Sleep heart rate monitoring device and self-adaptive non-contact detection method

Abstract

The invention relates to a sleep heart rate monitoring device and a self-adaptive non-contact detection method, and the detection method comprises the steps: S10, obtaining a pressure matrix sequence when a user lies on bedding through an array pressure sensor arranged on the bedding; s20, determining the pose of the user based on the pressure matrix sequence; s30, determining a weight corresponding to each pressure detection point of the array pressure sensor according to the pose of the user, and forming a weight matrix corresponding to the pressure matrix sequence; and S40, determining the sleep heart rate of the user according to the pressure matrix sequence and the weight matrix. According to the method, the stability of the detection precision is ensured, and the adaptability of the monitoring device to different users in different sleeping postures is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical monitoring, and in particular, to a sleep heart rate monitoring device and an adaptive non-contact detection method. Background Art

[0002] Monitoring sleep quality is of great significance to human health, and heart rate is one of the key physiological indicators for evaluating sleep quality. At present, although traditional contact heart rate monitoring devices have high measurement accuracy, they need to directly attach sensors to the human body surface, which is likely to affect normal sleep. Existing non-contact heart rate detection technologies, such as methods based on millimeter-wave radar, infrared imaging, etc., although avoiding direct contact, have problems such as signals being easily interfered by environmental noise, unstable detection accuracy, and difficulty in adapting to different sleep postures.

[0003] Specifically, firstly, traditional contact heart rate monitoring devices need to directly attach sensors to the human body surface. This measurement method causes discomfort to users and affects sleep quality, which is caused by the limitations of the measurement principle itself; secondly, existing non-contact heart rate detection technologies have problems of low detection accuracy and poor reliability because their signal processing algorithms are relatively simple and cannot effectively suppress the interference of human movement and environmental noise; finally, existing technologies lack an adaptive detection mechanism, resulting in the inability to automatically adjust detection parameters according to different users' sleep postures and individual differences, which is caused by the design of the detection method not considering individual differences. Summary of the Invention

[0004] (I) Technical Problems to be Solved

[0005] In view of the above-mentioned disadvantages and deficiencies of the prior art, the present invention provides a sleep heart rate monitoring device and an adaptive non-contact detection method, which solve the technical problems of unstable detection accuracy and difficulty in adapting to different sleep postures in the prior art.

[0006] (II) Technical Solutions

[0007] To achieve the above object, the main technical solutions adopted by the present invention include:

[0008] In the first aspect, an embodiment of the present invention provides an adaptive non-contact detection method for sleep heart rate, including:

[0009] S10. Obtain a pressure matrix sequence when the user lies on the bedding through an array pressure sensor disposed on the bedding;

[0010] S20. Determine the pose of the user based on the pressure matrix sequence;

[0011] S30. Determine the weight corresponding to each pressure detection point of the array pressure sensor according to the user's pose, and form a weight matrix corresponding to the pressure matrix sequence;

[0012] S40. Determine the user's sleep heart rate according to the pressure matrix sequence and the weight matrix.

[0013] Optionally, the S40 includes:

[0014] S401. Based on the pressure matrix sequence, obtain the original pressure time series corresponding to each pressure detection point;

[0015] S402. Perform preprocessing on the original pressure time series to obtain a standard pressure time series;

[0016] S403. Perform peak detection on the standard pressure time series to obtain a plurality of peaks and the corresponding moments of each peak; determine the time interval between two adjacent peaks as the heartbeat duration of one heartbeat, and based on the heartbeat duration, determine the original heart rate corresponding to the pressure detection point;

[0017] S404. Perform weighted averaging on the original heart rate corresponding to each pressure detection point according to the weight matrix to determine the user's sleep heart rate.

[0018] Optionally, the pressure matrix sequence includes a plurality of pressure matrices arranged in the order of acquisition time, and the S20 includes:

[0019] S201. Extract initial features from the pressure matrix; the initial features include: the pressure concentration area, the mirror similarity of the pressure matrix, the energy eigenvalue of the pressure matrix, and / or the centroid position of the user on the array pressure sensor;

[0020] S202. Input the initial features and the pressure matrix into a pose judgment model for classification to determine the user's pose; the pose includes: the user's posture, the user's position range, and the key area within the user's position range;

[0021] The pose judgment model is a deep learning model obtained through pre-training with appropriate model parameters.

[0022] Optionally, when the initial feature is the pressure concentration area, the S201 includes:

[0023] S201-A1. Select one or more significant pressure points in the pressure matrix as the initial points;

[0024] Among them, the significant pressure point is a pressure detection point whose pressure value is greater than a preset multiple of the average value of all elements of the pressure matrix;

[0025] S201 - A2. Take the adjacent elements of the initial point as the intermediate points. For each intermediate point, calculate the average value of the elements within the 3×3 neighborhood of the intermediate point as the neighborhood average value, and calculate the error rate between the intermediate point and its corresponding neighborhood average value. If the error rate is less than the first preset threshold, then include this intermediate point element in the candidate region corresponding to the initial point;

[0026] S201 - A3. Take the adjacent elements of the intermediate points included in the candidate region as new intermediate points, and repeat step S201 - A2 until no new intermediate points are included in the candidate region, to obtain the candidate region corresponding to the initial point;

[0027] S201 - A4. Repeat steps S201 - A2 and S201 - A3 to obtain the candidate region corresponding to each initial point;

[0028] S201 - A5. Calculate the variance of each candidate region according to formula (1), and take the candidate regions with variance greater than the preset second threshold as the pressure concentration regions;

[0029]

[0030] where, σ 2 represents the variance of the candidate region; k represents the number of pressure detection points in the candidate region; l represents the number of the pressure detection point in the candidate region; P l represents the pressure value of the pressure detection point numbered l; μ represents the average value of all pressure values in the candidate region.

[0031] Optionally, when the initial feature is the mirror similarity of the pressure matrix, S201 includes:

[0032] S201 - B1. Based on a preset third threshold, perform binarization processing on the pressure matrix to obtain a binarized matrix with element values of only 0 and 1;

[0033] S201 - B2. Perform mirror transformation on the binarized matrix along the longitudinal symmetry axis of the array pressure sensor to obtain its mirror matrix;

[0034] S201 - B3. Perform logical AND operation on the binarized matrix and its mirror matrix to obtain the number of intersection elements a, and perform logical OR operation on the binarized matrix and its mirror matrix to obtain the number of union elements b;

[0035] S201 - B4. Based on a and b, determine the mirror similarity J of the pressure matrix according to formula (2);

[0036] J = a / b (2).

[0037] Optionally, when the initial feature is the energy eigenvalue of the pressure matrix, S201 includes:

[0038] Based on the pressure matrix, according to formula (3), determine the energy eigenvalue E of the pressure matrix;

[0039]

[0040] where E represents the energy eigenvalue of the pressure matrix; i = 1, 2, … m represents the row number of the pressure detection points; j = 1, 2, … n represents the column number of the pressure detection points; P ij represents the pressure value at the i-th row and j-th column of the pressure matrix.

[0041] Optionally, when the initial feature is the centroid position of the user on the array pressure sensor, S201 includes:

[0042] Based on the pressure matrix, according to formulas (4) and (5), determine the centroid position (x c , y c ) of the user on the array pressure sensor;

[0043]

[0044] where x c represents the row number corresponding to the centroid position; y c represents the column number corresponding to the centroid position; i = 1, 2, … m represents the row number of the pressure detection points; j = 1, 2, … n represents the column number of the pressure detection points; P ij represents the pressure value at the i-th row and j-th column of the pressure matrix.

[0045] Optionally, S30 includes:

[0046] S301. Determine the non-critical areas within the user's position range according to the user's posture, the user's position range, and the key areas within the user's position range;

[0047] S302. Assign the weight a to the pressure detection points corresponding to the key areas within the user's position range, assign the weight b to the pressure detection points corresponding to the non-critical areas within the user's position range, and assign the weight 0 to the pressure detection points corresponding to the non-user range of the array pressure sensor, to obtain a weight matrix, where a > b and a + b = 1.

[0048] Optionally, S301 includes:

[0049] When the user's posture is supine, the key areas are: the chest area and / or the wrist area;

[0050] When the user's posture is lateral recumbent, the key areas are: the chest area, the arm area, and / or the wrist area;

[0051] When the posture of the user is the prone position, the key areas are: the chest area and / or the forehead area.

[0052] In a second aspect, an embodiment of the present invention provides a sleep heart rate monitoring device, including:

[0053] An array pressure sensor, disposed on the bedding, for acquiring a pressure matrix sequence when the user lies on the bedding;

[0054] An analysis device, configured to determine the posture of the user based on the pressure matrix sequence; determine the weight corresponding to each pressure detection point of the array pressure sensor according to the posture of the user, and form a weight matrix corresponding to the pressure matrix sequence; and determine the sleep heart rate of the user according to the pressure matrix sequence and the weight matrix.

[0055] (III) Advantageous Effects

[0056] The detection method and monitoring device proposed by the present invention obtain a pressure matrix sequence when the user lies on the bedding through an array pressure sensor disposed on the bedding; determine the posture of the user based on the pressure matrix sequence; determine the weight corresponding to each pressure detection point of the array pressure sensor according to the posture of the user, and form a weight matrix corresponding to the pressure matrix sequence; and determine the sleep heart rate of the user according to the pressure matrix sequence and the weight matrix.

[0057] Based on the above steps, the present invention determines the weight corresponding to each pressure detection point based on the posture of the user, sets a higher weight for the key areas within the user's position range, ensures the stability of the detection accuracy, and significantly improves the adaptability of the monitoring device to different users in different sleep postures, solving the problem that it is difficult for the prior art to cope with individual differences. Description of the Drawings

[0058] Figure 1 It is a schematic flowchart of an adaptive non-contact detection method for sleep heart rate provided in the embodiment;

[0059] Figure 2 It is a schematic flowchart of S40 in the embodiment;

[0060] Figure 3 It is a schematic flowchart of the process of extracting the pressure concentration area in the embodiment;

[0061] Figure 4 It is a schematic flowchart of the process of extracting the mirror similarity of the pressure matrix in the embodiment. Detailed Embodiments

[0062] To better understand the above technical solution, the exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more clear and thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0063] Embodiment 1

[0064] As Figure 1 shown, this embodiment provides an adaptive non-contact detection method for sleep heart rate, including:

[0065] S10. Obtain a pressure matrix sequence when the user lies on the bedding through an array pressure sensor disposed on the bedding.

[0066] S20. Determine the user's pose based on the pressure matrix sequence.

[0067] S30. According to the user's pose, determine the weight corresponding to each pressure detection point of the array pressure sensor to form a weight matrix corresponding to the pressure matrix sequence.

[0068] S40. Determine the user's sleep heart rate according to the pressure matrix sequence and the weight matrix.

[0069] Based on the above steps, in this embodiment, the weight corresponding to each pressure detection point is determined based on the user's pose, and a higher weight is set for the key area within the user's position range to ensure the stability of the detection accuracy, significantly improving the adaptability of the monitoring device to different users in different sleep postures and solving the problem that the prior art is difficult to cope with individual differences.

[0070] In a specific implementation manner of this embodiment, the pressure matrix sequence includes a plurality of pressure matrices arranged in the order of acquisition time, and the S20 includes:

[0071] S201. Extract initial features from the pressure matrix; the initial features include: the pressure concentration area, the mirror similarity of the pressure matrix, the energy eigenvalue of the pressure matrix, and / or the centroid position of the user on the array pressure sensor.

[0072] S202. Input the initial features and the pressure matrix into a pose judgment model for classification to determine the user's pose; the pose includes: the user's posture, the user's position range, and the key area within the user's position range.

[0073] Specifically, the postures of the user include: supine position, lateral position, and prone position. The key areas within the user's position range are actually the areas where the detection accuracy of the corresponding array pressure sensors is the highest.

[0074] When the user's posture is the supine position, the key areas are: the chest area and / or the wrist area. When lying supine, the chest is in direct contact with the bedding below. The pulsation of the heart will be transmitted through body tissues to the pressure sensor below the chest, causing slight pressure changes. This area is relatively stable and less disturbed by the movements of other body parts, and can better reflect the pressure fluctuations generated by the heart pulsation, thus calculating the heart rate more accurately. There is a radial artery at the wrist. When the wrist is naturally placed on the bedding, the arterial pulsation can also cause the pressure sensor below the wrist to detect pressure changes. Moreover, the skin of the wrist fits well with the sensor, and the signal is relatively stable, which helps to accurately calculate the heart rate.

[0075] When the user's posture is the lateral position, the key areas are: the chest area, the arm area, and / or the wrist area. When in the lateral position, the chest closer to the bed surface is closer to the pressure sensor, and the pressure signal transmitted by the heart pulsation is more direct. This area can effectively capture the pressure changes caused by the heart beating, reducing the pressure interference from other body parts, and is relatively accurate for calculating the heart rate. Taking the left lateral position as an example, the pressure sensor under the inner side of the right arm (the side closer to the bed surface) can better detect the pressure changes generated by the arm blood vessels due to the heart pulsation. Because the arm and the wrist are relatively stable in the lateral position, and the distance between the blood vessels and the sensor is relatively close, the signal quality is good, which is conducive to accurately obtaining heart rate information.

[0076] When the user's posture is the prone position, the key areas are: the chest area and / or the forehead area. Although the chest bears a large pressure when lying prone, the area in direct contact with the bedding in front of the chest can more sensitively perceive the pressure changes caused by the heart pulsation. This pressure change has a high correlation with the heart beating. By analyzing the pressure values in this area, the heart rate can be calculated relatively accurately. The forehead is in contact with the bedding when lying prone, and the blood vessels in the head will generate pressure fluctuations due to the heart pulsation. The pressure sensor below the forehead can detect this fluctuation. The forehead area is relatively flat, has good contact with the sensor, and is less affected by the movements of other body parts, and can provide relatively accurate pressure data for heart rate calculation.

[0077] The pose judgment model is a deep learning model obtained through pre-training with appropriate model parameters.

[0078] Specifically, the deep learning model can specifically be models such as CNN, AlexNet, VGG, etc.

[0079] The training dataset of the deep learning model can be obtained in advance by different testers lying on the array pressure sensor in different postures, and annotation tools such as labelme or manual annotation are used to mark the corresponding postures, position ranges, and key areas for the obtained pressure matrices, resulting in a sequence of annotated pressure matrices.

[0080] Input into the deep learning model together with the pressure matrix is also the initial features extracted from the pressure matrix itself; the initial features include: the pressure concentration area, the mirror similarity of the pressure matrix, the energy eigenvalue of the pressure matrix, and / or the centroid position of the user on the array pressure sensor. Specifically, the extraction method for each of the above initial features is as follows:

[0081] When the initial feature is the pressure concentration area, different sleeping postures will result in different pressure concentration patterns. For example, when lying on the back, the pressure concentration area may be relatively dispersed and approximately rectangular, while when lying on the stomach, the pressure concentration area may be more concentrated under the chest and abdomen, and the shape will also be different. These features can be used as a basis for distinguishing different postures. Specifically, the method for extracting the pressure concentration area includes steps A1 to A5:

[0082] A1. Select one or more significant pressure points in the pressure matrix as the initial points.

[0083] Among them, the significant pressure point is a pressure detection point whose pressure value is greater than a preset multiple of the average value of all elements in the pressure matrix. Specifically, the range of the preset multiple can be [3, 6].

[0084] A2. Take the adjacent elements of the initial point as the intermediate points. For each intermediate point, calculate the average value of the elements in the 3×3 neighborhood of the intermediate point as the neighborhood average value, and calculate the error rate between the intermediate point and its corresponding neighborhood average value. If the error rate is less than the first preset threshold, then classify the intermediate point element into the candidate area corresponding to the initial point. Specifically, the first preset threshold can be [0, 0.2]. That is, calculate the error between the average pressure value of the small neighborhood centered on the current intermediate point and the intermediate point. If the error is small, it is considered that the pressure values of the current intermediate point and the initial point are relatively close.

[0085] A3. Take the adjacent elements of the intermediate points classified into the candidate area as the new intermediate points, and repeat step A2 until no new intermediate points are classified into the candidate area, obtaining the candidate area corresponding to the initial point.

[0086] A4. Repeat steps A2 and A3 to obtain the candidate area corresponding to each initial point.

[0087] A5. Calculate the variance of each candidate region according to formula (1), and use the candidate regions with variances greater than a preset second threshold as pressure concentration regions. Specifically, the second threshold can be 1.5 to 3 times the overall variance of the pressure matrix. When the variance of a candidate region is greater than the preset second threshold, it is considered that the degree of uneven pressure distribution in this region is relatively prominent compared to the entire pressure array and can be regarded as the user's position area. Setting the upper limit of the second threshold can exclude external environmental interference.

[0088]

[0089] Among them, σ 2 represents the variance of the candidate region; k represents the number of pressure detection points in the candidate region; l represents the number of the pressure detection point in the candidate region; P l represents the pressure value of the pressure detection point numbered l; μ represents the average value of all pressure values in the candidate region.

[0090] When the initial feature is the mirror similarity of the pressure matrix, it can better capture the user's posture. For example, the supine position usually has good left - right symmetry, while the lateral position shows obvious asymmetry in the left - right direction. The symmetry can be measured by calculating indicators such as the mirror similarity of the pressure matrix, which is helpful for distinguishing different sleep postures. The methods for extracting the mirror similarity of the pressure matrix include:

[0091] B1. Based on a preset third threshold, perform binarization processing on the pressure matrix to obtain a binarized matrix with element values of only 0 and 1. Specifically, all non - zero values of the pressure matrix can be set to 1. To exclude environmental interference, the third threshold can also be set to the maximum pressure value when items such as blankets and quilts are laid on the array pressure sensor.

[0092] B2. Mirror - transform the binarized matrix along the longitudinal symmetry axis of the array pressure sensor to obtain its mirror matrix.

[0093] B3. Perform a logical AND operation on the binarized matrix and its mirror matrix to obtain the number of intersection elements a, and perform a logical OR operation on the binarized matrix and its mirror matrix to obtain the number of union elements b.

[0094] B4. Based on a and b, determine the mirror similarity J of the pressure matrix according to formula (2).

[0095] J = a / b (2).

[0096] When the initial feature is the energy eigenvalue of the pressure matrix, the pressure matrix can be regarded as a two-dimensional signal, and its energy can be calculated. The energy can be represented by the sum of the squares of all element values in the pressure matrix. Different sleep postures will result in different contact areas and pressure distributions between the body and the bedding, thus leading to different energies of the pressure matrix. For example, when lying on the side, the contact area between the body and the bedding is relatively small, and the energy of the pressure matrix may be relatively low; while when lying on the back, the contact area is larger, and the energy may be relatively high. The calculation method of the energy eigenvalue of the pressure matrix includes:

[0097] Based on the pressure matrix, according to formula (3), determine the energy eigenvalue E of the pressure matrix;

[0098]

[0099] where E represents the energy eigenvalue of the pressure matrix; i = 1, 2, … m represents the row number of the pressure detection point; j = 1, 2, … n represents the column number of the pressure detection point; P ij represents the pressure value at the i-th row and j-th column of the pressure matrix.

[0100] The centroid position of the user on the array pressure sensor can better determine the user's position. The calculation method of the centroid position of the user on the array pressure sensor is:

[0101] Based on the pressure matrix, according to formulas (4) and (5), determine the centroid position (x c , y c ) of the user on the array pressure sensor.

[0102]

[0103] where x c represents the row number corresponding to the centroid position; y c represents the column number corresponding to the centroid position; i = 1, 2, … m represents the row number of the pressure detection point; j = 1, 2, … n represents the column number of the pressure detection point; P ij represents the pressure value at the i-th row and j-th column of the pressure matrix.

[0104] Training the deep learning model based on the above-mentioned labeled pressure array and initial features can enable the deep learning model to accurately obtain the features of the user's individual information and posture information, and more accurately judge the user's posture, the user's position range, and the key areas within the user's position range, providing a data basis for the weight matrix used in subsequent calculation of the user's sleep heart rate.

[0105] In another specific implementation manner of this embodiment, based on the user's pose determined in step S20, S30 includes:

[0106] S301. Determine the non-critical areas within the user's position range based on the user's posture, the user's position range, and the key areas within the user's position range.

[0107] S302. Assign the weight corresponding to the pressure detection point in the key area within the user's position range as a, assign the weight corresponding to the pressure detection point in the non-critical area within the user's position range as b, and assign the weight corresponding to the pressure detection point in the non-user range of the array pressure sensor as 0, to obtain a weight matrix, where a > b and a + b = 1. Specifically, the allocation of a and b is divided according to the importance degree of the key area. Preferably, the value range of a is [0.7, 1].

[0108] The specific steps of S40 include:

[0109] S401. Based on the pressure matrix sequence, obtain the original pressure time series corresponding to each pressure detection point.

[0110] S402. Preprocess the original pressure time series to obtain a standard pressure time series.

[0111] Specifically, use a Butterworth filter to preprocess the data to remove noise and unnecessary frequency components. Use the butter function to generate the coefficients of the filter, and then use the filtfilt function to perform zero-phase filtering on the data. Calculate the rolling average and standard deviation of the data to determine the threshold for peak detection. Window size: 50 data points. Threshold: rolling average plus 1.5 times the standard deviation.

[0112] S403. Perform peak detection on the standard pressure time series to obtain multiple peaks and the corresponding moments of each peak; determine the time interval between two adjacent peaks as the heartbeat duration of one heartbeat, and based on the heartbeat duration, determine the original heart rate corresponding to this pressure detection point.

[0113] Specifically, use the find_peaks function to detect the peaks in the data. Minimum interval: 0.5 seconds interval for 50 data points. Height threshold: determined by an adaptive threshold. Prominence: 0.5, indicating the prominence of the peak. Width: 3 data points, indicating the minimum width of the peak.

[0114] If at least 3 peaks are detected, calculate the intervals (in seconds) between adjacent peaks, remove abnormal intervals (intervals less than 0.5 seconds or greater than 2.0 seconds), calculate the original heart rate using the median to avoid the influence of outliers, verify whether the heart rate is within a reasonable range (40 - 200 bpm), and return the heart rate value rounded to one decimal place.

[0115] The core idea of step S403 is as follows: First, the main frequency range (0.5 - 3.0 Hz) of the heartbeat signal is cleaned out by a band - pass filter. Then, the peaks in the waveform are identified using an adaptive threshold (mean + standard deviation). Finally, the heart rate is deduced by calculating the time intervals between adjacent peaks, and multiple validations (number of peaks, interval range, final heart rate range) are used to ensure the reliability of the result.

[0116] S404. For the original heart rate corresponding to each pressure detection point, weighted averaging is performed according to the weight matrix to determine the user's sleep heart rate.

[0117] Embodiment 2

[0118] An embodiment of the present invention provides a sleep heart rate monitoring device, including:

[0119] An array pressure sensor, disposed on the bedding, for obtaining a pressure matrix sequence when the user lies on the bedding.

[0120] An analysis device, for determining the user's posture based on the pressure matrix sequence; determining the weight corresponding to each pressure detection point of the array pressure sensor according to the user's posture, forming a weight matrix corresponding to the pressure matrix sequence; and determining the user's sleep heart rate according to the pressure matrix sequence and the weight matrix.

[0121] Specifically, the analysis device includes:

[0122] A data acquisition module, for obtaining a pressure matrix sequence when the user lies on the bedding through the array pressure sensor disposed on the bedding.

[0123] A posture module, for determining the user's posture based on the pressure matrix sequence.

[0124] A weight module, for determining the weight corresponding to each pressure detection point of the array pressure sensor according to the user's posture, forming a weight matrix corresponding to the pressure matrix sequence.

[0125] A heart rate module, for determining the user's sleep heart rate according to the pressure matrix sequence and the weight matrix.

[0126] In addition, the sleep heart rate monitoring device may further include a display device, for displaying the user's sleep heart rate, as well as information such as the breathing rate, blood oxygen concentration, body temperature, blood pressure, etc. monitored by the heart rate monitoring device through other sensors.

[0127] Since the system / apparatus described in the above embodiments of the present invention is the system / apparatus adopted for implementing the method of the above embodiments of the present invention, those skilled in the art can understand the specific structure and variations of the system / apparatus based on the method described in the above embodiments of the present invention, and thus will not be elaborated herein. Any system / apparatus adopted for the method of the above embodiments of the present invention falls within the scope of protection of the present invention.

[0128] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can be implemented in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be realized by computer program instructions.

[0130] It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in the claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention can be realized by means of hardware including several different elements and by means of a suitably programmed computer. In the claims listing several apparatuses, several of these apparatuses can be embodied by the same hardware. The use of the words first, second, third, etc. is only for convenience of expression and does not indicate any order. These words can be construed as part of the element name.

[0131] In addition, it should be noted that in the description of this specification, the description of terms such as "an embodiment", "some embodiments", "embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0132] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications after learning the basic creative concepts. Therefore, the claims should be construed to cover the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0133] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and its equivalent technologies, the present invention should also cover these modifications and variations.

Claims

1. An adaptive non-contact detection method for sleep heart rate, characterized in that, Including: S10. Obtain a pressure matrix sequence when a user lies on a bedding through an array pressure sensor disposed on the bedding; S20. Determine the pose of the user based on the pressure matrix sequence; S30. Determine the weight corresponding to each pressure detection point of the array pressure sensor according to the pose of the user, and form a weight matrix corresponding to the pressure matrix sequence; S40. Determine the sleep heart rate of the user according to the pressure matrix sequence and the weight matrix.

2. The detection method according to claim 1, characterized in that, The S40 includes: S401. Based on the pressure matrix sequence, obtain an original pressure time series corresponding to each pressure detection point; S402. Perform preprocessing on the original pressure time series to obtain a standard pressure time series; S403. Perform peak detection on the standard pressure time series to obtain a plurality of peaks and the moments corresponding to each peak; determine the time interval between two adjacent peaks as the heartbeat duration of one heartbeat, and based on the heartbeat duration, determine the original heart rate corresponding to this pressure detection point; S404. Perform weighted averaging on the original heart rate corresponding to each pressure detection point according to the weight matrix to determine the sleep heart rate of the user.

3. The detection method according to claim 1, wherein The pressure matrix sequence includes a plurality of pressure matrices arranged in the order of acquisition time, and the S20 includes: S201. Extract initial features from the pressure matrix; the initial features include: a pressure concentration region, the mirror similarity of the pressure matrix, the energy eigenvalue of the pressure matrix, and / or the centroid position of the user on the array pressure sensor; S202. Input the initial features and the pressure matrix into a pose judgment model for classification to determine the pose of the user; the pose includes: the posture of the user, the position range of the user, and the key regions within the user's position range; The pose judgment model is a deep learning model obtained through pre-training with appropriate model parameters.

4. The detection method according to claim 3, wherein When the initial feature is the pressure concentration region, the S201 includes: S201-A1. Select one or more significant pressure points in the pressure matrix as initial points; Wherein, the significant pressure point is a pressure detection point whose pressure value is greater than a preset multiple of the average value of all elements of the pressure matrix; S201-A2. Take the adjacent elements of the initial point as intermediate points. For each intermediate point, calculate the average value of the elements within the 3×3 neighborhood of the intermediate point as the neighborhood average value, calculate the error rate between the intermediate point and its corresponding neighborhood average value. If the error rate is less than the first preset threshold, then classify this intermediate point element into the candidate region corresponding to the initial point; S201-A3. Take the adjacent elements of the intermediate point included in the candidate region as new intermediate points, and repeat step S201-A2 until no new intermediate points are included in the candidate region to obtain the candidate region corresponding to this initial point; S201-A4. Repeat steps S201-A2 and S201-A3 to obtain the candidate region corresponding to each initial point; S201-A5. Calculate the variance of each candidate region according to formula (1), and take the candidate region with a variance greater than the preset second threshold as the pressure concentration region; Among them, σ 2 represents the variance of the candidate region; k represents the number of pressure detection points in the candidate region; l represents the number of the pressure detection point in the candidate region; P l represents the pressure value of the pressure detection point numbered l; μ represents the average value of all pressure values in the candidate region.

5. The detection method according to claim 3, wherein When the initial feature is the mirror similarity of the pressure matrix, the S201 includes: S201 - B1. Binarize the pressure matrix based on a preset third threshold to obtain a binarized matrix with element values of only 0 and 1; S201 - B2. Mirror - transform the binarized matrix along the longitudinal symmetry axis of the array pressure sensor to obtain its mirror matrix; S201 - B3. Perform a logical AND operation on the binarized matrix and its mirror matrix to obtain the number of intersection elements a, and perform a logical OR operation on the binarized matrix and its mirror matrix to obtain the number of union elements b; S201 - B4. Based on a and b, determine the mirror similarity J of the pressure matrix according to formula (2); J = a / b (2).

6. The detection method according to claim 3, characterized in that, When the initial feature is the energy eigenvalue of the pressure matrix, S201 includes: Based on the pressure matrix, determine the energy eigenvalue E of the pressure matrix according to formula (3); Among them, E represents the energy eigenvalue of the pressure matrix; i = 1, 2, … m represents the row number of the pressure detection points; j = 1, 2, … n represents the column number of the pressure detection points; P ij represents the pressure value at the i-th row and j-th column of the pressure matrix.

7. The detection method according to claim 3, characterized in that, When the initial feature is the centroid position of the user on the array pressure sensor, S201 includes: Based on the pressure matrix, according to formulas (4) and (5), determine the centroid position (x c , y c ) of the user on the array pressure sensor; Among them, x c represents the row number corresponding to the centroid position; y c represents the column number corresponding to the centroid position; i = 1, 2, … m represents the row numbers of the pressure detection points; j = 1, 2, … n represents the column numbers of the pressure detection points; P ij represents the pressure value at the i-th row and j-th column of the pressure matrix.

8. The detection method according to any one of claims 3 to 7, characterized in that S30 includes: S301. Determine the non - critical areas within the user's position range according to the user's posture, the user's position range, and the key areas within the user's position range; S302. Assign the weight corresponding to the pressure detection points in the key areas within the user's position range as a, assign the weight corresponding to the pressure detection points in the non - critical areas within the user's position range as b, and assign the weight corresponding to the pressure detection points outside the user's range of the array pressure sensor as 0 to obtain a weight matrix, where a > b and a + b = 1.

9. The detection method according to claim 8, wherein S301 includes: When the user's posture is supine, the key areas are: the chest area and / or the wrist area; When the user's posture is lateral recumbent, the key areas are: the chest area, the arm area, and / or the wrist area; When the user's posture is prone, the key areas are: the chest area and / or the forehead area.

10. A sleep heart rate monitoring device, characterized in that, Includes: An array pressure sensor, disposed on the bedding, for obtaining a sequence of pressure matrices when the user lies on the bedding; An analysis device, for determining the user's pose based on the sequence of pressure matrices; determining the weight corresponding to each pressure detection point of the array pressure sensor according to the user's pose to form a weight matrix corresponding to the sequence of pressure matrices; and determining the user's sleep heart rate according to the sequence of pressure matrices and the weight matrix.

Images