Non-contact heart rate and respiration recorder

Abstract

The present application relates to the field of respiratory heart rate monitoring, and discloses a non-contact heart rate and respiration recorder, which comprises a host computer, a wireless charging base, a charging cable and a terminal device, wherein the host computer is internally provided with a non-contact sensor module, a processor, a battery and a wireless communication module, and the shell adopts a waterproof sealing structure. In use, the host computer is placed under a pillow, and a heart beat and respiration channel signal is collected through the non-contact sensor, and after filtering, normalization and artifact removal, the signal forms a heart beat observation and a respiration observation under the action of parameters such as a sampling rate, a time window length, a vertical offset and a coupling stiffness, and is subjected to joint optimization under the constraints of spectral matching and cross-modal coherence, and finally outputs a heart rate and a respiration rate. The device does not need to wear an electrode, is stable in detection, and is suitable for long-term home health monitoring and clinical assistance.

Description

Technical Field

[0001] This invention relates to the field of respiratory and heart rate monitoring, specifically a non-contact heart rate and respiratory recorder. Background Technology

[0002] In recent years, non-contact heart rate and respiration monitoring technology has gradually become a research hotspot. Compared with traditional electrode patches or chest strap contact monitoring methods, non-contact monitoring can collect heart and respiratory signals using sensors under the pillow, mattress sensors, radar, acoustic sensors, etc., without increasing the burden of wearing them. Therefore, it is more suitable for long-term continuous home health management and sleep monitoring.

[0003] Existing non-contact heart rate and respiration monitoring devices typically employ single-channel detection, such as relying solely on the BCG channel or solely on the acoustic channel for signal acquisition and analysis. These methods often depend on fixed sampling rates and empirical filtering parameters, failing to account for the diversity of individuals and environments. In practical use, insufficient sampling rates reduce the time-frequency resolution of cardiac and respiratory signals, easily leading to inaccurate peak detection; excessively high sampling rates increase power consumption, affecting device battery life. Furthermore, fixed analysis windows cannot balance stability and sensitivity; short windows are susceptible to transient noise interference, while long windows result in response lag. These issues make it difficult for existing technologies to simultaneously achieve stability and real-time performance in different scenarios.

[0004] Existing pillow-based or non-contact monitoring solutions do not adequately consider the spatial positional differences between the user and the sensor. Due to variations in pillow thickness and head positioning, the vertical offset between the sensor and the head can range from a few millimeters to several centimeters. Traditional solutions often assume the sensor is in close contact or at a fixed distance, failing to model and compensate for this parameter. This results in significant signal amplitude attenuation and poor applicability across individuals or scenarios. Furthermore, different pillow materials and structures exhibit significant differences in acoustic or mechanical transmission characteristics. Existing methods typically do not introduce mechanical coupling parameters to correct signal drift, leading to non-stationary shifts in the respiratory wave baseline and unclear respiratory peaks, thus affecting the accuracy of respiratory rate estimation.

[0005] Most existing non-contact monitoring methods detect heart rate and respiratory rate independently, lacking cross-modal consistency constraints. This leads to frequent jumps or erroneous locking in single-channel results when subjected to body motion interference or signal degradation. In such cases, the lack of mutual verification mechanisms between heart rate and respiratory rate estimates makes it difficult to guarantee the stability of the overall output. Furthermore, most existing technologies employ simple thresholding or single spectral peak selection methods, failing to introduce a joint optimization objective function within a mathematical framework to simultaneously constrain the spectral matching and cross-modal coherence of heart rate and respiration. Consequently, they often fail to converge stably in complex environments, affecting the reliability of monitoring. Summary of the Invention

[0006] The purpose of this invention is to provide a non-contact heart rate and respiration recorder to solve the technical problems mentioned in the background section.

[0007] Based on the above ideas, the present invention provides the following technical solution:

[0008] A non-contact heart rate and respiration recorder, comprising:

[0009] The host computer has a built-in non-contact sensor module, microprocessor, memory and wireless communication module;

[0010] A wireless charging dock for wirelessly charging the main unit;

[0011] A power module, including a battery, provides power to the host unit;

[0012] The outer casing, with a waterproof and sealed structure, covers the main unit, achieving an IP68 protection rating.

[0013] The non-contact sensor module includes a low-frequency acoustic piezoelectric sensor and a BCG cardiac pulse detection unit, used to collect human heart rate and respiratory signals;

[0014] The microprocessor is configured with a time-frequency domain joint algorithm for feature extraction and data processing of the signals acquired by the sensor module.

[0015] The wireless communication module is a low-power Bluetooth module, used to transmit the processed monitoring data to an external terminal device;

[0016] The terminal device runs data management software to receive and display the heart rate and respiratory rate data.

[0017] Preferably, the non-contact sensor module is configured as follows:

[0018] S1. Acquire cardiac and respiratory channel signals at the suboccipital location using the sampling rate;

[0019] S2. Bandpass filtering and normalization are performed on the signal to obtain cardiac candidate components and respiratory candidate components;

[0020] S3. Perform artifact removal and robust segmentation within the time window length;

[0021] S4. Under the sampling rate, time window length and vertical offset parameters, the cardiac candidate component and the respiratory candidate component are correlated and fused to obtain the cardiac observation.

[0022] S5. Under the sampling rate, time window length and coupling stiffness parameters, the respiratory candidate components are stabilized and amplitude normalized to obtain respiratory observations;

[0023] S6. Using the cardiac and respiratory observations as inputs, perform joint optimization based on spectral matching and cross-modal coherence constraints to output heart rate and respiratory rate.

[0024] Preferably, step S4 includes: performing mean-reduction and normalization processing on the cardiac candidate component and the respiratory candidate component within the time window length, calculating the correlation using a sliding window, and finely locating the correlation peak.

[0025] Preferably, the cardiac observations satisfy the following relationship:

[0026]

[0027] τ∈[-τ0,τ0

[0028] in:

[0029] X(t) represents cardiac observations;

[0030] b bcg (t) represents the filtered and normalized result of the cardiac channel signal;

[0031] b ac (t) represents the filtered and normalized result of the respiratory channel signal;

[0032] τ is a delayed variable;

[0033] τ0 is the maximum allowable delay range;

[0034] Δz is the vertical offset parameter;

[0035] α is the attenuation coefficient;

[0036] <·,·> represent inner product operations;

[0037] ||·|| represents the signal energy norm.

[0038] Preferably, step S5 includes: performing envelope extraction on the respiratory candidate components, and performing low-pass filtering and amplitude normalization within the time window length to obtain respiratory observations.

[0039] Preferably, the respiratory observation Y satisfies the following relationship:

[0040]

[0041] in:

[0042] Y(t) represents the respiratory observation;

[0043] h lp (t;T w ) is based on the time length T w A low-pass filter with a scale of 1;

[0044] * indicates convolution operation;

[0045] H{·} represents the Hilbert transform;

[0046] k is the coupling stiffness parameter.

[0047] Preferably, step S6 includes: using the spectral peak of cardiac observations as a candidate heart rate and the spectral peak of respiratory observations as a candidate respiratory rate, performing iterative optimization under the condition of satisfying the cross-modal coherence threshold, and applying temporal smoothing constraints between continuous time windows.

[0048] Preferably, the optimization satisfies the following objective function:

[0049] J(HR,RR)=w1D(PSD(X),HR)+w2D(PSD(Y),RR)+w3(1-Coh(X,Y;RR))

[0050] in,

[0051] J(HR,RR) is the objective function;

[0052] w1, w2, and w3 are positive weighting coefficients;

[0053] D(·) is a measure of spectral difference;

[0054] PSD(·) represents the power spectral density;

[0055] X represents cardiac observations;

[0056] Y represents respiratory observations;

[0057] Coh(X,Y;RR) is the coherence calculated at the respiratory rate RR.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] This invention solves the problems of discomfort and signal interference caused by the skin electrode contact quality in existing contact monitoring devices by incorporating a non-contact sensor module within the main unit and integrating it with the main unit, wireless charging base, and waterproof sealed shell. Because this invention uses a non-contact detection method placed under the pillow, users do not need to wear additional electrodes or chest straps during sleep to achieve continuous and stable heart rate and respiratory rate monitoring, significantly improving comfort and compliance. It is suitable for long-term home health monitoring and auxiliary medical diagnosis.

[0060] This invention introduces sampling rate and time window length as common system parameters into the signal processing flow, and vertical offset and coupling stiffness as differential parameters, forming a dual-channel processing for cardiac and respiratory observations. By introducing spectral matching and cross-modal coherence constraints in the final optimization step, mutual verification and complementarity of information from different channels are achieved, effectively overcoming the problems of single-channel susceptibility to motion artifacts, differences in pillow core material, or placement offsets, thereby improving the robustness and accuracy of the detection results. Compared with existing technologies, the processing framework of this invention combines parameter winding and multi-channel fusion, demonstrating significant ingenuity.

[0061] This invention, by applying joint optimization and temporal smoothing constraints to cardiac and respiratory detection results, not only enables real-time output of heart rate and respiratory rate but also maintains the stability of results across time windows, avoiding misjudgments caused by frequent jumps. On the hardware side, it employs low-power Bluetooth communication and wireless charging to meet the application needs of everyday home monitoring and medical scenarios. On the software side, a data management platform enables result storage and analysis, supporting long-term individual trend monitoring and clinical reference. Therefore, this invention combines non-invasiveness, stability, and practicality, possessing significant value for widespread application. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of the overall structure of the non-contact heart rate and respiration recorder of the present invention.

[0063] Figure 2 This is a schematic diagram of the internal structure of the non-contact heart rate and respiration recorder of the present invention.

[0064] Figure 3 This is a schematic diagram illustrating the application scenario of the non-contact heart rate and respiration recorder of the present invention.

[0065] Figure 4 This is a flowchart illustrating the use of a non-contact sensor module in a non-contact heart rate and respiration recorder according to Embodiment 2 of the present invention.

[0066] In the diagram: 1. Main unit; 2. Wireless charging base; 3. Charging cable; 4. Terminal device; 5. Non-contact sensor module; 6. Battery. Detailed Implementation

[0067] Example 1

[0068] A non-contact heart rate and respiration recorder, comprising:

[0069] Host 1, which has a built-in non-contact sensor module 5, a microprocessor, a memory and a wireless communication module;

[0070] The wireless charging base 2 is used to wirelessly charge the host, and the wireless charging base 2 is connected to the charging cable 3;

[0071] A power module, including battery 6, provides power to the host unit;

[0072] The outer casing, with a waterproof and sealed structure, covers the main unit, achieving an IP68 protection rating.

[0073] The non-contact sensor module 5 includes a low-frequency acoustic piezoelectric sensor and a BCG cardiac pulse detection unit, used to collect human heart rate signals and respiratory signals;

[0074] The microprocessor is configured with a time-frequency domain joint algorithm for feature extraction and data processing of the signals acquired by the sensor module.

[0075] The wireless communication module is a low-power Bluetooth module, used to transmit the processed monitoring data to the external terminal device 4;

[0076] The terminal device 4 runs data management software for receiving and displaying the heart rate and respiratory rate data.

[0077] In use, the main unit 1 is placed under the pillow or inside the pillow core near the user's head. The non-contact sensor module 5 does not need to directly contact the skin, but can capture the user's heart rate and breathing information through body micro-movements and low-frequency acoustic signals. The main unit 1 transmits the processed data to the terminal device 4 in real time via Bluetooth Low Energy. The terminal device 4 can be a mobile phone, tablet, or medical monitoring platform. The data management software running on it is used to display real-time heart rate and respiratory rate curves, store historical data, and provide trend analysis and abnormality alerts. This embodiment avoids the discomfort of wearing traditional electrode patches, and at the same time, it realizes the visualization and remote management of individual health data through terminal software.

[0078] Specifically, the non-contact sensor module 5 is configured as follows:

[0079] S1. Acquire cardiac and respiratory channel signals at the suboccipital location using the sampling rate;

[0080] S2. Bandpass filtering and normalization are performed on the signal to obtain cardiac candidate components and respiratory candidate components;

[0081] S3. Perform artifact removal and robust segmentation within the time window length;

[0082] S4. Under the sampling rate, time window length and vertical offset parameters, the cardiac candidate component and the respiratory candidate component are correlated and fused to obtain the cardiac observation.

[0083] S5. Under the sampling rate, time window length and coupling stiffness parameters, the respiratory candidate components are stabilized and amplitude normalized to obtain respiratory observations;

[0084] S6. Using the cardiac and respiratory observations as inputs, perform joint optimization based on spectral matching and cross-modal coherence constraints to output heart rate and respiratory rate.

[0085] Specifically, the sensor module acquires cardiac and respiratory channel signals at the location under the pillow at a sampling rate. These signals are then bandpass filtered and normalized to obtain candidate cardiac and respiratory components. Artifact removal and robust segmentation are performed within a set time window to ensure the stability of the input data. Subsequently, under the control of parameters such as sampling rate, time window length, and vertical bias, cross-modal correlation fusion is performed on the candidate cardiac and respiratory components to form a cardiac observation X. Simultaneously, under the control of parameters such as sampling rate, time window length, and coupling stiffness, envelope stabilization and amplitude normalization are performed on the candidate respiratory component to obtain a respiratory observation Y. Finally, using observations X and Y as inputs, joint optimization is performed based on spectral matching and cross-modal coherence constraints to output heart rate (HR) and respiratory rate (RR). In this embodiment, the sampling rate and time window length act as common parameters for both cardiac and respiratory detection, while the vertical bias and coupling stiffness act as differential parameters for X and Y, respectively. In the final joint optimization, a two-layer entanglement of common and differential parameters is formed, ensuring the results remain stable despite placement deviations and pillow core variations.

[0086] Specifically, step S4 includes: performing mean-reduction and normalization processing on the cardiac candidate component and the respiratory candidate component within the time window length, calculating the correlation using a sliding window, and finely locating the correlation peak.

[0087] First, the cardiac and respiratory candidate components, after artifact removal, are subjected to mean-reduction and unit energy normalization. Within a time window, an overlapping sliding window is used to calculate the correlation between the two channels, and the correlation peaks are finely located within each window to obtain the cardiac observation X. By introducing a vertical bias parameter for occipital attenuation compensation, the significance and stability of the correlation peaks are significantly improved, thereby increasing the confidence of cardiac event detection.

[0088] Specifically, the cardiac observations satisfy the following relationship:

[0089]

[0090] τ∈[-τ0,τ0

[0091] in:

[0092] X(t) represents cardiac observations;

[0093] b bcg (t) represents the filtered and normalized result of the cardiac channel signal;

[0094] b ac(t) represents the filtered and normalized result of the respiratory channel signal;

[0095] τ is a delayed variable;

[0096] τ0 is the maximum allowable delay range;

[0097] Δz is the vertical offset parameter;

[0098] α is the attenuation coefficient;

[0099] <·,·> represent inner product operations;

[0100] ||·|| represents the signal energy norm.

[0101] The principle is that the micro-motions generated by the heartbeat are simultaneously reflected in the BCG and acoustic signals. However, due to differences in pillow thickness and sensor placement, the acoustic channel experiences significant attenuation and time delay. This embodiment achieves cross-channel amplitude alignment by introducing a parameter Δz and constructing an exponential compensation term αΔz, making the peak value of the cardiac component more prominent during correlation calculation. Maximizing the inner product ratio is essentially finding the optimal delay τ to align the cardiac signals of the two channels. This is equivalent to embedding the geometric parameter (Δz), sampling rate, and time window together into a correlation function, thus forming parameter entanglement. This method ensures that cardiac detection remains highly robust even under different pillow thicknesses and placement conditions, reducing the vulnerability of traditional single-channel peak detection to artifact interference.

[0102] Specifically, step S5 includes: performing envelope extraction on the respiratory candidate components, and performing low-pass filtering and amplitude normalization within the time window length to obtain respiratory observations.

[0103] First, the candidate respiratory components are subjected to Hilbert transform to extract the envelope. Then, a low-pass filter is applied within the time window to preserve the main respiratory frequency band components. Subsequently, the envelope signal is normalized in amplitude and drift suppressed using a coupling stiffness parameter to obtain the respiratory observation Y. In this embodiment, by introducing a coupling stiffness parameter, the amplitude drift caused by pillow core hardness or user body movement is effectively reduced, making the respiratory baseline more stable and the main peak of the respiratory spectrum more prominent.

[0104] Specifically, the respiratory observation Y satisfies the following relationship:

[0105]

[0106] in:

[0107] Y(t) represents the respiratory observation;

[0108] h lp (t;T w ) is based on the time length T w A low-pass filter with a scale of 1;

[0109] * indicates convolution operation;

[0110] H{·} represents the Hilbert transform;

[0111] k is the coupling stiffness parameter.

[0112] The principle is that respiratory signals mainly manifest as low-frequency envelope changes, but the amplitude is prone to non-stationary drift under different pillow materials and head pressure conditions. After extracting the envelope using Hilbert transform, if no correction is made, the drift will mask the true respiratory peak. This embodiment introduces the parameter κ into the denominator to dynamically normalize the signal amplitude. When the pillow is harder (κ is larger), the excessively amplified low-frequency drift is suppressed; when the pillow is softer (κ is smaller), more true respiratory fluctuations are preserved. Combined with low-pass filtering of the time window length, the respiratory wave baseline becomes stable and the main frequency is clear. Thus, the sampling rate fs and the time window length are intertwined with κ in the formula, ensuring both time-frequency resolution and introducing mechanical property correction, ultimately making the respiratory detection results more adaptable to the environment, superior to traditional fixed threshold methods.

[0113] Specifically, step S6 includes: using the spectral peak of cardiac observations as a candidate heart rate and the spectral peak of respiratory observations as a candidate respiratory rate, performing iterative optimization under the condition of satisfying the cross-modal coherence threshold, and applying temporal smoothing constraints between continuous time windows.

[0114] Specifically, the optimization satisfies the following objective function:

[0115] J(HR,RR)=w1D(PSD(X),HR)+w2D(PSD(Y),RR)+w3(1-Coh(X,Y;RR))

[0116] in,

[0117] J(HR,RR) is the objective function;

[0118] w1, w2, and w3 are positive weighting coefficients;

[0119] D(·) is a measure of spectral difference;

[0120] PSD(·) represents the power spectral density;

[0121] X represents cardiac observations;

[0122] Y represents respiratory observations;

[0123] Coh(X,Y;RR) is the coherence calculated at the respiratory rate RR.

[0124] The principle is to achieve robust estimation of heart rate and respiratory rate through three joint constraints: the first ensures the consistency of the spectral peaks of HR with cardiac observation X; the second ensures the consistency of the spectral peaks of RR with respiratory observation Y; and the third verifies cross-modal consistency by calculating the coherence of X and Y at respiratory frequency. This optimization function indirectly couples common parameters (sampling rate fs, time window length T_w) and differential parameters (vertical bias Δz, coupling stiffness κ) under the same optimization framework, so that the solutions for heart rate and respiratory rate are no longer independent of each other, but form a mutual verification mechanism through parameter entanglement. This mechanism ensures that the final solution can still converge stably in the presence of body motion interference or channel degradation, significantly improving the robustness and clinical reliability of non-contact monitoring, which is a technological advancement that was difficult for those skilled in the art to predict.

[0125] Compared with existing technologies, this invention introduces a two-layer parameter system (sampling rate, time window length) and differential parameters (vertical offset, coupling stiffness) into the signal processing framework for the first time. Common parameters act simultaneously on both cardiac observation X and respiratory observation Y, ensuring consistency in time and frequency resolution between the two signals. Differential parameters compensate for weaknesses in the cardiac and respiratory channels respectively. Vertical offset corrects for attenuation caused by differences in pillow thickness and placement, while coupling stiffness suppresses drift caused by varying pillow hardness and body movement. In the final joint optimization step, X and Y are simultaneously incorporated into the objective function, causing the effects of common and differential parameters to intertwine and entangle during the optimization process. This "parameter entanglement" allows heart rate and respiratory rate detection to no longer rely on a single channel or fixed compensation, but rather achieves robust output through the coupling of cross-modal mutual verification and differential compensation. Therefore, this invention not only overcomes the shortcomings of traditional single-channel detection, which is easily affected by body movement, pillow material, and placement, but also forms a systematic parameter design and algorithm framework that differs from existing solutions.

Claims

1. A non-contact heart rate and respiration recorder, characterized in that, include: The host (1) has a built-in non-contact sensor module (5), microprocessor, memory and wireless communication module; Wireless charging dock (2) for wirelessly charging the host; A power module, including a battery (6), provides power to the host; The outer casing, with a waterproof and sealed structure, covers the main unit, achieving an IP68 protection rating. The non-contact sensor module (5) includes a low-frequency acoustic piezoelectric sensor and a BCG cardiac pulse detection unit, used to collect human heart rate signals and respiratory signals; The microprocessor is configured with a time-frequency domain joint algorithm for feature extraction and data processing of the signals acquired by the sensor module. The wireless communication module is a low-power Bluetooth module, used to transmit the processed monitoring data to an external terminal device (4). The terminal device (4) runs data management software for receiving and displaying the heart rate and respiratory rate data; The non-contact sensor module (5) is configured as follows: S1. Acquire cardiac and respiratory channel signals at the suboccipital location using the sampling rate; S2. Bandpass filtering and normalization are performed on the signal to obtain cardiac candidate components and respiratory candidate components; S3. Perform artifact removal and robust segmentation within the time window length; S4. Under the sampling rate, time window length and vertical offset parameters, the cardiac candidate component and the respiratory candidate component are correlated and fused to obtain the cardiac observation. S5. Under the sampling rate, time window length and coupling stiffness parameters, the respiratory candidate components are stabilized and amplitude normalized to obtain respiratory observations; S6. Using the cardiac and respiratory observations as inputs, perform joint optimization based on spectral matching and cross-modal coherence constraints to output heart rate and respiratory rate.

2. The non-contact heart rate and respiration recorder according to claim 1, characterized in that, Step S4 includes: within the time window length, performing mean-reduction and normalization processing on the cardiac candidate components and respiratory candidate components, calculating the correlation using a sliding window, and finely locating the correlation peaks.

3. A non-contact heart rate and respiration recorder according to claim 2, characterized in that, The cardiac observations satisfy the following relationship: ； in: X(t) represents cardiac observations; b bcg (t) represents the filtered and normalized result of the cardiac channel signal; b ac (t) represents the filtered and normalized result of the respiratory channel signal; τ is a delayed variable; τ0 is the maximum allowable delay range; Δz is the vertical offset parameter; α is the attenuation coefficient; This is an inner product operation; Let be the signal energy norm.

4. A non-contact heart rate and respiration recorder according to claim 3, characterized in that, Step S5 includes: performing envelope extraction on the respiratory candidate components, and performing low-pass filtering and amplitude normalization within the time window length to obtain respiratory observations.

5. A non-contact heart rate and respiration recorder according to claim 4, characterized in that, The respiratory observation Y satisfies the following relationship: ； in: Y(t) represents the respiratory observation; h lp (t; T) ) is based on the time length T A low-pass filter with a scale of 1; * indicates convolution operation; For Hilbert transform; k is the coupling stiffness parameter.

6. A non-contact heart rate and respiration recorder according to claim 5, characterized in that, Step S6 includes: using the spectral peak of cardiac observations as a candidate heart rate and the spectral peak of respiratory observations as a candidate respiratory rate, performing iterative optimization under the condition of satisfying the cross-modal coherence threshold, and applying temporal smoothing constraints between continuous time windows.

7. A non-contact heart rate and respiration recorder according to claim 6, characterized in that, The optimization satisfies the following objective function: J(HR,RR)=w1D(PSD(X),HR)+w2D(PSD(Y),RR)+w3(1 Coh(X,Y;RR)); in, J(HR,RR) is the objective function; w1, w2, and w3 are positive weighting coefficients; As a measure of spectral difference; Power spectral density; X represents cardiac observations; Y represents respiratory observations; Coh(X,Y;RR) is the coherence calculated at the respiratory rate RR.

Images