A data storage, data reconstruction method, device, equipment and medium

Abstract

The application provides a data storage and data reconstruction method, device, equipment and medium, wherein the method comprises: acquiring multiple groups of photoplethysmography (PPG) signals; distinguishing the multiple groups of PPG signals into a group of reference signals and at least one group of candidate signals; for each group of candidate signals, determining a target signal corresponding to the candidate signal from the multiple groups of PPG signals; calculating a conversion relationship between the target signal and the candidate signal; storing the reference signal and the conversion relationship between each group of candidate signals and the target signal corresponding thereto. Through the above method, the storage efficiency can be improved while ensuring the integrity of the key information of the signal.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more specifically, to a method, apparatus, device, and medium for data storage and data reconstruction. Background Technology

[0002] Photoplethysmography (PPG) is a physiological signal that uses optical principles to non-invasively detect changes in blood volume and pulse waves in the microvascular bed of tissue. The conventional acquisition method involves using a single-wavelength or multi-wavelength photoelectric sensor in contact with the body surface. A photoelectric converter then transforms the transmitted or reflected light intensity signal into an electrical signal, thereby acquiring and analyzing multiple physiological parameters such as heart rate, blood oxygen saturation, vascular elasticity, and respiratory rate.

[0003] However, during their in-depth research, the inventors discovered that with the extension of continuous monitoring time and the expansion of application scenarios, the storage and transmission of massive PPG signal data has become a bottleneck restricting its large-scale application. The high sampling rate and large data volume of the original PPG signal place excessive demands on the storage medium capacity, embedded system resources, and wireless communication bandwidth, significantly increasing hardware costs and system power consumption.

[0004] Therefore, a solution is needed. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method, apparatus, device, and medium for data storage and data reconstruction to efficiently store multiple sets of PPG signals.

[0006] In a first aspect, embodiments of this application provide a data storage method, the method comprising: Acquire multiple sets of photoplethysmography (PPG) signals; at least two sets of PPG signals show some differences. Multiple PPG signals are divided into a reference signal and at least one candidate signal; the signal quality of the reference signal is higher than that of the candidate signal. For each set of candidate signals, a target signal corresponding to the candidate signal is determined from the multiple sets of PPG signals; the similarity between the target signal and the candidate signal meets a preset requirement, and the target signal is the reference signal or a candidate signal whose transformation relationship has been calculated; Calculate the conversion relationship between the target signal and the candidate signal; the conversion relationship is used to reconstruct the candidate signal based on the target signal. The reference signal and the conversion relationship between each group of candidate signals and their corresponding target signal are stored.

[0007] In one feasible implementation, the multiple sets of PPG signals are obtained by at least one acquisition device at different acquisition locations, and / or based on light of different colors; Storing the reference signal includes: The core frequency component of each group of PPG signals is determined to obtain target frequency bands for multiple groups of PPG signals; each target frequency band includes each core frequency component. The reference signal is bandpass filtered based on the core frequency component set; Store the filtered reference signal.

[0008] In one feasible implementation, the method further includes: For each group of PPG signals, it is divided into multiple windows; any two windows have at least partial overlap. For each window, perform a Fourier transform on the data in the window to obtain the frequency domain signal; Based on the frequency domain signal, each group of PPG signals is compressed to obtain multiple compressed groups of PPG signals. Multiple PPG signals are divided into a reference signal and at least one candidate signal, including: The compressed PPG signals are divided into a reference signal and at least one candidate signal.

[0009] In one feasible implementation, compression of each group of PPG signals is performed based on the frequency domain signal, including: For each group of PPG signals in the frequency domain, a scaling factor is dynamically assigned to it based on the amplitude of each frequency domain coefficient; the scaling factor is proportional to the amplitude. The frequency domain signal is non-uniformly scaled based on the scaling factor corresponding to each frequency domain coefficient.

[0010] In one feasible implementation, calculating the transformation relationship between the target signal and the candidate signal includes: For each group of candidate signals, when the target signal is a reference signal, the transformation relationship between the reference signal and the group of candidate signals is calculated based on a preset algorithm; the preset algorithm includes: a generalized linear model and / or cross-correlation analysis; The first signal obtained based on the transformation relationship and the reference signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the first signal, the conversion relationship between the first signal and the group of candidate signals is calculated based on the preset algorithm; The second signal obtained based on the transformation relationship and the first signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the second signal, the conversion relationship between the second signal and the group of candidate signals is calculated based on the preset algorithm.

[0011] In one feasible implementation, the method further includes: Calculate the residual between the target signal and the candidate signal.

[0012] Based on the signal quality of the candidate signal, the degree of compression of the residual is determined; the degree of compression is inversely proportional to the signal quality. Based on the compression level, the residual of the candidate signal is compressed; Store the compressed residual.

[0013] Secondly, embodiments of this application also provide a data reconstruction method, the method comprising: Acquire a signal to be processed; the signal to be processed is stored based on any of the methods described in the first aspect; the signal to be processed includes at least: a reference signal, and a conversion relationship between a target signal and a candidate signal, wherein the target signal is the reference signal, or a candidate signal whose conversion relationship has been calculated; Based on the target signal and the conversion relationship between the target signal and the candidate signal to be reconstructed, each group of candidate signals is reconstructed.

[0014] Thirdly, embodiments of this application also provide a data storage device, the device comprising: The acquisition module is used to acquire multiple sets of photoplethysmography (PPG) signals; at least two sets of PPG signals have some differences between them. A differentiation module is used to differentiate multiple sets of PPG signals into a reference signal and at least one set of candidate signals; the signal quality of the reference signal is higher than that of the candidate signal. The determination module is used to determine the target signal corresponding to the candidate signal from the plurality of PPG signals for each candidate signal; the similarity between the target signal and the candidate signal meets a preset requirement, and the target signal is the reference signal or a candidate signal whose transformation relationship has been calculated; The calculation module is used to calculate the conversion relationship between the target signal and the candidate signal; the conversion relationship is used to reconstruct the candidate signal based on the target signal; A storage module is used to store the reference signal and the conversion relationship between each group of candidate signals and their corresponding target signal.

[0015] In one feasible implementation, the multiple sets of PPG signals are obtained by at least one acquisition device at different acquisition locations, and / or based on light of different colors; The storage module is used to store the reference signal, for the following purposes: The core frequency component of each group of PPG signals is determined to obtain target frequency bands for multiple groups of PPG signals; each target frequency band includes each core frequency component. The reference signal is bandpass filtered based on the core frequency component set; Store the filtered reference signal.

[0016] In one feasible implementation, the device further includes: A window module is used to divide each group of PPG signals into multiple windows; any two windows have at least partial overlap. The frequency domain conversion module is used to perform Fourier transform on the data in each window to obtain the frequency domain signal. A compression module is used to compress each group of PPG signals based on the frequency domain signal to obtain multiple compressed PPG signals. The differentiation module is used to distinguish multiple sets of PPG signals into a reference signal and at least one set of candidate signals, for the following purposes: The compressed PPG signals are divided into a reference signal and at least one candidate signal.

[0017] In one feasible implementation, the compression module is used to compress each group of PPG signals based on the frequency domain signal, for the following purposes: For each group of PPG signals in the frequency domain, a scaling factor is dynamically assigned to it based on the amplitude of each frequency domain coefficient; the scaling factor is proportional to the amplitude. The frequency domain signal is non-uniformly scaled based on the scaling factor corresponding to each frequency domain coefficient.

[0018] In one feasible implementation, the calculation module is used to calculate the transformation relationship between the target signal and the candidate signal, for the following purposes: For each group of candidate signals, when the target signal is a reference signal, the transformation relationship between the reference signal and the group of candidate signals is calculated based on a preset algorithm; the preset algorithm includes: a generalized linear model and / or cross-correlation analysis; The first signal obtained based on the transformation relationship and the reference signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the first signal, the conversion relationship between the first signal and the group of candidate signals is calculated based on the preset algorithm; The second signal obtained based on the transformation relationship and the first signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the second signal, the conversion relationship between the second signal and the group of candidate signals is calculated based on the preset algorithm.

[0019] In one feasible implementation, the computing module is further used for: Calculate the residual between the target signal and the candidate signal; Also includes: The first module is used to determine the degree of compression of the residual based on the signal quality of the candidate signal; the degree of compression is inversely proportional to the signal quality. The second module is used to compress the residual of the candidate signal based on the compression degree. The compression module is also used to store the compressed residual.

[0020] Fourthly, embodiments of this application also provide a data reconstruction apparatus, the apparatus comprising: A signal acquisition module is used to acquire a signal to be processed; the signal to be processed is stored based on any of the methods described in the first aspect above; the signal to be processed includes at least: a reference signal, and a conversion relationship between a target signal and a candidate signal, wherein the target signal is the reference signal, or a candidate signal whose conversion relationship has been calculated; The reconstruction module is used to reconstruct each set of candidate signals based on the target signal and the conversion relationship between the target signal and the candidate signals to be reconstructed.

[0021] Fifthly, embodiments of this application also provide an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the method as described in any one of the first or second aspects.

[0022] In a sixth aspect, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the method as described in any one of the first or second aspects.

[0023] This application provides a method, apparatus, device, and medium for data storage and reconstruction. After obtaining multiple sets of PPG signals, it first determines the set of reference signals with the best signal quality, then constructs a complete signal baseline using this set of reference signals. The remaining candidate signals are efficiently represented by establishing conversion relationships with the reference signals. Specifically, this solution transforms the stored content from multiple complete sets of PPG signals into a lightweight architecture of "one complete reference signal + multiple sets of conversion relationships." When it is necessary to restore the original data, the system can accurately reconstruct each candidate signal using the reference signal and the corresponding conversion relationships. This design achieves a qualitative leap in storage efficiency while ensuring the integrity of key signal information.

[0024] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A flowchart illustrating a data storage method provided in an embodiment of this application is shown.

[0027] Figure 2 A flowchart of a data reconstruction method provided by an embodiment of this application is shown.

[0028] Figure 3 A schematic diagram of the structure of a data storage device provided in an embodiment of this application is shown.

[0029] Figure 4 A schematic diagram of the structure of a data reconstruction apparatus provided in an embodiment of this application is shown.

[0030] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0032] Wearable devices such as smartwatches and smart rings commonly use photoplethysmography (PPG) technology for health monitoring. For example, they deduce key physiological parameters such as heart rate, blood oxygen saturation, and heart rate variability (HRV) by analyzing raw PPG signals.

[0033] With the development of sensor technology and sensing technology, the acquisition of PPG signals is no longer limited to a single device or a single mode: Within a single high-end wearable device, multiple wavelengths (such as green, red, and infrared light) light sources and light receivers are integrated, forming multiple acquisition channels. These PPG signals of different wavelengths carry complementary physiological information due to their specific absorption characteristics of blood components (such as oxyhemoglobin and deoxyhemoglobin).

[0034] At a broader level of personal health networks, users may simultaneously wear smartwatches, smart rings, ear-worn monitoring devices, and more. These heterogeneous devices constitute a distributed, multi-channel PPG acquisition system that captures multidimensional information about cardiovascular activity asynchronously or synchronously from different parts of the body (such as the wrist, fingertips, and earlobes) at the same or different wavelengths.

[0035] However, whether it's multiple photoelectric channels in a single device or a data acquisition network composed of multiple devices, they all face the same core contradiction: the huge gap between the rapidly increasing data dimensionality and information density and the limited storage, computing, and communication resources of terminal devices. While advanced algorithms such as deep learning crave all this raw data to train more robust and accurate models, the physical limitations of hardware make directly storing and transmitting all raw signals impractical.

[0036] Against this backdrop, this application provides a data storage method, which is briefly described below: For single-device, multi-wavelength scenarios, the wavelength channel with the best signal quality (such as green PPG with the highest signal-to-noise ratio in motion scenarios) can be dynamically selected as the reference signal, while other wavelengths (such as red and infrared PPG used for blood oxygen calculation) are used as candidate signals, and only the conversion relationship between them and the reference signal is stored. This essentially encapsulates the physiological and physical correlation between different wavelength signals using a mathematical model.

[0037] In multi-device collaborative scenarios, the signal collected by the device with the best acquisition location, least interference, and highest signal quality from the distributed device group can be selected as the global reference signal. Signals collected by other devices, regardless of their wavelength or acquisition location, are considered candidate signals and are characterized by calculating their conversion relationship with the global reference signal (or an intermediate signal with an established conversion relationship).

[0038] In this way, this solution compresses massive, heterogeneous raw PPG data into a highly efficient dataset consisting of "a core reference signal + a series of lightweight transformation relationships." This not only greatly alleviates storage and transmission pressure, but more importantly, it establishes a unified data representation paradigm, enabling the organic integration and management of PPG signals from different sources and with different characteristics.

[0039] When training data or advanced analysis is needed for cloud-based AI models, the system can quickly and faithfully reconstruct the original PPG waveform of any channel, any device, and any point in time based on this data structure.

[0040] The following is a detailed introduction. For example... Figure 1 As shown, it includes the following steps: Step 101: Acquire multiple sets of photoplethysmography (PPG) signals; at least two sets of PPG signals have some differences between them.

[0041] For example, this section describes a scenario where multiple PPG signals are acquired from multiple channels. At least two of the multiple PPG signals will have some differences. For example, their difference could be that they have different wavelengths.

[0042] In a typical application scenario, wearable devices such as smartwatches integrate multi-wavelength photoelectric sensing systems. These systems can emit light of different wavelengths (e.g., green, red, and infrared) almost synchronously and receive light signals reflected or transmitted from subcutaneous tissue, thus forming multiple independent signal acquisition channels. Although the PPG signals acquired by the green, red, and infrared channels are time-synchronized and reflect the same pulse rhythm, their waveform morphology, DC component, AC component amplitude, and the degree of influence from motion noise will exhibit significant and quantifiable differences.

[0043] If each PPG signal is stored directly, a huge amount of storage space would be required to acquire multiple PPG signals in this situation.

[0044] Step 102: Divide the multiple PPG signals into a reference signal and at least one candidate signal; the signal quality of the reference signal is higher than that of the candidate signal.

[0045] By analyzing the signal quality of each group of PPG signals, multiple groups of PPG signals are divided into a reference signal with the best signal quality, and the remaining PPG signals are determined as candidate signals. The candidate signals are the signals that need to be compressed and stored in this scheme.

[0046] The selection of a reference signal is crucial because its quality directly determines the overall fidelity of subsequent signal reconstruction.

[0047] The following example illustrates a comprehensive scheme for evaluating signal quality, integrating multiple quantization metrics from the time, frequency, and cepstral domains to comprehensively and objectively assess the quality of PPG signals: (1) Cepstral index: Cepstral peak significance: Calculate the peak intensity corresponding to the heart rate cycle in the cepstral domain of the signal. A significant and isolated peak indicates that the signal contains a strong and regular pulse wave component, and the signal quality is high.

[0048] Cepstral peak to mean ratio: The higher this ratio, the more prominent the heart rate component is relative to background noise and artifacts, and the purer the signal.

[0049] (2) Frequency domain indicators: Spectral entropy: Used to measure the uncertainty (or disorder) of a signal in the frequency domain. A high-quality PPG signal should have its power spectral energy concentrated near the heart rate and its harmonic frequencies, resulting in a low spectral entropy value. Conversely, a noisy signal has a dispersed spectral energy distribution and a high entropy value.

[0050] Signal-to-noise ratio (SNR): Estimated by analyzing the signal power spectrum, it is the ratio of signal power within a narrow band centered on the heart rate fundamental frequency and its harmonics to the noise power across the entire frequency band. A higher SNR indicates better signal quality.

[0051] (3) Time-domain indicators: Kurtosis: A measure of the shape of the amplitude distribution of a signal. A high-quality PPG pulse sequence typically has a more "sharp" amplitude distribution (i.e., high kurtosis) because it contains regular pulse peaks and low baseline fluctuations. Noise signals, on the other hand, have an amplitude distribution that is closer to a normal distribution (kurtosis close to 3).

[0052] Coefficient of variation (COV) of peak-to-peak interval: The ratio of the standard deviation to the mean of the peak-to-peak intervals (corresponding to heart rate intervals) of a continuous pulse wave. A value that is too low may indicate a rigid signal (loss of physiological authenticity), while a value that is too high indicates severe interference or a large number of missed / false detections of pulse waves. A moderate COV value that conforms to physiological patterns is a sign of a high-quality signal.

[0053] When evaluating the signal quality of each group of PPG signals, all or some of the above indicators are calculated for each PPG signal to be evaluated. Then, methods such as weighted summation, rule-based decision trees, and mathematical models can be used to obtain the comprehensive quality index for each group of PPG signals. Finally, the PPG signal with the highest comprehensive quality index is selected as the reference signal, and the remaining signals are automatically classified as candidate signals.

[0054] Step 103: For each group of candidate signals, determine the target signal corresponding to the candidate signal from the multiple groups of PPG signals; the similarity between the target signal and the candidate signal meets the preset requirements, and the target signal is the reference signal or a candidate signal whose conversion relationship has been calculated.

[0055] This step aims to find the most suitable "template" or "reference", i.e., the target signal, for each set of candidate signals that need to be compressed, so as to ensure that the transformation relationship of subsequent calculations is the most efficient and concise.

[0056] The similarity between the target signal and the candidate signal can be determined using a preset algorithm. A core approach is to calculate the correlation coefficient between the two signals in the time domain, frequency domain, or feature space; the higher the correlation coefficient, the higher the similarity. For example, the target signal can be determined based on the similarity of wavelengths.

[0057] Here, the "candidate signal whose transformation relationship has been calculated" is a virtual signal generated by a mathematical process. It is not the original data itself, but a high-precision copy reconstructed by combining the reference signal (or the previous such signal) with the corresponding transformation relationship.

[0058] In other words, it allows subsequent candidate signals to no longer be limited to the initial "gold standard" (i.e., the reference signal), but instead to choose a "predecessor" that is more similar to itself and has a closer wavelength as the target. This simplifies the transformation relationship describing the differences between the two signals, requires less storage space, and thus achieves extremely high overall compression efficiency. Essentially, it upgrades the compression process from a single "central radial" mode to a highly efficient, chain-like, or tree-like "pipeline" mode.

[0059] Step 104: Calculate the conversion relationship between the target signal and the candidate signal; the conversion relationship is used to reconstruct the candidate signal based on the target signal.

[0060] In one feasible implementation, for each group of candidate signals, when the target signal is a reference signal, the transformation relationship between the reference signal and the group of candidate signals is calculated based on a preset algorithm; the preset algorithm includes: a generalized linear model, and / or cross-correlation analysis; The first signal obtained based on the transformation relationship and the reference signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the first signal, the conversion relationship between the first signal and the group of candidate signals is calculated based on the preset algorithm; The second signal obtained based on the transformation relationship and the first signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the second signal, the conversion relationship between the second signal and the group of candidate signals is calculated based on the preset algorithm.

[0061] Specifically, the above describes a chain-like or tree-like signal processing and dependency establishment process. Its core idea is that not all candidate signals must be directly converted to the original reference signal. Instead, they can be converted to an "intermediate signal" that is closer to it in feature similarity. This "intermediate signal" itself is derived from the reference signal or a previous-level intermediate signal.

[0062] Generalized linear models (such as finding the mapping coefficients from the target signal to the candidate signal through linear regression) or cross-correlation analysis (used to find the best alignment and similarity between signals) are mathematical tools for calculating the transformation relationship between signals.

[0063] The "first signal" and "second signal" are not new original signals, but virtual signals reconstructed or estimated from the reference signal or the signal at the previous level by applying transformation relationships. They are labeled as "candidate signals whose transformation relationships have been calculated", meaning that they can serve as the "benchmark" for calculating transformation relationships for other candidate signals, that is, the target signal.

[0064] When this solution is applied to scenarios involving multiple wavelengths and multiple sets of PPG signals, a more intelligent strategy based on spectral physical characteristics can be introduced. The following example illustrates this: Because PPG signals of different wavelengths exhibit continuous morphological changes, signals with closer wavelengths typically have more similar waveform characteristics.

[0065] The following example, using a system encompassing four wavelengths—green (526nm), orange (600nm), red (665nm), and infrared (940nm)—illustrates two possible reference selection paths: (1) Scenario using the green light channel as the reference signal: When the green light signal is selected as the global reference signal, the determination of the target signal can follow the order of wavelength from near to far: a. For the "orange light" candidate signal, since its wavelength is closest to that of green light, the green light reference signal can be selected as the target signal, and the conversion relationship between the orange light candidate signal and the green light reference signal can be directly calculated.

[0066] b. For the "red light" candidate signal, the estimated "near-orange light" signal (i.e., the first signal) generated by linear transformation of the green light signal can be selected as the target signal, or the green light reference signal can be selected as the target signal. In this case, the target signal is the first signal. Calculate the transformation relationship between the estimated "near-orange light" signal and the red light candidate signal. Then, determine the near-red light generated by linear transformation of the red light candidate signal (based on the calculated transformation relationship) as the second signal.

[0067] c. For infrared candidate signals, near-infrared light (the second signal) generated by linear transformation (calculated transformation relationship) of red candidate signals can be selected as the target signal. Calculate the transformation relationship between near-infrared light and infrared candidate signals.

[0068] (2) Scenarios using the red light channel as the reference signal: When the red light signal is selected as the global reference signal, the procedure can be adjusted as follows: a. When selecting a red light reference signal as the target signal from among the infrared candidate signals, the target signal serves as the reference signal. Calculate the conversion relationship between the infrared candidate signal and the red light reference signal.

[0069] b. The orange light candidate signal selects the red light reference signal as the target signal. At this time, the target signal is the reference signal. Calculate the conversion relationship between the red light reference signal and the orange light candidate signal.

[0070] c. For the green light candidate signal with the largest wavelength difference, the estimated "near-orange light" signal (first signal) generated from the red light signal can be selected as the target signal. The conversion relationship between the "near-orange light" and the green light candidate signal is then calculated.

[0071] Step 105: Store the reference signal and the conversion relationship between each group of candidate signals and their corresponding target signal.

[0072] At this point, it is no longer necessary to store multiple sets of PPG signals, but only one set of reference signals plus the conversion relationship between each set of candidate signals and its corresponding target signal.

[0073] Taking the example of using "red light as a reference signal" as an example, we only need to store the following: Data 1: A complete red PPG signal (reference signal).

[0074] Data 2. The conversion relationships of the other three sets of signals are: "conversion relationship of infrared light relative to red light", "conversion relationship of orange light relative to red light", and "conversion relationship of green light relative to 'near orange light generated by red light'".

[0075] In other words, this solution does not store the original infrared, orange, and green light signals, but it can perfectly reconstruct all signals when needed by using reference signals and corresponding conversion relationships, thus greatly saving storage space.

[0076] This application provides a method, apparatus, device, and medium for data storage and reconstruction. After obtaining multiple sets of PPG signals, it first determines the set of reference signals with the best signal quality, then constructs a complete signal baseline using this set of reference signals. The remaining candidate signals are efficiently represented by establishing conversion relationships with the reference signals. Specifically, this solution transforms the stored content from multiple complete sets of PPG signals into a lightweight architecture of "one complete reference signal + multiple sets of conversion relationships." When it is necessary to restore the original data, the system can accurately reconstruct each candidate signal using the reference signal and the corresponding conversion relationships. This design achieves a qualitative leap in storage efficiency while ensuring the integrity of key signal information.

[0077] In an alternative implementation, the multiple sets of PPG signals are acquired by at least one acquisition device at different acquisition locations and / or based on different colors of light.

[0078] Storing the reference signal includes: The core frequency components of each group of PPG signals are determined to obtain target frequency bands for multiple groups of PPG signals; each target frequency band includes each core frequency component; the reference signal is bandpass filtered based on the set of core frequency components; and the filtered reference signal is stored.

[0079] In this step, to further reduce the amount of data that needs to be stored, the reference signal to be stored needs to be further filtered. That is, instead of storing the complete reference signal, only its essential part is stored. The specific method is as follows: first, analyze all PPG signals to find the common set of the core pulse frequency range of each group of PPG signals to obtain the target frequency band. Then, perform bandpass filtering on the reference signal, retaining only the waveform within this frequency band. Finally, store this filtered and simplified signal.

[0080] For example, if analysis reveals that the effective pulse components of all signals fall within the 0.8-2.5Hz range, the system will use this frequency band to filter the reference signal, eliminating low-frequency drift and high-frequency noise. Ultimately, the stored waveforms are those falling within the 0.8-2.5Hz range of the reference signal, which retain all the key features required for reconstruction while further compressing the data volume.

[0081] For each group of PPG signals, the core frequency components can be determined in the following way: A three-stage filtering approach is used to precisely pinpoint the core frequency components of the PPG signal. First, leveraging the characteristics of the PPG physiological signal (effective pulse frequency is typically much lower than 10Hz), a 10Hz high-frequency cutoff threshold is set to directly eliminate high-frequency noise. Next, energy filtering is performed in the frequency domain, retaining only significant frequency components with power exceeding 5% of the total energy to filter out weak background interference. Finally, to ensure the integrity of core physiological information, the remaining components are accumulated from high to low until they cover 98% of the total energy, thus preserving all core frequency components characterizing heart rate and its harmonics. This combined strategy achieves an optimal balance between noise reduction and fidelity preservation.

[0082] For each selected core frequency component, record its index information so that these key core frequency components can be accurately located during subsequent reconstruction, thereby reconstructing a better candidate signal.

[0083] In an optional implementation, the method further includes: For each group of PPG signals, it is divided into multiple windows; any two windows have at least partial overlap; for each window, the data in the window is subjected to Fourier transform to obtain a frequency domain signal; based on the frequency domain signal, each group of PPG signals is compressed to obtain multiple compressed PPG signals.

[0084] For each PPG signal group, consecutive PPG signals are divided into sliding, partially overlapping time windows. This facilitates local spectral analysis (because the frequency characteristics of the signal may change over time) and lays the foundation for real-time, streaming compression.

[0085] After obtaining multiple windows, the time-domain signal within each window is transformed from the time domain to the frequency domain by performing a Fast Fourier Transform. This transformation concentrates the signal energy onto a few core frequency components, creating conditions for efficient compression.

[0086] Next, lossy compression is performed on the frequency coefficients of the frequency domain signal obtained from the Fourier transform. For example, the frequency domain signal can be compressed by the same amplitude.

[0087] Alternatively, different frequency domain coefficients can be compressed by different magnitudes.

[0088] For example, for each group of PPG signals in the frequency domain, a scaling ratio is dynamically assigned to it based on the amplitude of each frequency domain coefficient; the scaling ratio is proportional to the amplitude; and the frequency domain signal is non-uniformly scaled based on the scaling factor corresponding to each frequency domain coefficient.

[0089] In other words, a scaling factor proportional to the amplitude of each frequency domain coefficient is dynamically assigned. For high-amplitude coefficients (such as the DC component that determines the signal baseline and the core frequency that characterizes the main pulse), a larger scaling factor is assigned, which is equivalent to using a coarser quantization step size for processing, thus achieving stronger compression. For low-amplitude coefficients, a smaller scaling factor (i.e., a finer quantization step size) is assigned, thus preserving more precision during compression. This strategy of "coarse quantization for high amplitude and fine quantization for low amplitude" essentially optimizes the allocation of limited data storage resources. It ensures that the main energy components of the signal are efficiently compressed with acceptable distortion, while preserving the low-amplitude components containing detailed information, thereby maximizing the quality and usability of the reconstructed signal even with an overall high compression ratio.

[0090] At this point, step 102 distinguishes multiple sets of PPG signals into a reference signal and at least one set of candidate signals, including: The compressed PPG signals are divided into a reference signal and at least one candidate signal.

[0091] This reduces the overall storage and transmission burden of multiple PPG signals.

[0092] In an optional implementation, the method further includes: Calculate the residual between the target signal and the candidate signal; determine the degree of compression of the residual based on the signal quality of the candidate signal; the degree of compression is inversely proportional to the signal quality; compress the residual of the candidate signal based on the degree of compression; and store the compressed residual.

[0093] The residual can be understood as the difference between the "candidate signal" and the "estimated value of the target signal after transformation," as shown in the following formula: Candidate signal = (target signal × transformation relationship) + residual The smaller the residual, the more accurate the transformation relationship and the more similar the candidate signal is to the target signal.

[0094] Assuming the target signal has a value of 100 at a certain moment, and the transformation relationship is (a=0.9, b=5), then the estimated value is 0.9*100+5=95. If the true value of the candidate signal at this moment is 97, then the residual is 97-95=2. During reconstruction, we perfectly recover the original value using 95+2=97. During storage, we only need to store the transformation relationship (0.9, 5) and the residual 2, without needing to store the complete candidate signal.

[0095] Since low-quality signals contain a large amount of noise, a significant portion of their residuals is also noise, making it unworthy of high-precision storage. Therefore, setting the compression level inversely proportional to the signal quality allows for more aggressive compression of the residuals of low-quality signals.

[0096] For example, one calculated residual is 22.214.

[0097] If the candidate signal quality is high, it is stored as 22. If the candidate signal quality is low, coarser quantization is performed, and it is directly stored as 20.

[0098] Based on the same technical concept, embodiments of this application also provide a method for data reconstruction. For example... Figure 2 As shown, it includes the following steps: Step 201: Obtain the signal to be processed; the signal to be processed is stored based on any of the aforementioned data storage methods; the signal to be processed includes at least: a reference signal, and a conversion relationship between a target signal and a candidate signal, wherein the target signal is the reference signal, or a candidate signal whose conversion relationship has been calculated.

[0099] Step 202: Based on the target signal and the conversion relationship between the target signal and the candidate signal to be reconstructed, reconstruct each group of candidate signals.

[0100] The steps 201-202 described above are actually the reverse process of the aforementioned data storage method. The core objective is to use the stored compressed data (reference signal, conversion relationship, etc.) to reconstruct the original waveforms of all candidate signals with high fidelity.

[0101] The data reconstruction process is illustrated using a preferred embodiment: Read the selected frequency index: First, determine which frequency components need to be reconstructed. During compression, only the core frequency components may have been saved to remove noise. In this case, the system reads the indexes of these saved frequency points, focuses on the core frequency components, and ignores irrelevant or noisy frequency bands, laying the foundation for accurate reconstruction.

[0102] Load stored spectral coefficients: Read compressed frequency domain data for these key frequency points, mainly including amplitude and phase information. This is the "raw material" for reconstructing the signal.

[0103] Zero-padding is used to generate a complete fundamental spectrum: Since only some key frequency points were retained during storage, zero-padding is now needed to fill in the discarded frequency points, thereby reconstructing a complete spectrum. This step prepares for the subsequent inverse Fourier transform, ensuring that it can be correctly converted back to the time domain signal.

[0104] The key to achieving efficient compression is reconstructing the spectra of other channels layer by layer using stored linear transformation parameters and residuals. Reconstruction is not performed independently, but follows a dependency chain established during storage: starting with the complete spectrum of the reference signal. The stored transformation relationships (such as parameters of the linear model) are applied to convert the reference signal spectrum into an estimated spectrum of the first candidate signal. The compressed residual stored for this candidate signal is then loaded and added to the estimated spectrum, thus accurately reconstructing the complete spectrum of the first candidate signal. Using this newly reconstructed candidate signal as the new "reference," this process is repeated layer by layer until the complete spectra of all channels are reconstructed. This significantly improves the overall compression efficiency.

[0105] Obtain and apply the scaling factor to inversely scale the spectrum: reverse the normalization or quantization operations performed during compression to save storage space, restore the spectrum values ​​to their original magnitude, and correct the amplitude scaling caused by compression.

[0106] Perform inverse Fourier transform to restore the time-domain signal: For each complete spectrum reconstructed in the above steps, convert it from the frequency domain back to the time domain through inverse fast Fourier transform to obtain the reconstructed signal of each time segment.

[0107] Discard overlapping portions and stitch windows: Reverse the "overlapping segmentation" operation used during compression. Discard the overlapping portions at the beginning and end of each time segment, retain only the most effective portions in the middle, and then smoothly stitch these effective portions together to ultimately recover the original waveforms of all PPG channels completely and continuously.

[0108] In this way, the original signals of the stored multiple sets of PPG signals can be reconstructed.

[0109] Based on the same technical concept, embodiments of this application also provide a data storage device, such as... Figure 3As shown, the device includes: The acquisition module 301 is used to acquire multiple sets of photoplethysmography (PPG) signals; at least two sets of PPG signals have some differences between them.

[0110] The differentiation module 302 is used to differentiate multiple PPG signals into a reference signal and at least one candidate signal; the signal quality of the reference signal is higher than that of the candidate signal.

[0111] The determining module 303 is used to determine the target signal corresponding to the candidate signal from the plurality of PPG signals for each candidate signal; the similarity between the target signal and the candidate signal meets a preset requirement, and the target signal is the reference signal or a candidate signal whose conversion relationship has been calculated.

[0112] The calculation module 304 is used to calculate the conversion relationship between the target signal and the candidate signal; the conversion relationship is used to reconstruct the candidate signal based on the target signal.

[0113] The storage module 305 is used to store the reference signal and the conversion relationship between each group of candidate signals and their corresponding target signals.

[0114] In one feasible implementation, the multiple sets of PPG signals are acquired by at least one acquisition device at different acquisition locations, and / or based on light of different colors.

[0115] The storage module is used to store the reference signal, for the following purposes: The core frequency component of each group of PPG signals is determined to obtain target frequency bands for multiple groups of PPG signals; the target frequency bands include each of the core frequency components.

[0116] The reference signal is bandpass filtered based on the core frequency component set.

[0117] Store the filtered reference signal.

[0118] In one feasible implementation, the device further includes: A window module is used to divide each group of PPG signals into multiple windows; any two windows have at least partial overlap.

[0119] The frequency domain transformation module is used to perform Fourier transform on the data in each window to obtain the frequency domain signal.

[0120] The compression module is used to compress each group of PPG signals based on the frequency domain signal to obtain multiple compressed PPG signals.

[0121] The differentiation module is used to distinguish multiple sets of PPG signals into a reference signal and at least one set of candidate signals, for the following purposes: The compressed PPG signals are divided into a reference signal and at least one candidate signal.

[0122] In one feasible implementation, the compression module is used to compress each group of PPG signals based on the frequency domain signal, for the following purposes: For each group of PPG signals in the frequency domain, a scaling factor is dynamically assigned to it based on the amplitude of each frequency domain coefficient; the scaling factor is proportional to the amplitude.

[0123] The frequency domain signal is non-uniformly scaled based on the scaling factor corresponding to each frequency domain coefficient.

[0124] In one feasible implementation, the calculation module is used to calculate the transformation relationship between the target signal and the candidate signal, for the following purposes: For each group of candidate signals, when the target signal is a reference signal, the transformation relationship between the reference signal and the group of candidate signals is calculated based on a preset algorithm; the preset algorithm includes: a generalized linear model and / or cross-correlation analysis.

[0125] The first signal obtained based on the transformation relationship and the reference signal is determined as a candidate signal for which the transformation relationship has been calculated.

[0126] Alternatively, for each group of candidate signals, when the target signal is the first signal, the conversion relationship between the first signal and the group of candidate signals is calculated based on the preset algorithm.

[0127] The second signal, obtained based on the transformation relationship and the first signal, is determined as a candidate signal for which the transformation relationship has been calculated.

[0128] Alternatively, for each group of candidate signals, when the target signal is the second signal, the conversion relationship between the second signal and the group of candidate signals is calculated based on the preset algorithm.

[0129] In one feasible implementation, the computing module is further used for: Calculate the residual between the target signal and the candidate signal.

[0130] Also includes: The first module is used to determine the degree of compression of the residual based on the signal quality of the candidate signal; the degree of compression is inversely proportional to the signal quality.

[0131] The second module is used to compress the residual of the candidate signal based on the compression degree.

[0132] The compression module is also used to store the compressed residual.

[0133] In one feasible implementation, embodiments of this application also provide a data reconstruction apparatus, such as... Figure 4 As shown, the device includes: The signal acquisition module 401 is used to acquire a signal to be processed; the signal to be processed is stored based on any of the methods described in the first aspect above; the signal to be processed includes at least: a reference signal, and a conversion relationship between a target signal and a candidate signal, wherein the target signal is the reference signal, or a candidate signal whose conversion relationship has been calculated.

[0134] The reconstruction module 402 is used to reconstruct each group of candidate signals based on the target signal and the conversion relationship between the target signal and the candidate signals to be reconstructed.

[0135] Figure 5 A schematic diagram of an electronic device provided in this application embodiment includes: a processor 501, a storage medium 502, and a bus 503. The storage medium 502 stores machine-readable instructions executable by the processor 501. When the electronic device runs the method as described in the embodiment, the processor 501 communicates with the storage medium 502 through the bus 503, and the processor 501 executes the machine-readable instructions to perform the steps as described in the embodiment.

[0136] In this embodiment, the storage medium 502 may also execute other machine-readable instructions to perform other methods as described in the embodiment. For details on the specific execution steps and principles, please refer to the description of the embodiment, which will not be repeated here.

[0137] This application also provides a computer-readable storage medium storing a computer program that is executed by a processor to perform the steps as described in the embodiments.

[0138] In this embodiment, the computer program, when run by the processor, can also execute other machine-readable instructions to perform other methods as described in the embodiments. For details on the specific execution steps and principles, please refer to the description of the embodiments, which will not be repeated here.

[0139] 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. The apparatus embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interface; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0140] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0141] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0142] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, 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, ROM, RAM, magnetic disks, or optical disks.

[0143] The above are merely specific embodiments 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.

Claims

1. A method for data storage, characterized in that, The method includes: Acquire multiple sets of photoplethysmography (PPG) signals; at least two sets of PPG signals show some differences. Multiple PPG signals are divided into a reference signal and at least one candidate signal; the signal quality of the reference signal is higher than that of the candidate signal. For each set of candidate signals, a target signal corresponding to the candidate signal is determined from the multiple sets of PPG signals; the similarity between the target signal and the candidate signal meets a preset requirement, and the target signal is the reference signal or a candidate signal whose transformation relationship has been calculated; Calculate the conversion relationship between the target signal and the candidate signal; the conversion relationship is used to reconstruct the candidate signal based on the target signal. Store the reference signal, and the conversion relationship between each group of candidate signals and their corresponding target signal; Calculating the transformation relationship between the target signal and the candidate signal includes: For each group of candidate signals, when the target signal is a reference signal, the conversion relationship between the reference signal and the group of candidate signals is calculated based on a preset algorithm; The first signal obtained based on the transformation relationship and the reference signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the first signal, the conversion relationship between the first signal and the group of candidate signals is calculated based on the preset algorithm; The second signal obtained based on the transformation relationship and the first signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the second signal, the conversion relationship between the second signal and the group of candidate signals is calculated based on the preset algorithm.

2. The method according to claim 1, characterized in that, The multiple sets of PPG signals are obtained by at least one acquisition device at different acquisition locations, and / or based on light of different colors; Storing the reference signal includes: The core frequency component of each group of PPG signals is determined to obtain target frequency bands for multiple groups of PPG signals; each target frequency band includes each core frequency component. The reference signal is bandpass filtered based on the core frequency component set; Store the filtered reference signal.

3. The method according to claim 1, characterized in that, The method further includes: For each group of PPG signals, it is divided into multiple windows; any two windows have at least partial overlap. For each window, perform a Fourier transform on the data in the window to obtain the frequency domain signal; Based on the frequency domain signal, each group of PPG signals is compressed to obtain multiple compressed groups of PPG signals. Multiple PPG signals are divided into a reference signal and at least one candidate signal, including: The compressed PPG signals are divided into a reference signal and at least one candidate signal.

4. The method according to claim 3, characterized in that, Based on the frequency domain signal, each group of PPG signals is compressed, including: For each group of PPG signals in the frequency domain, a scaling factor is dynamically assigned to it based on the amplitude of each frequency domain coefficient; the scaling factor is proportional to the amplitude. The frequency domain signal is non-uniformly scaled based on the scaling factor corresponding to each frequency domain coefficient.

5. The method according to claim 1, characterized in that, The preset algorithm includes: a generalized linear model, and / or cross-correlation analysis.

6. The method according to claim 5, characterized in that, The method further includes: Calculate the residual between the target signal and the candidate signal; Based on the signal quality of the candidate signal, the degree of compression of the residual is determined; the degree of compression is inversely proportional to the signal quality. Based on the compression level, the residual of the candidate signal is compressed; Store the compressed residual.

7. A method for data reconstruction, characterized in that, The method includes: Acquire a signal to be processed; the signal to be processed is stored based on the method described in any one of claims 1-6; the signal to be processed includes at least: a reference signal, and a conversion relationship between a target signal and a candidate signal, wherein the target signal is the reference signal, or a candidate signal whose conversion relationship has been calculated; Based on the target signal and the conversion relationship between the target signal and the candidate signal to be reconstructed, each group of candidate signals is reconstructed.

8. A data storage device, characterized in that, The device includes: The acquisition module is used to acquire multiple sets of photoplethysmography (PPG) signals; at least two sets of PPG signals have some differences between them. A differentiation module is used to differentiate multiple sets of PPG signals into a reference signal and at least one set of candidate signals; the signal quality of the reference signal is higher than that of the candidate signal. The determination module is used to determine the target signal corresponding to the candidate signal from the plurality of PPG signals for each candidate signal; the similarity between the target signal and the candidate signal meets a preset requirement, and the target signal is the reference signal or a candidate signal whose transformation relationship has been calculated; The calculation module is used to calculate the conversion relationship between the target signal and the candidate signal; the conversion relationship is used to reconstruct the candidate signal based on the target signal; A storage module is used to store the reference signal, as well as the conversion relationship between each group of candidate signals and their corresponding target signal; The calculation module is used to calculate the transformation relationship between the target signal and the candidate signal, and is used for: For each group of candidate signals, when the target signal is a reference signal, the conversion relationship between the reference signal and the group of candidate signals is calculated based on a preset algorithm; The first signal obtained based on the transformation relationship and the reference signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the first signal, the conversion relationship between the first signal and the group of candidate signals is calculated based on the preset algorithm; The second signal obtained based on the transformation relationship and the first signal is determined as a candidate signal for which the transformation relationship has been calculated; Alternatively, for each group of candidate signals, when the target signal is the second signal, the conversion relationship between the second signal and the group of candidate signals is calculated based on the preset algorithm.

9. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is in operation, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the method as described in any one of claims 1 to 7.

Images