Intelligent heart rate detection mattress based on BCG signal and GAN model

Abstract

The invention relates to the technical field of intelligent medical monitoring and signal processing, in particular to an intelligent heart rate detection mattress based on a BCG signal and a GAN model, and a piezoelectric sensor array, a signal preprocessing module, a signal optimization module, a calculation module and a data storage module are arranged in a body of the intelligent heart rate detection mattress; the piezoelectric sensor array is used for collecting a BCG signal of a user in a specified time period when the user lies on the intelligent mattress; the signal preprocessing module is used for preprocessing the collected BCG signals to obtain corresponding preprocessed signals; the signal optimization module is used for optimizing the pre-processed signal by using a pre-trained GAN model to obtain an optimized BCG signal; and the heart rate calculation module is used for calculating an optimized user heart rate value based on the optimized BCG signal, and storing the optimized user heart rate value to the data storage module.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent medical monitoring and signal processing, and in particular to an intelligent heart rate detection mattress based on BCG signals and a GAN model. Background Art

[0002] Heart rate is one of the important physiological indicators of human health, and can reflect an individual's cardiovascular health, emotional state, and sleep quality. Currently, heart rate detection methods mainly include photoplethysmography (PPG), electrocardiogram (ECG), and body movement electrocardiogram (BCG)-based detection methods. Among them, PPG and ECG detection require wearing devices, such as smart bracelets, smart watches, or ECG patches, which may have a certain impact on the user's comfort, and long-term wearing may cause skin discomfort. In contrast, the heart rate detection method based on BCG signals does not require wearing a device, and can obtain heart rate data through non-contact measurement. It is highly user-friendly and has therefore attracted widespread attention.

[0003] The BCG signal is a vibration signal from the mattress or seat surface caused by cardiac contraction, blood circulation, and subtle body movements. This signal is typically acquired by piezoelectric sensors, strain sensors, or accelerometers. However, BCG signals are weak and susceptible to interference from environmental noise, body motion artifacts, and other factors, resulting in reduced detection accuracy. Existing BCG signal processing methods typically use filtering and signal decomposition techniques, such as wavelet transforms and adaptive filtering, to improve signal quality. However, while these methods remove noise and enhance the signal, they may also result in information loss or increased artifacts, affecting the accuracy of the final heart rate calculation. Summary of the Invention

[0004] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides an intelligent heart rate detection mattress based on BCG signals and GAN models.

[0005] In order to achieve the above objectives, the main technical solutions adopted by the present invention include:

[0006] An embodiment of the present invention provides an intelligent heart rate detection mattress based on BCG signals and a GAN model. The body of the intelligent heart rate detection mattress is provided with a piezoelectric sensor array, a signal preprocessing module, a signal optimization module, a heart rate calculation module, and a data storage module.

[0007] The piezoelectric sensor array is used to collect the BCG signal of the user during a specified time period when the user lies on the smart mattress;

[0008] A signal preprocessing module is used to preprocess the collected BCG signal to obtain a corresponding preprocessed signal;

[0009] The signal optimization module is used to optimize the preprocessed signal using the pre-trained GAN model to obtain the optimized BCG signal;

[0010] The heart rate calculation module calculates the optimized user heart rate value based on the optimized BCG signal and stores the optimized user heart rate value in the data storage module.

[0011] Preferably, the signal preprocessing module preprocesses the collected BCG signal by: filtering;

[0012] The signal preprocessing module uses formula (1) to filter the collected BCG signal; the formula (1) is:

[0013]

[0014] S raw (t) BCG signal of the user at time t;

[0015] S pre (t) is the BCG signal S of the user at time t raw (t) the corresponding pre-processed signal;

[0016] Wherein, γ is a preset nonlinear smoothing factor used to control the smoothness of the signal;

[0017] μ is the preset dynamic baseline offset used to remove low-frequency drift.

[0018] Preferably, the GAN model includes:

[0019] Discriminator, used to calculate the BCG signal S of the user at time t using formula (2) raw (t) corresponds to the pre-processed signal S pre (t) characteristic consistency index L auth (t), and remove the feature consistency index L auth (t) The preprocessed signal is lower than the preset threshold, and a high-confidence signal is obtained;

[0020] The formula (2) is:

[0021]

[0022] S ref (t) is a predefined standard BCG signal template at time t used for comparison and evaluation of actual signal quality;

[0023] is the BCG signal S of the user at time t raw (t) the first derivative of the corresponding preprocessed signal;

[0024] α is a pre-set nonlinear penalty factor;

[0025] Generator, used to perform data enhancement on the filtered high-confidence signals using formula (3) and generate optimized BCG signals;

[0026] The formula (3) is:

[0027]

[0028] represents the nonlinear transformation of the generator;

[0029] S true (t) is the BCG signal S of the user at time t after screening raw (t) the corresponding high-confidence signal;

[0030] For the generator The BCG signal S of the user at time t after the input is filtered raw (t) corresponds to the high-confidence signal S true (t), the enhanced signal generated;

[0031] ε is the signal adaptive adjustment parameter;

[0032] S GAN (t) is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) Corresponding optimized BCG signal;

[0033] N(0,σ 2 ) has zero mean and variance σ 2 Gaussian noise.

[0034] Preferably, where σ 2 Calculated by formula (4);

[0035] The formula (4) is:

[0036]

[0037] is the pre-acquired initial noise variance;

[0038] τ is the adjustment parameter for controlling the noise variation amplitude;

[0039] P signal (t) is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true(t) the instantaneous power;

[0040] P noise (t) is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) is the noise power.

[0041] Preferably, the initial noise variance Determined by formula (5);

[0042] The formula (5) is:

[0043]

[0044] is the average value of the instantaneous power of all high-confidence signals;

[0045] k is a pre-set empirical coefficient.

[0046] Preferably, the heart rate calculation module calculates the optimized user heart rate value based on the optimized BCG signal, specifically including:

[0047] The heart rate calculation module calculates the corresponding initial heart rate value based on all optimized BCG signals;

[0048] The heart rate calculation module optimizes the initial heart rate value to obtain the optimized user heart rate value.

[0049] Preferably, the heart rate calculation module calculates the corresponding initial heart rate value based on all optimized BCG signals using formula (6);

[0050] The formula (6) is:

[0051]

[0052] HR cale is the initial heart rate value;

[0053] t j is the time of the jth heart beat detected from all optimized BCG signals;

[0054] t j-1 is the time of the j-1th heart beat detected from all optimized BCG signals;

[0055] β j For t j The corresponding dynamic weight factor.

[0056] Preferably, the heart rate calculation module optimizes the initial heart rate value using formula (7) to obtain an optimized user heart rate value;

[0057] The formula (7) is:

[0058]

[0059] HR final To optimize the user's heart rate value;

[0060] ω is the adjustment coefficient;

[0061] is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) The corresponding optimized BCG signal S GAN The second derivative of (t).

[0062] Preferably, the GAN model is trained using a training dataset containing real BCG signals, wherein the training dataset includes BCG signals of healthy users and users with abnormal heart rates, external interference signals, and synchronized electrocardiogram (ECG) data as reference labels. During the training process, the generator and discriminator are optimized using an adversarial training method so that the optimized BCG signals are close to the real signals in the time and frequency domains.

[0063] Training termination conditions include signal quality convergence, heart rate calculation error decreasing to a preset threshold, and training loss being stable without mode collapse.

[0064] Preferably, the smart heart rate detection mattress further comprises: a remote transmission module;

[0065] The remote transmission module is used to communicate with the mobile terminal and transmit the optimized user heart rate value stored in the data storage module to the mobile terminal.

[0066] The beneficial effects of the present invention are:

[0067] The intelligent heart rate detection mattress based on BCG signals and GAN models of the present invention adopts a collaborative architecture of a piezoelectric sensor array, a signal preprocessing module, a GAN signal optimization module, and a heart rate calculation module. Compared with the single sensor or unoptimized BCG detection solutions in the prior art, it can realize the full process automation from signal acquisition to heart rate calculation, achieving the effect of non-contact, high-precision heart rate monitoring.

[0068] The present invention discloses an intelligent heart rate detection mattress based on BCG signals and a GAN model. By preprocessing the original BCG signal using a filtering formula with a nonlinear smoothing factor and a dynamic baseline offset, compared to traditional fixed threshold filtering, it can adaptively eliminate low-frequency drift and high-frequency noise, thereby achieving the effect of retaining effective heart beat characteristics while reducing signal distortion.

[0069] The present invention discloses an intelligent heart rate detection mattress based on BCG signals and GAN models. Since a discriminator is used to calculate feature consistency indicators and a generator is used for data enhancement, compared with directly using the original signal to calculate the heart rate, it can eliminate low-credibility signals and generate optimized BCG signals, thereby achieving the effect of stably extracting heart rate features in weak signal or interference environments.

[0070] The present invention discloses an intelligent heart rate detection mattress based on BCG signals and a GAN model. Due to the use of dynamic noise variance calculation and initial noise variance optimization, compared to the existing technology with a fixed noise model, it can adaptively adjust the noise suppression intensity according to the instantaneous power of the signal, thereby maintaining signal purity when the user has different body shapes or the mattress pressure is unevenly distributed. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 This is a schematic diagram of the internal module connections of an intelligent heart rate detection mattress based on BCG signals and GAN models of the present invention;

[0072] Figure 2 This is a flow chart of an execution method of an intelligent heart rate detection mattress based on BCG signals and a GAN model in the second embodiment. DETAILED DESCRIPTION

[0073] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation methods in conjunction with the accompanying drawings.

[0074] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. Instead, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0075] Example 1

[0076] See also Figure 1This embodiment provides an intelligent heart rate detection mattress based on BCG signals and GAN models. The body of the intelligent heart rate detection mattress is provided with a piezoelectric sensor array, a signal preprocessing module, a signal optimization module, a heart rate calculation module, and a data storage module;

[0077] The piezoelectric sensor array is used to collect the BCG signal of the user during a specified time period when the user lies on the smart mattress;

[0078] Specifically, in actual applications, the piezoelectric sensor array is evenly embedded in the upper layer of the mattress in the form of a 5×5 matrix with a spacing of 10 cm. It can collect BCG signals in the chest and abdominal areas when the user lies on the smart mattress. Compared with the single-point sensor or sparse array in the existing technology, it can significantly improve the spatial resolution and anti-interference ability of the BCG signal, achieving the effect of stably capturing the heart beat signal when the user turns over or the local pressure changes.

[0079] A signal preprocessing module is used to preprocess the collected BCG signal to obtain a corresponding preprocessed signal;

[0080] In this embodiment, the signal preprocessing module preprocesses the collected BCG signal by: filtering;

[0081] The signal preprocessing module uses formula (1) to filter the collected BCG signal; the formula (1) is:

[0082]

[0083] S raw (t) The BCG signal of the user at time t; S pre (t) is the BCG signal S of the user at time t raw (t) corresponds to the preprocessed signal; where γ is a preset nonlinear smoothing factor used to control the degree of signal smoothing. It can dynamically control the smoothness of the BCG signal, retaining more effective information when noise interference is low, and performing stronger smoothing when noise is high, thereby optimizing signal quality. μ is a preset dynamic baseline offset used to remove low-frequency drift.

[0084] The signal optimization module is used to optimize the preprocessed signal using the pre-trained GAN model to obtain the optimized BCG signal;

[0085] The GAN model includes:

[0086] Discriminator, used to calculate the BCG signal S of the user at time t using formula (2) raw (t) corresponds to the pre-processed signal S pre (t) characteristic consistency index Lauth (t), and remove the feature consistency index L auth (t) The preprocessed signal is lower than the preset threshold, and a high confidence signal is obtained;

[0087] The formula (2) is:

[0088]

[0089] S ref (t) is a predefined standard BCG signal template at time t used for comparison and evaluation of actual signal quality; is the BCG signal S of the user at time t raw (t) is the first-order derivative of the corresponding preprocessed signal; α is a pre-set nonlinear penalty factor;

[0090] Feature consistency index L auth (t) measures the difference in derivatives between the optimized and reference signals. This metric considers not only the signal amplitude but also its dynamic characteristics. This ensures that the optimized BCG signal matches the reference signal in overall morphology, not just numerical similarity.

[0091] Compared to traditional filtering methods, signal optimization using GANs can learn and generate optimization results that are closer to the true BCG signal, reducing information loss and improving signal quality. Through continuous training, GANs can better adapt to the characteristics of BCG signals from different users, improving their applicability. α controls the degree of amplification of consistency errors, making the optimization process more stable and preventing GAN overfitting or signal distortion.

[0092] The generator is used to perform data enhancement on the filtered high-confidence signals using formula (3) and generate an optimized BCG signal; the formula (3) is:

[0093]

[0094] represents the nonlinear transformation of the generator; S true (t) is the BCG signal S of the user at time t after screening raw (t) the corresponding high-confidence signal; For the generator The BCG signal S of the user at time t after the input is filtered raw (t) corresponds to the high-confidence signal S true (t), the enhanced signal generated; ε is the signal adaptive adjustment parameter; S GAN (t) is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal Strue (t) corresponds to the optimized BCG signal; N(0, σ 2 ) has zero mean and variance σ 2 Gaussian noise.

[0095] Generator By learning the characteristics of high-confidence BCG signals, we can generate optimization results that are closer to the real BCG signals and reduce measurement errors. 2 ) prevents GAN overfitting while improving signal robustness, ensuring that the optimized BCG signal maintains high credibility under different environments. The signal adaptively adjusts the parameter ε to control the impact of noise on the signal, ensuring that the GAN increases data diversity during data augmentation without excessively affecting signal quality, thereby maintaining signal stability and reliability.

[0096] Among them, σ 2 Calculated by formula (4); the formula (4) is:

[0097]

[0098] is the initial noise variance obtained in advance; τ is the adjustment parameter that controls the noise variation amplitude;

[0099] P signal (t) is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) instantaneous power; P noise (t) is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) is the noise power.

[0100] When the instantaneous power of the signal P signal (t) is significantly higher than the noise power P noise (t) (high signal-to-noise ratio), σ 2 Exponential decay reduces the noise suppression strength and avoids feature loss caused by excessive smoothing. When the noise power P noise When (t) is dominant (low signal-to-noise ratio), σ 2 Approaching the initial variance Enhanced noise suppression protects valid signals. Compared to existing solutions that use a fixed noise model, this system dynamically adjusts noise suppression strength based on the real-time ratio of the instantaneous signal power to the noise power, maintaining the purity of the heart rate characteristic signal even in complex scenarios such as user movement and breathing interference.

[0101] Among them, the initial noise variance Determined by formula (5); the formula (5) is:

[0102]

[0103] is the average value of the instantaneous power of all high-confidence signals; k is a preset empirical coefficient.

[0104] In this embodiment, a smart heart rate detection mattress based on BCG signal and GAN model adopts the average value of the instantaneous power of high-reliability signal. Combined with the empirical coefficient k to determine the initial noise variance Compared with traditional solutions with fixed noise variance or manually preset values, it can dynamically adapt to the signal strength differences between different users, achieving the effect of adaptively initializing noise suppression parameters in personalized usage scenarios, significantly improving the stability of subsequent GAN signal optimization.

[0105] The heart rate calculation module calculates the optimized user heart rate value based on the optimized BCG signal and stores the optimized user heart rate value in the data storage module.

[0106] In this embodiment, the heart rate calculation module calculates the optimized user heart rate value based on the optimized BCG signal, specifically including:

[0107] The heart rate calculation module calculates the corresponding initial heart rate value based on all optimized BCG signals;

[0108] Specifically, the heart rate calculation module uses formula (6) to calculate the corresponding initial heart rate value based on all optimized BCG signals;

[0109] The formula (6) is:

[0110]

[0111] HR cale is the initial heart rate value; t j is the time of the jth heart beat detected from all optimized BCG signals; t j-1 is the time of the j-1th heart beat detected from all optimized BCG signals; β j For t j The corresponding dynamic weight factor.

[0112] In this embodiment, due to the use of the initial heart rate calculation method of dynamic weight factors and time interval weighted averaging (Formula 6), compared with the traditional fixed time window or simple peak detection scheme, it can dynamically adjust the weights of different heart beat intervals, effectively compensate for the calculation errors caused by local signal distortion or short-term interference, and achieve the effect of stably outputting high-precision heart rate values ​​under non-ideal signal conditions.

[0113] The BCG signal may cause inaccurate detection of some heart beats (such as missed detection or false detection) due to user body movement, breathing interference or sensor noise. j Assign a weight to each heartbeat interval (e.g., based on signal quality, amplitude, or signal-to-noise ratio) to reduce the impact of low-quality intervals on the overall heart rate calculation. For example, if the jth heartbeat signal quality is poor (e.g., low signal-to-noise ratio), then β j Decrease to reduce the contribution of this interval in the average calculation.

[0114] The heart rate calculation module optimizes the initial heart rate value to obtain an optimized user heart rate value.

[0115] Specifically, the heart rate calculation module uses formula (7) to optimize the initial heart rate value to obtain the optimized user heart rate value;

[0116] The formula (7) is:

[0117]

[0118] HR final To optimize the user's heart rate value;

[0119] ω is the adjustment coefficient used to correct the amplitude, usually ranging from 0.01 to 0.05, to balance sensitivity and stability. is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) The corresponding optimized BCG signal S GAN The second derivative of (t). The local curvature change of the BCG signal is directly related to the dynamic characteristics of cardiac contraction / diastole (such as the acceleration of the aortic valve opening and closing). By analyzing the extreme points of curvature, the true heartbeat moment can be more accurately located, making up for the lack of sensitivity of the simple time interval detection (Formula 6) to morphological changes. When the absolute value of the second-order derivative is large (strong curvature change, such as a violent heartbeat), the correction term Significantly affects the final heart rate value and strengthens the response to abnormal beats; when the second-order derivative approaches zero (stationary signal), the result approaches the initial heart rate HR cale , avoid overcorrection.

[0120] Traditional peak detection is susceptible to misjudgment due to motion artifacts or breathing interference, while the second-order derivative has a natural suppression effect on high-frequency noise (derivative smoothing effect).

[0121] In the practical application of this embodiment, the GAN model is trained using a training dataset containing real BCG signals. The training dataset includes BCG signals of healthy users and users with abnormal heart rates, external interference signals, and synchronized electrocardiogram (ECG) data as reference labels. During the training process, the generator and discriminator are optimized using an adversarial training method, so that the optimized BCG signals are close to the real signals in the time and frequency domains.

[0122] Training termination conditions include signal quality convergence (the frequency domain correlation coefficient (0.5-5Hz band) between the generated signal and the real signal is ≥ 0.95), the heart rate calculation error is reduced to a preset threshold (on the validation set, the heart rate calculation error of the GAN-optimized signal is ≤ 1.5BPM), and the training loss is stable without mode collapse (the discriminator's accuracy on the generated signal is stable at 48%-52% (indicating generative diversity).

[0123] For example, the training data set includes: Healthy user data: BCG signals of 100 healthy adults (20-60 years old) are collected, and each person is continuously monitored for 30 minutes, covering states such as lying still, turning over, and deep breathing, and ECG is recorded simultaneously as the gold standard label. Abnormal heart rate data: Contains BCG signals of 50 patients with arrhythmia (atrial fibrillation, premature beats, etc.), and the hospital's professional equipment synchronously collects ECG to mark the abnormal heartbeat location. Interference signal: Common environmental noise (such as mattress friction, air conditioning vibration) and motion artifacts (simulating user turning over) are added, accounting for 20% to enhance robustness.

[0124] The smart heart rate detection mattress further includes: a remote transmission module; the remote transmission module is used to communicate with a mobile terminal and transmit the optimized user heart rate value stored in the data storage module to the mobile terminal.

[0125] For example, a patient with hypertension sleeps on a smart mattress. The mattress monitors heart rate in real time and transmits the data to their child's mobile app via a remote transmission module (e.g., Wi-Fi / Bluetooth). The heart rate calculation module generates optimized heart rate values ​​every five minutes and stores them in the data storage module. The remote transmission module pushes encrypted data to the home gateway on the user's nightstand via Bluetooth Low Energy (BLE 5.0), which then relays it to the child's mobile phone via the cloud.

[0126] Example 2

[0127] See also Figure 2 This embodiment provides a method for implementing a smart heart rate detection mattress based on BCG signals and a GAN model, including:

[0128] Step 1: After the user lies down, the piezoelectric sensor array in the smart heart rate detection mattress begins to collect BCG signals at a sampling rate of 100 Hz for 30 minutes.

[0129] The collected BCG signals include heart beats (0.5-5 Hz), respiratory interference (0.1-0.3 Hz) and turning noise (sudden high frequency).

[0130] Step 2: The signal preprocessing module in the intelligent heart rate detection mattress preprocesses the collected BCG signal using formula (1) to obtain a corresponding preprocessed signal;

[0131] The formula (1) is:

[0132]

[0133] S raw (t) The BCG signal of the user at time t; S pre (t) is the BCG signal S of the user at time t raw (t) corresponds to the preprocessed signal; where γ is a preset nonlinear smoothing factor used to control the degree of signal smoothing. It can dynamically control the smoothness of the BCG signal, retaining more effective information when noise interference is low, and performing stronger smoothing when noise is high, thereby optimizing signal quality. μ is a preset dynamic baseline offset used to remove low-frequency drift.

[0134] Step 3: The signal optimization module in the smart heart rate detection mattress uses the pre-trained GAN model to optimize the pre-processed signal to obtain the optimized BCG signal;

[0135] The discriminator of the GAN model was screened, and the feature consistency index was calculated. Low-quality signal segments with a feature consistency index less than 0.8 (such as the turning period) were eliminated. High-confidence signal segments (SNR>10dB) were retained, accounting for 85%.

[0136] The generator (U-Net structure) of the GAN model performs data enhancement on high-confidence signals and outputs optimized signals.

[0137] Step 4: The heart rate calculation module in the intelligent heart rate detection mattress detects the heart beat time point in the optimization signal and calculates the initial heart rate HR using formula (6) calc Then use formula (7) to correct the initial heart rate to obtain the optimized user heart rate value.

[0138] Step 5: The remote transmission module in the smart heart rate monitoring mattress transmits the optimized user's heart rate value to the home gateway via BLE every 5 minutes. The encrypted value is then uploaded to the cloud. The mobile app connected to the cloud receives the data and displays the corresponding heart rate trend graph.

[0139] In this embodiment, through GAN model optimization and a dynamic heart rate correction algorithm, medical-grade detection accuracy can be achieved even in complex scenarios such as respiratory interference and rolling noise. The discriminator in the GAN model removes low-quality signals, while the generator (U-Net) enhances heartbeat features. The optimized signal SNR reaches 22dB.

[0140] Example 3

[0141] This embodiment provides a smart heart rate detection mattress based on BCG signals and a GAN model, including:

[0142] The piezoelectric sensor array consists of multiple high-sensitivity piezoelectric sensors evenly distributed inside the mattress. It is used to collect the BCG signal generated by the user's heartbeat and convert it into an analog electrical signal. Through the signal conditioning circuit (amplification, filtering, A / D conversion), high-quality digital BCG signals are obtained.

[0143] The signal preprocessing module performs multi-stage filtering on the digitized BCG signal, including bandpass filtering (to remove noise) and wavelet transform denoising (to enhance the signal). Adaptive noise suppression technology is used to eliminate interference factors such as mattress vibration and respiratory movement.

[0144] The signal optimization module (based on the GAN model) includes a generator using a recurrent neural network (RNN) structure, including a long short-term memory (LSTM) or gated recurrent unit (GRU), to learn the time series characteristics of BCG signals and generate high-quality, denoised and enhanced BCG signals.

[0145] The discriminator uses an RNN structure to distinguish the authenticity of the input BCG signal, ensuring that the optimized signal is close to the real signal. Through iterative training, the signal optimization capability of the generator is continuously improved.

[0146] In this embodiment, the signal optimization module receives the data output by the signal preprocessing module and further enhances and optimizes the denoised BCG signal.

[0147] The signal optimization module provides high-quality signal input to the heart rate calculation module to ensure the accuracy of heart rate feature extraction.

[0148] Reduce redundant information in the data transmission module, optimize the quality of transmitted data, and improve the reliability of remote monitoring.

[0149] The heart rate calculation module uses FFT (Fast Fourier Transform) and time-domain analysis to extract heart rate-related features, including physiological parameters such as heart rate (HR) and heart rate variability (HRV). This is combined with anomaly detection algorithms (such as KNN, SVM, or deep learning) to identify abnormal events such as arrhythmia and cardiac arrest.

[0150] In addition, the heart rate calculation module in this embodiment also has an intelligent early warning function, sending an alert to the user or guardian when an abnormal heart rate is detected. Relying on the high-quality signal provided by the signal optimization module, the accuracy of the calculation results is ensured and the false alarm rate is reduced.

[0151] The data transmission module supports multiple wireless communication protocols, such as Wi-Fi, Bluetooth, and Zigbee, and is compatible with various smart devices. It can transmit monitoring data in real time to a mobile app or cloud server, allowing users to view historical data and health trends. The heart rate calculation module obtains final monitoring results and efficiently transmits them to remote devices.

[0152] This embodiment combines piezoelectric sensors, GAN signal optimization, wireless data transmission and other technologies to realize a high-precision, low-interference heart rate monitoring system.

[0153] The mattress in this embodiment utilizes a GAN model to generate high-quality BCG signals, effectively removing noise and improving the signal-to-noise ratio, making heart rate detection more accurate and reliable. This mattress is not only suitable for home environments but can also be widely used in hospitals, nursing homes, and other settings, demonstrating its high universality and flexibility. By monitoring users' health data over a long period of time, it can provide personalized health management recommendations for each user, helping them better understand and manage their health.

[0154] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0155] In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they may refer to fixed connection, detachable connection, or integration; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; and internal communication between two components or interaction between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0156] In the present invention, unless otherwise expressly specified or limited, when a first feature is "above" or "below" a second feature, it may mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Furthermore, when a first feature is "above," "above," or "above" a second feature, it may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is at a higher level than the second feature. When a first feature is "below," "below," or "below" a second feature, it may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is at a lower level than the second feature.

[0157] In the description of this specification, the terms "one embodiment", "some embodiments", "embodiments", "examples", "specific examples" or "some examples" refer to the specific features, structures, materials or characteristics described in conjunction with the embodiment or example and included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and features of different embodiments or examples, unless they are mutually inconsistent.

[0158] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may alter, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. An intelligent heart rate detection mattress based on BCG signal and GAN model, characterized by: The body of the intelligent heart rate detection mattress is provided with a piezoelectric sensor array, a signal preprocessing module, a signal optimization module, a heart rate calculation module and a data storage module; The piezoelectric sensor array is used to collect the BCG signal of the user during a specified time period when the user lies on the smart mattress; A signal preprocessing module is used to preprocess the collected BCG signal to obtain a corresponding preprocessed signal; The signal optimization module is used to optimize the preprocessed signal using the pre-trained GAN model to obtain the optimized BCG signal; The heart rate calculation module calculates the optimized user heart rate value based on the optimized BCG signal and stores the optimized user heart rate value in the data storage module.

2. The intelligent heart rate detection mattress based on BCG signal and GAN model according to claim 1 is characterized in that: The signal preprocessing module preprocesses the collected BCG signals as follows: filtering; The signal preprocessing module uses formula (1) to filter the collected BCG signal; the formula (1) is: S raw (t) BCG signal of the user at time t; S pre (t) is the BCG signal S of the user at time t raw (t) the corresponding pre-processed signal; Wherein, γ is a preset nonlinear smoothing factor used to control the smoothness of the signal; μ is the preset dynamic baseline offset used to remove low-frequency drift.

3. The intelligent heart rate detection mattress based on BCG signal and GAN model according to claim 1 is characterized in that: The GAN model includes: Discriminator, used to calculate the BCG signal S of the user at time t using formula (2) raw (t) corresponds to the pre-processed signal S pre (t) characteristic consistency index L auth (t), and remove the feature consistency index L auth (t) The preprocessed signal is lower than the preset threshold, and a high-confidence signal is obtained; The formula (2) is: S ref (t) is a predefined standard BCG signal template at time t used for comparison and evaluation of actual signal quality; is the BCG signal S of the user at time t raw (t) the first derivative of the corresponding preprocessed signal; α is a pre-set nonlinear penalty factor; Generator, used to perform data enhancement on the filtered high-confidence signals using formula (3) and generate optimized BCG signals; The formula (3) is: represents the nonlinear transformation of the generator; S true (t) is the BCG signal S of the user at time t after screening raw (t) the corresponding high-confidence signal; For the generator The BCG signal S of the user at time t after filtering the input raw (t) corresponds to the high-confidence signal S true (t), the enhanced signal generated; ε is the signal adaptive adjustment parameter; S GAN (t) is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) Corresponding optimized BCG signal; N(0,σ 2 ) has zero mean and variance σ 2 Gaussian noise.

4. The intelligent heart rate detection mattress based on BCG signal and GAN model according to claim 3 is characterized in that: in, σ 2 Calculated by formula (4); The formula (4) is: is the pre-acquired initial noise variance; τ is the adjustment parameter for controlling the noise variation amplitude; P signal (t) is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) the instantaneous power; P noise (t) is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) is the noise power.

5. The intelligent heart rate detection mattress based on BCG signal and GAN model according to claim 4 is characterized in that: in, Initial noise variance Determined by formula (5); The formula (5) is: is the average value of the instantaneous power of all high-confidence signals; k is a pre-set empirical coefficient.

6. The intelligent heart rate detection mattress based on BCG signal and GAN model according to claim 1, characterized in that: The heart rate calculation module calculates the optimized user heart rate value based on the optimized BCG signal, specifically including: The heart rate calculation module calculates the corresponding initial heart rate value based on all optimized BCG signals; The heart rate calculation module optimizes the initial heart rate value to obtain the optimized user heart rate value.

7. The intelligent heart rate detection mattress based on BCG signal and GAN model according to claim 6, characterized in that: The heart rate calculation module uses formula (6) to calculate the corresponding initial heart rate value based on all optimized BCG signals; The formula (6) is: HR cale is the initial heart rate value; t j is the time of the jth heart beat detected from all optimized BCG signals; t j-1 is the time of the j-1th heart beat detected from all optimized BCG signals; β j For t j The corresponding dynamic weight factor.

8. The intelligent heart rate detection mattress based on BCG signal and GAN model according to claim 7, characterized in that: The heart rate calculation module uses formula (7) to optimize the initial heart rate value to obtain the optimized user heart rate value; The formula (7) is: HR final To optimize the user's heart rate value; ω is the adjustment coefficient; is the BCG signal S of the user at time t after screening raw (t) corresponds to the high-confidence signal S true (t) The corresponding optimized BCG signal S GAN The second derivative of (t).

9. The intelligent heart rate detection mattress based on BCG signal and GAN model according to claim 1, characterized in that: The GAN model is trained using a training dataset containing real BCG signals, which includes BCG signals of healthy users and users with abnormal heart rates, external interference signals, and synchronized electrocardiogram (ECG) data as reference labels. During the training process, the adversarial training method is used to optimize the generator and discriminator, so that the optimized BCG signal is close to the real signal in the time and frequency domains; Training termination conditions include signal quality convergence, heart rate calculation error decreasing to a preset threshold, and training loss being stable without mode collapse.

10. The intelligent heart rate detection mattress based on BCG signal and GAN model according to claim 1, characterized in that: The smart heart rate detection mattress also includes: a remote transmission module; The remote transmission module is used to communicate with the mobile terminal and transmit the optimized user heart rate value stored in the data storage module to the mobile terminal.

Images