. When real-time deformation field data is generated, the strain value of each sensing unit carries row and column numbers and coordinate information, thus allowing direct retrieval of this mapping to find the set of sensing units within a preset range surrounding a given physiological signal acquisition unit.

[0066] Once the mapping relationship between real-time deformation field data and the acquisition unit is established, a certain spatial range is selected during actual operation. The strain data of all sensing units within this range are used as the local deformation input of the physiological signal acquisition unit. This spatial range is set in the form of a radius, for example, 2 cm or 3 cm. The specific value is determined during clothing design based on the fabric elasticity and the density of the sensing units. Taking a radius of 3 cm as an example, when a certain acquisition unit is located at a coordinate point in a two-dimensional plane, all sensing units with an Euclidean distance of less than or equal to 3 cm from that point are retrieved, and their strain values ​​at the current moment are extracted to form a data set. This set includes the stretching, compression, or torsion of the area surrounding the acquisition unit. As time progresses, this process is repeated for each sampling cycle, thus continuously generating a time-series local strain data set. To ensure data stability, obvious invalid values ​​in the sensor unit output are excluded when extracting the set. For example, extreme abnormal data caused by open circuits or short circuits can be removed by using a preset threshold (e.g., data exceeding ±20% of the normal strain range is considered invalid). The resulting set can fully reflect the fabric deformation state of the local area of ​​the acquisition unit at the current moment.

[0067] After acquiring the strain data set around the acquisition unit, statistical calculations are performed to obtain the average strain value and strain fluctuation degree of the region. Specifically, the strain values ​​of all sensing units within the set are arithmetically averaged; the result is the average strain value of the region, used to characterize the overall tensile or compressive trend of the area. Then, the deviation of all strain values ​​within the set from the average value is calculated, and the fluctuation is expressed as the standard deviation; this standard deviation represents the strain fluctuation degree. A greater strain fluctuation degree indicates stronger local differences and uneven stretching in the region; for example, some areas may experience significant stretching while others remain almost unchanged. The statistical process is performed once in each sampling period, thus forming a time-series of average strain values ​​and strain fluctuation degree.

[0068] Further identification of the main directional characteristics of strain within the region is achieved. Specifically, this involves creating a vector set by combining the position coordinates of all sensing units within a preset radius with their corresponding strain values, and inputting this vector into principal component analysis (PCA). PCA constructs a covariance matrix and solves for the eigenvectors to obtain the principal direction of strain in the two-dimensional plane. This direction represents the most significant tensile or compressive trend in the local deformation distribution. For example, if the principal direction angle is 75 degrees at a given sampling moment, it indicates that the region is primarily stretched along an axis at a 75-degree angle to the horizontal. This principal deformation direction is then integrated with the aforementioned average strain value and strain fluctuation degree to form the local deformation feature vector of the acquisition unit at that moment.

[0069] In step S3, the credibility weight of the original physiological signal is calculated based on the local deformation characteristics, and the amount of interference of deformation on the physiological signal is estimated through a pre-established deformation-interference mapping relationship model.

[0070] Under experimental conditions, different types of standardized deformation patterns were applied to smart clothing to ensure coverage of the main deformation types that might occur during actual wear. Specifically, the clothing was fixed to a standard human body simulation model or an adjustable mechanical stretching platform, and three predetermined deformation patterns—longitudinal stretching, transverse twisting, and combined longitudinal and transverse deformation—were applied sequentially. Longitudinal stretching refers to stretching the clothing by a certain percentage along the long axis of the human body, for example, setting the stretching range to 5%, 10%, and 15% of the original length. Transverse twisting involves fixing one edge and rotating the other edge to create 15-degree, 30-degree, and 45-degree lateral twists. Combined deformation applies stretching and twisting simultaneously in both the longitudinal and transverse directions, for example, a 10% longitudinal stretch combined with a 30-degree transverse twist. Each deformation state was maintained for a fixed time period (e.g., 30 seconds) to ensure sufficient data collection. During this process, a flexible deformation sensing network embedded in the clothing continuously collected the strain output of each matrix sensing unit at a fixed sampling frequency (e.g., 100Hz), forming a time-series data set. Simultaneously, high-precision reference physiological sensors, not embedded in the clothing, are attached to the subject's skin surface to synchronously acquire baseline physiological signal sequences. For example, standard electrocardiogram waveforms are acquired using medical-grade electrocardiogram electrodes, and respiratory baseline curves are acquired using a high-precision breathing belt. These baseline signals are strictly aligned with the signals embedded in the clothing on the time axis and are used as a reference for subsequent difference calculations.

[0071] After acquiring the strain data set and the baseline physiological signal sequence, the raw data were standardized to ensure data comparability under different deformation modes. Specifically, the strain value of each sensing unit was subtracted from the static initial value, and then divided by the measurement range to obtain the normalized dimensionless strain value. Subsequently, using the reference physiological signal as the true value, the difference between the signal from the clothing's built-in acquisition unit and the reference signal was calculated. The differences were divided into two categories: amplitude deviation, obtained by comparing the amplitudes of the two signals point by point; and waveform distortion, measured by calculating the similarity between the two signals, using the degree of difference in correlation reduction or key waveform feature points (such as QRS width and peak position difference) to generate a numerical distortion index. These two types of differences were combined as signal interference, corresponding to amplitude deviation and waveform distortion, respectively. Then, a piecewise regression method was used to establish the mapping relationship between normalized strain and interference. Specifically, the strain data interval was divided into several continuous small segments, and a linear regression function was independently fitted within each segment to output a linear approximate relationship between strain characteristics and interference. The fitting results for each segment are recorded in the form of regression coefficients and error indices, forming a segmented regression model database.

[0072] After completing piecewise regression fitting under different deformation modes, all regression relationship data are integrated into a unified three-dimensional feature lookup table. The construction process involves calculating three core feature values ​​from the local strain data: average strain value, strain fluctuation degree, and principal deformation direction angle. These three features are used as the input dimensions of the lookup table, forming a three-dimensional index space. Each set of indices corresponds to a set of output values, namely, amplitude deviation estimate and waveform distortion estimate. By successively filling in data under different deformation modes and different sampling intervals, a comprehensive lookup table is finally formed. Due to the limited number of experimental points actually collected, cubic spline interpolation is performed on the lookup table to ensure its continuity and smoothness in the input space. Specifically, within the three-dimensional input space, cubic spline interpolation is used to smoothly fit the output values ​​of adjacent sample points, generating a continuous function surface, thus ensuring smooth and stable output results at any input point. After processing, the final deformation-disturbance mapping relationship model is obtained, which reflects both the local average tensile state and the fluctuation and directional characteristics. In this way, the model can quickly look up tables and output the corresponding disturbance estimates during actual operation.

[0073] Based on the degree of strain fluctuation in local deformation characteristics, dynamically adjusted reliability weights are assigned to the original physiological signals. The reliability weights are determined by the magnitude of the strain fluctuation. The degree of strain fluctuation is obtained by calculating the dispersion of strain data within the acquisition area. Smaller fluctuations indicate a more stable clothing fit and a more reliable physiological signal, thus warranting a higher weight; larger fluctuations indicate uneven local deformation and increased signal interference, thus requiring a lower weight. The degree of strain fluctuation is divided into three intervals: low, medium, and high. In the low fluctuation interval, the weight is close to 1, decreasing monotonically with increasing fluctuation, and approaching 0 in the high fluctuation interval. This mapping relationship is determined through fitting pre-experimental data to ensure stable and consistent weight values ​​under different deformation conditions.

[0074] The average strain value, strain fluctuation degree, and characteristic quantities corresponding to the main deformation direction are input into a pre-trained deformation-interference mapping model. The model outputs estimates of the original physiological signal amplitude deviation and waveform distortion, corresponding to the current local deformation characteristics. These estimates are then combined to generate a complete signal interference description vector. After obtaining the amplitude deviation and waveform distortion estimates, they are integrated into a single interference description vector. Specifically, the amplitude deviation estimate is used as a benchmark to describe the overall signal intensity shift; for example, in ECG monitoring, this deviation corresponds to the rise or fall of the overall QRS complex potential. Secondly, the waveform distortion estimate is used as an indicator to describe signal morphological changes; for example, in respiratory monitoring, this distortion reflects the broadening of the inspiratory peak or the asymmetry of the expiratory curve. To ensure that the combined result reflects both amplitude changes and morphological distortion, the two estimates are standardized to make them comparable within the same numerical range. Finally, they are concatenated as a vector to form an interference description vector containing two dimensions.

[0075] In step S4, the original physiological signal is dynamically compensated according to the amount of interference to obtain the corrected physiological signal sequence.

[0076] Amplitude compensation is performed on the original physiological signal to eliminate the overall amplitude drift caused by local deformation of the clothing fabric. Specifically, an amplitude deviation estimate is extracted from the signal interference description vector. This estimate represents the overall increase or decrease in signal amplitude relative to a reference value at that moment. The compensation process employs two methods: first, directly subtracting the deviation value to restore the signal to the expected amplitude level; second, when the deviation exhibits a proportional change, scaling is used to adjust the overall signal by a scaling factor. To ensure the stability of the compensation, upper and lower limits are set during subtraction or scaling. For example, when the deviation estimate exceeds a set range (e.g., exceeding 30% of the reference amplitude), saturation processing is performed to avoid overcompensation.

[0077] After amplitude compensation, further correction is performed on waveform distortion. Specifically, the corresponding compensation filter parameters are constructed based on the waveform distortion estimate in the interference description vector. The distortion estimate reflects the morphological differences between the original signal and the reference signal, such as QRS complex width expansion in an ECG signal or blunting of the inspiratory segment of a respiratory curve. The design principle of the compensation filter is to suppress these distortion frequency bands introduced by deformation while preserving the effective components of the signal. The specific process involves establishing a correspondence between the distortion estimate and filter parameters using a lookup table in preliminary experiments. During actual operation, this parameter set is directly called to generate a set of filter coefficients. Subsequently, the amplitude-compensated signal is convolved with this filter. The result of the convolution selectively corrects the frequency components of the original signal, attenuating the distortion.

[0078] After convolution processing, although the amplitude and morphology of the signal are largely restored, its phase characteristics may still be shifted. Phase shift manifests as a time delay or advance of key points such as signal peaks and troughs relative to the reference signal. To ensure the signal's temporal consistency with the reference physiological signal, phase correction is performed. Specifically, a standard phase characteristic sequence of the reference physiological signal is established. For example, in an ECG signal, the standard time distribution of the P wave, QRS wave, and T wave is determined; in a respiratory signal, the standard positions of the inspiratory peak and expiratory trough are determined. During actual operation, the convolution-processed signal is compared with this standard phase characteristic sequence, and the phase difference between key points is calculated. The sampling time axis of the signal is adjusted according to the phase difference. For example, if an overall delay of 2 milliseconds is detected, the entire signal sequence is shifted forward by 2 milliseconds; if a local segment is detected as advanced, a local interpolation shift is performed within that segment. After correction, the signal's key time points are aligned with the reference characteristics, ensuring consistent phase characteristics. Finally, the signal output after amplitude compensation, convolution filtering, and phase correction forms the corrected physiological signal sequence.

[0079] In step S5, the cumulative value of the confidence weight is calculated for the corrected physiological signal sequences in the regions where different physiological signal acquisition units are located, and the optimal estimated signal is extracted.

[0080] The outputs of each physiological signal acquisition unit, after interference compensation and amplitude correction, are collected to form the corrected physiological signal sequences. These sequences have undergone dynamic compensation and phase correction in previous stages, enabling them to more accurately reflect the physiological state at the corresponding anatomical locations. Simultaneously, each sequence is associated with a confidence weight value, derived from real-time calculations based on local deformation characteristics. This weight value is limited to a range of 0 to 1, with values ​​closer to 1 indicating higher stability and reliability. In practice, the outputs of all acquisition units are archived at a fixed sampling frequency. Each sampling point includes the signal amplitude, waveform morphology, and corresponding confidence weight. This creates a multi-channel parallel signal data stream based on the acquisition units. Each data stream contains both time-series physiological signal values ​​and their paired confidence weight values.

[0081] After acquiring multiple signal sequences, the timelines of all acquisition units are aligned. Specifically, using the global sampling clock as a reference, the sampling points of different signal channels are precisely matched according to their timestamps. If a channel is missing a point due to delay or sampling offset, it is filled in using interpolation of adjacent points, ensuring that all channels have valid data at the same timestamp. After time alignment, weight accumulation begins. At each timestamp, the reliability weight value corresponding to all acquisition units is extracted, and these weights are summed to obtain the cumulative weight value for that time point. This cumulative value reflects the overall reliability level of the signal group at that moment: a higher cumulative value indicates that more signal sources are stable and consistent at that moment; a lower cumulative value indicates a decrease in signal reliability in a local area. During the calculation process, to avoid the influence of a single extremely abnormal channel on the overall performance, a maximum limit is set on the weight of a single channel, which cannot exceed 30% of the total weight, thus ensuring that the cumulative value reflects the overall characteristics rather than the abrupt changes of individual signals. Finally, a cumulative weight sequence corresponding to each timestamp is output.

[0082] After obtaining the complete cumulative weight sequence, it is filtered within a set sliding time window to determine the optimal estimation period. The length of the sliding time window is 5 to 10 seconds, ensuring sufficient coverage of the signal period without introducing excessive delay. Specifically, the time window is slid across the cumulative weight sequence sequentially, and the average cumulative weight is calculated for all time points within that window at each position. Then, the average values ​​of all window positions are compared, and the window with the largest value is selected as the optimal estimation period. This period corresponds to the interval with the highest overall signal reliability. After determining the optimal period, the corrected physiological signal sequences of all acquisition units within this time window are extracted as candidate data and fused into the optimal estimation signal. The fusion method is not a simple superposition, but rather combines the weight values ​​of each signal acquisition unit, filtering out channels with consistently low weights within the period, retaining only signal trajectories with high weights, and integrating them into the target output. The resulting optimal estimation signal not only has the highest overall reliability within the period but also takes into account the complementarity between different channels, thus providing stable data input for subsequent wearability status determination or other diagnostic analyses.

[0083] In step S6, when the parameters of the optimal estimated signal exceed a preset reasonable range, a clothing wearing status adjustment prompt message is generated.

[0084] The optimal estimated signals from all physiological signal acquisition units are continuously monitored. The monitoring process involves two types of parameters: amplitude range and waveform morphology. Amplitude range is determined by recording the difference between the peak and trough values ​​of the signal within a continuous sampling period. For example, for ECG signals, the amplitude is typically stable between 0.5 mV and 2 mV; a long-term amplitude below 0.3 mV may indicate loose electrode contact, while an amplitude above 3 mV indicates abnormal gain. For photoplethysmography (PPG) signals, the amplitude is typically between 0.2 and 1.0 units; a value below 0.1 indicates optical path obstruction, while a value above 1.5 may indicate excessive sensor pressure. For respiratory impedance signals, normal amplitude fluctuations are between 0.1 and 0.5 ohms; a value below 0.05 ohms indicates poor fit, while a value above 1 ohm indicates sensor misalignment. Waveform morphology is determined by comparing key feature points of the actual signal with a pre-stored standard waveform feature set. For example, an electrocardiogram (ECG) signal should have a clear P wave, QRS complex, and T wave structure. If a segment is continuously missing or the peak width is excessively extended, it is considered distorted. A pulsed image (PPG) should have regular rising and falling edges. If a plateau-like or broken waveform appears, it is abnormal. By combining the above amplitude range and waveform morphology, a closed loop of regional localization and abnormality detection is achieved.

[0085] When any parameter is detected to exceed the reasonable threshold, a prompt message is immediately generated, and the prompt category is distinguished according to the type of exceedance. The triggering process is as follows: based on the judgment result, determine which type of signal has exceeded the limit, and lock the corresponding physiological signal acquisition unit area. For example, when the ECG signal amplitude is too low and the waveform is severely distorted, locate the electrode unit in the chest area; when the respiratory signal amplitude is too high, locate the abdominal or subcostal impedance band unit. Then proceed to the category determination stage. Category 1 is a clothing fit adjustment prompt, triggered when the abnormality is due to loose fit or insufficient contact area. For example, if the PPG signal amplitude is below the lower limit, it means that the optical probe has detached from the skin, and the user needs to be prompted to adjust the tightness of the clothing. Category 2 is a clothing wearing method correction prompt, triggered when the abnormality is due to incorrect wearing position or local misalignment. For example, if the respiratory impedance signal amplitude is consistently too high, it indicates that the impedance band position has shifted to a non-target anatomical site, and the user is prompted to readjust the wearing position. The prompt message includes three items when it is generated: the abnormal collection unit area, the abnormal type, and the adjustment category. It is output through the user terminal in a clear text or graphic prompt manner to ensure that the user can quickly understand and perform the corresponding operation.

[0086] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0087] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0088] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0089] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0090] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0091] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0092] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0093] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0094] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0095] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart clothing management method based on multi-sensor information fusion, characterized in that, Includes the following steps: S1. Acquire raw physiological signals through the physiological signal acquisition unit and collect fabric deformation distribution data through the flexible deformation sensor network. S2. Based on the fabric deformation distribution data, extract the local deformation features of the area where the physiological signal acquisition unit is located; S3. Based on local deformation features, calculate the confidence weight of the original physiological signal, and estimate the amount of interference of deformation on the physiological signal through a pre-established deformation-interference mapping relationship model. S4. Dynamically compensate the original physiological signal based on the amount of interference to obtain the corrected physiological signal sequence; S5. Calculate the cumulative value of the confidence weight for the corrected physiological signal sequences in the regions where different physiological signal acquisition units are located, and extract the optimal estimated signal. S6. When the parameters of the optimal estimated signal exceed the preset reasonable range, generate a clothing wearing status adjustment prompt message; In step S2, extracting the local deformation features of the area where the physiological signal acquisition unit is located based on the fabric deformation distribution data specifically includes: Based on the initial static positional relationship between the physiological signal acquisition unit and the corresponding sensor unit matrix position of the flexible deformation sensing network, a mapping association table between real-time deformation field data and physiological signal acquisition unit is established. Acquire the strain data set of each physiological signal acquisition unit within a preset radius area, and use it as the fabric deformation distribution data corresponding to the area where the physiological signal acquisition unit is located; Statistical analysis of strain data within the area where the physiological signal acquisition unit is located yields the average strain value and strain fluctuation degree of the corresponding area. Principal component analysis was used to determine the main direction of strain distribution in the region where each physiological signal acquisition unit is located, and the main deformation direction was recorded. The average strain value, strain fluctuation degree, and main deformation direction of the area where each physiological signal acquisition unit is located are integrated into local deformation characteristics. In step S3, based on local deformation features, the confidence weight of the original physiological signal is calculated, and the amount of interference of deformation on the physiological signal is estimated through a pre-established deformation-interference mapping model. Specifically, this includes: The confidence weight value of the original physiological signal is set and adjusted according to the degree of strain fluctuation in the local deformation characteristics, and the confidence weight value is inversely proportional to the degree of strain fluctuation. The average strain value, strain fluctuation degree and feature quantity corresponding to the main deformation direction are input into the pre-trained deformation-interference mapping relationship model, and the output is the original physiological signal amplitude deviation estimate and signal waveform distortion estimate corresponding to the current local deformation characteristics. The amplitude deviation estimate and the waveform distortion estimate are combined to generate a complete signal interference description vector; The specific process for establishing the pre-trained deformation-interference mapping relationship model is as follows: Three predetermined deformation modes are applied to the garment: longitudinal stretching, lateral twisting, and combined deformation. Record the strain data set output by the flexible deformation sensing network under various deformation modes, and use a reference physiological sensor that is not built into the clothing to collect the reference signal sequence of physiological signals. The collected strain data set is standardized, and the amplitude deviation and waveform distortion index between the reference signal sequence and the output signal sequence of the physiological signal acquisition unit built into the clothing are calculated as signal interference. A piecewise regression method was used to establish a mathematical mapping relationship between standardized strain data and signal interference. The mapping relationship data established under different deformation modes are integrated into a three-dimensional feature lookup table, which records the numerical relationship between three feature values—average strain, strain fluctuation degree, and main deformation direction angle—and the corresponding signal amplitude deviation estimate and waveform distortion estimate. The feature lookup table is subjected to cubic spline interpolation to generate a continuous and smooth deformation-interference mapping model. In step S4, the dynamic compensation of the original physiological signal based on the interference level to obtain the corrected physiological signal sequence specifically includes: The signal amplitude deviation estimate is extracted from the signal interference description vector, and amplitude compensation is performed on the original physiological signal. The amplitude compensation method includes subtracting the amplitude deviation or scaling proportionally. The parameters of the compensation filter are constructed based on the waveform distortion estimate of the original physiological signal, and the compensation filter is used to perform convolution operation on the amplitude-compensated signal. Phase correction is performed on the original physiological signal after convolution based on the phase of the baseline physiological signal. The signal, after amplitude compensation, convolution operation, and phase correction, is output as a corrected physiological signal sequence.

2. The intelligent clothing management method based on multi-sensor information fusion according to claim 1, characterized in that, In step S1, the acquisition of raw physiological signals through the physiological signal acquisition unit and the acquisition of fabric deformation distribution data through the flexible deformation sensing network specifically include: Raw physiological signals are collected based on the built-in fixed physiological signal acquisition unit of the clothing, with each physiological signal acquisition unit corresponding to a target anatomical location; The original physiological signal includes amplitude, waveform morphology, and sampling timestamp; Based on the sensor unit group arranged in matrix form in the flexible deformation sensor network, the fabric strain data of each sensor unit location is collected at a fixed sampling interval and then filtered. The filtered fabric strain data is bound to the matrix position of the sensing unit to generate real-time deformation field data showing the relationship between strain magnitude and spatial distribution on the two-dimensional static plane of the garment.

3. The intelligent clothing management method based on multi-sensor information fusion according to claim 1, characterized in that, In step S5, the cumulative value calculation of the confidence weight values ​​of the corrected physiological signal sequences in the regions where different physiological signal acquisition units are located, and the extraction of the optimal estimated signal, specifically includes: Obtain the corrected physiological signal sequences and corresponding confidence weight values ​​for the regions where different physiological signal acquisition units are located; Time alignment is performed on the corrected physiological signal sequences of all physiological signal acquisition units in the region, and the cumulative value of the confidence weight corresponding to each timestamp is calculated. Within a preset sliding time window, the period with the largest cumulative average value of the confidence weight is taken as the optimal estimation period, and the corrected physiological signal sequence of each physiological signal acquisition unit within the corresponding period is extracted as the optimal estimation signal.

4. The intelligent clothing management method based on multi-sensor information fusion according to claim 1, characterized in that, In step S6, when the parameters of the optimal estimated signal exceed a preset reasonable range, generating a clothing wearing status adjustment prompt message specifically includes: The amplitude range and waveform morphology of the optimal estimated signal of all physiological signal acquisition units are continuously monitored. When any parameter exceeds the preset reasonable range threshold, the area where the corresponding physiological signal acquisition unit is located is located and a prompt message is generated. The type of adjustment prompt is determined based on the type of parameter exceeding the limit. The categories include clothing fit adjustment prompts and clothing wearing method correction prompts.

Images