A sleep lying posture recognition method, device, equipment and medium

Abstract

The application discloses a sleep posture recognition method, device and equipment and a medium. The method comprises the following steps: acquiring a radar echo signal matrix reflected by a target object at least at one time; processing the radar echo signal matrix at the current time to obtain a to-be-used signal matrix; taking the current time as the cutoff time of a historical preset first time length, determining a body movement index of the target object at the current time based on the to-be-used signal matrix at each time within the historical preset first time length; in the case that the body movement index meets a preset condition, generating a to-be-used spectrogram based on the to-be-used signal matrix at the current time; and determining a posture recognition result corresponding to the target object based on the to-be-used spectrogram. The application solves the problem that the sleep posture recognition accuracy is low in the prior art because the object speed, angle and other information are determined by analyzing the radar signal to recognize the posture, improves the sleep posture recognition accuracy and efficiency, and meets the user demand.

Description

Technical Field

[0001] This invention relates to the field of computer processing technology, and in particular to a method, apparatus, device, and medium for sleep posture recognition. Background Technology

[0002] In order to better meet users' needs for sleep quality, the variety of sleep posture recognition devices is becoming increasingly diverse. In application, sleep posture recognition devices are usually used to identify users' sleeping postures in order to analyze users' sleep patterns based on their sleeping postures and provide users with better sleep services.

[0003] Currently, methods for sleep posture recognition typically use radar sensors to emit electromagnetic wave signals and receive the radar echo signals reflected by objects. The received signals are then analyzed to obtain information such as the object's speed and angle, which is used to identify the sleeping posture. However, in a static state, the information in the radar echo is sparse, making feature extraction difficult. Furthermore, during sleep, the sleeping posture changes slightly, resulting in minimal changes to the object's speed and angle, leading to poor accuracy in sleep posture recognition. Summary of the Invention

[0004] This invention provides a method, device, equipment, and medium for sleep posture recognition, which improves the accuracy and efficiency of sleep posture recognition, thereby meeting user needs.

[0005] According to one aspect of the present invention, a method for recognizing sleep posture is provided, the method comprising:

[0006] Obtain the radar echo signal matrix reflected by the target object at at least one moment;

[0007] For each time moment, the radar echo signal matrix at the current time is processed to obtain the signal matrix to be used;

[0008] Using the current moment as the end time of the first historical preset duration, the body movement index of the target object at the current moment is determined based on the signal matrix to be used at each moment within the first historical preset duration.

[0009] When the body motion index meets preset conditions, a spectrum to be used is generated based on the signal matrix to be used at the current moment; wherein, the spectrum to be used includes a first spectrum and a second spectrum; the first spectrum includes the relative distance and azimuth information between the target object and the radar; the second spectrum includes the relative distance and velocity information between the target object and the radar;

[0010] Based on the spectrum to be used, the reclining posture recognition result corresponding to the target object is determined.

[0011] According to another aspect of the present invention, a sleep posture recognition device is provided, the device comprising:

[0012] The signal acquisition module is used to acquire the radar echo signal matrix reflected by the target object at at least one moment;

[0013] The module for determining the signal matrix to be used is used to process the radar echo signal matrix at the current time for each time moment to obtain the signal matrix to be used.

[0014] The body movement index determination module is used to determine the body movement index of the target object at the current time, based on the signal matrix to be used at each time within the historical preset first time, with the current time as the cutoff time of the historical preset first time.

[0015] The spectrum determination module is used to generate a spectrum to be used based on the signal matrix to be used at the current moment, provided that the body motion index meets preset conditions; wherein, the spectrum to be used includes a first spectrum and a second spectrum; the first spectrum includes the relative distance and azimuth information between the target object and the radar; the second spectrum includes the relative distance and velocity information between the target object and the radar;

[0016] The posture recognition result determination module is used to determine the posture recognition result corresponding to the target object based on the spectrum to be used.

[0017] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0018] At least one processor; and

[0019] A memory communicatively connected to the at least one processor; wherein,

[0020] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the sleep posture recognition method according to any embodiment of the present invention.

[0021] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the sleep posture recognition method according to any embodiment of the present invention.

[0022] The technical solution of this invention involves acquiring at least one radar echo signal matrix reflected by a target object at any given time; processing the radar echo signal matrix at the current time to obtain a signal matrix to be used; using the current time as the cutoff time of a historical preset first time period, determining the body movement index of the target object at the current time based on the signal matrix to be used at each time within the historical preset first time period; generating a spectrum to be used based on the signal matrix to be used at the current time when the body movement index meets preset conditions; and determining the recumbent posture recognition result corresponding to the target object based on the spectrum to be used. This solves the problem of low accuracy in recumbent posture recognition caused by analyzing radar signals to determine object speed, angle, and other information in the prior art, and achieves the goal of obtaining a recumbent posture recognition result by analyzing the radar signal matrix at the current time. The radar echo signal matrix reflected by the target object at the previous moment is processed to obtain the signal matrix to be used. Then, combined with the signal matrix to be used at the current moment and the historical moments before the current moment, the body movement index of the target object at the current moment is determined. This allows the body movement index to accurately reflect the changes in the target object's sleeping posture. The body movement index that does not meet the preset conditions is filtered out, and the moment when the body movement index meets the preset conditions is extracted. This moment is when the target object is in a relatively static state. The signal matrix to be used at the moment when the target object is in a relatively static state is used to generate the spectrum to be used, and the corresponding sleeping posture recognition result is determined. This achieves high recognition efficiency while improving the accuracy of sleeping posture recognition, thus meeting the user's needs.

[0023] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of a sleep posture recognition method according to Embodiment 1 of the present invention;

[0026] Figure 2 This is a schematic diagram of the lying posture recognition method provided in Embodiment 1 of the present invention;

[0027] Figure 3 This is a flowchart of a sleep posture recognition method according to Embodiment 2 of the present invention;

[0028] Figure 4This is a flowchart of a sleep posture recognition method according to Embodiment 3 of the present invention;

[0029] Figure 5 This is a schematic diagram of a sleep posture recognition device according to Embodiment 4 of the present invention;

[0030] Figure 6 This is a schematic diagram of the structure of an electronic device that implements the sleep posture recognition method of this invention. Detailed Implementation

[0031] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] Example 1

[0034] Figure 1 This is a flowchart of a sleep posture recognition method according to Embodiment 1 of the present invention. This embodiment is applicable to the recognition of human sleep postures. The method can be executed by a sleep posture recognition device, which can be implemented in hardware and / or software. The sleep posture recognition device can be configured in radar equipment. Figure 1 As shown, the method includes:

[0035] S110. Obtain at least one radar echo signal matrix reflected by the target object at any given time.

[0036] The target object can be a user whose lying posture needs to be identified.

[0037] In this embodiment, the radar can be controlled to emit electromagnetic wave signals towards the space where the target object is located. After being reflected by the target object, the emitted electromagnetic wave signals are received by the radar receiver. All signals received at the same time constitute a radar echo signal matrix, which may include the signal energy values ​​of multiple points in space. Optionally, the radar can be a low-complexity millimeter-wave radar. Low-complexity millimeter-wave radar has a simplified circuit design, which helps to reduce cost and power consumption. At the same time, millimeter-wave radar has high resolution, strong penetration and high bandwidth, enabling it to transmit more data, improving the stability of sleep posture recognition while ensuring recognition accuracy.

[0038] S120. For each time moment, process the radar echo signal matrix at the current time moment to obtain the signal matrix to be used.

[0039] It should be noted that when performing prone positioning recognition of a target object based on radar signals, the radar echo signal matrix at at least one moment can be acquired in real time or periodically. The method for processing the radar echo signal matrix at each moment to obtain the signal matrix to be used is the same. Any moment can be taken as the current moment, and the process of processing the radar echo signal matrix at the current moment to obtain the signal matrix to be used will be used as an example for the introduction.

[0040] In this embodiment, the radar echo signal matrix at the current moment can be processed, such as through signal amplification, denoising, and sampling, to obtain a processed signal matrix to be used. Specifically, the processing of the radar echo signal matrix at the current moment to obtain the signal matrix to be used can be achieved by: preprocessing the radar echo signal matrix at the current moment to obtain a discrete echo signal matrix; performing a Fast Fourier Transform on the discrete echo signal matrix to obtain a frequency domain signal matrix; and performing clutter suppression processing on the frequency domain signal matrix to obtain the signal matrix to be used.

[0041] In practical applications, the signals in the radar echo signal matrix can be amplified to increase their amplitude and compensate for potential attenuation during transmission, resulting in an amplified output signal. Then, the amplified signal undergoes frequency mixing, such as multiplying it with a locally generated oscillator signal to down-convert it to an intermediate frequency (IF) or baseband signal, yielding a mixed signal. Next, the mixed signal can be sampled according to a preset sampling frequency, and then subjected to analog-to-digital conversion (ADC). That is, since the sampled signal is analog, it can be mapped to discrete digital values, converting the sampled signal into a digital signal, thus achieving ADC. All digital signals can be used as a discrete echo signal matrix. Furthermore, a Fast Fourier Transform (FFT) algorithm is used to perform a FFT on the discrete echo signal matrix to extract features from the signal. These features can include the range information of the target object, resulting in a frequency domain signal matrix containing range features. By performing clutter suppression processing on the frequency domain signal matrix and retaining the signal information of the target object, the signal matrix after clutter suppression is used as the signal matrix to be used. By filtering out interference signals, the user's lying posture can be identified, which can effectively improve the accuracy of lying posture recognition.

[0042] For example, see Figure 2 After receiving the radar echo signal matrix, it can be sequentially processed through a signal amplifier, mixer, and ADC (Analog-to-Digital Converter) to obtain a discrete echo signal matrix. The echo signal in the discrete echo signal matrix can be represented as y(l,n,k), where k is the index of the antenna channel; l is the index of the linear frequency modulated continuous wave (or slow time), representing the position in the continuously transmitted radar pulse wave sequence; n is the index of the sampling point (or fast time), representing the sampling position in the radar pulse wave; y(l,n,k) represents the nth sampling point signal in the lth linear frequency modulated continuous wave under the kth antenna channel, and these sampling point signals are the echo signals. A Fast Fourier Transform (FFT) is performed on the discrete echo signal matrix in the fast time dimension (i.e., the sampling time within the linear frequency modulated continuous wave) to transform the signal from the time domain to the frequency domain. In the frequency domain, each frequency component corresponds to a range cell, resulting in a frequency domain signal matrix. The signal in the frequency domain signal matrix can be represented as y ′ (l,r,k), r∈[1,N], where r refers to the index of the range cell. Then, a clutter suppression algorithm can be used to suppress clutter in the range dimension of the signal in the frequency domain signal matrix, resulting in a clutter-suppressed multidimensional signal. The signal matrix to be used. Optionally, the clutter suppression algorithm can be a moving average algorithm, a subtractive average algorithm, etc., but it is not limited to these algorithms.

[0043] It should be noted that different radar systems and application scenarios may require different clutter suppression methods. In practical applications, the appropriate method can be selected according to specific needs and conditions, and corresponding optimizations and adjustments can be made.

[0044] S130. Using the current moment as the end time of the first preset historical duration, determine the body movement index of the target object at the current moment based on the signal matrix to be used at each moment within the first preset historical duration.

[0045] The historical preset first duration can be a pre-defined duration, but the time period corresponding to the historical preset first duration will dynamically change as the current time changes. The body movement index can be used to reflect an individual's body movement in a static state. For example, in a sleep scenario, the body movement index can reflect the degree of change in a user's sleeping posture; a higher body movement index indicates a more active sleeping posture, while a lower body movement index indicates a more static sleeping posture.

[0046] In this embodiment, the current time can be used as the end time of a historically preset first duration. Based on the historically preset first duration, the starting time is calculated backward from the current time. That is, the historically preset first duration includes the current time and the times before the current time. For example, if the historically preset first duration is 10s and the current time is 20s, if 20s is the end time, then 9s is calculated backward from 20s, and the time period from 11s to 20s can be used as the period of the historically preset first duration. To improve calculation efficiency, a sliding window approach can also be used, with the historically preset first duration as the length of each sliding window. When determining the body movement index of the target object at the current time, the current time is used as the end time of the sliding window to obtain each time within the historically preset first duration. This allows the body movement index of the target object at the current time to be determined based on the signal matrix to be used at each time within the historically preset first duration. When determining the body movement index of the target object at the next time after the current time, the sliding window slides forward one time, and the time after the current time is used as the end time of the sliding window to obtain each time within the next historically preset first duration. It should be noted that the method for determining the signal matrix to be used at each moment within the historical preset first time period is the same as the method for determining the signal matrix to be used at the current moment in step S120, and will not be repeated here.

[0047] S140. If the body motion index meets the preset conditions, generate the spectrum to be used based on the signal matrix to be used at the current time.

[0048] The preset conditions can be used to determine whether a signal is stationary or in motion. These preset conditions include a decision threshold parameter, which can be a threshold or range used to distinguish and determine the signal state; the threshold or range is the threshold value of the decision threshold parameter. The spectra to be used include a first spectrum and a second spectrum. The first spectrum can be a range-azimuth spectrum, which includes the relative distance and azimuth information between the target and the radar. The second spectrum can be a range-Doppler spectrum, which includes the relative distance and velocity information between the target and the radar.

[0049] In this embodiment, preset conditions can be configured in advance. For example, the preset condition can be that when the body movement index is less than the threshold value of the decision threshold parameter, the body movement index is considered to meet the preset condition; when the body movement index is not less than the threshold value of the decision threshold parameter, the body movement index is considered not to meet the preset condition. The threshold value of the decision threshold parameter can be preset. For example, one or more threshold values ​​corresponding to the decision threshold parameter can be set according to the characteristics of the signal, such as amplitude, frequency, or phase. The body movement index of the target object at the current moment is compared with the preset parameter value. When the body movement index is less than the parameter value, it is determined that the body movement index meets the preset condition, and the target object is considered to be in a relatively static state at this time; if the body movement index is not less than the parameter value, it is determined that the body movement index does not meet the preset condition, and the target object is considered to be in a moving state, and a change in sleeping posture has occurred.

[0050] In this embodiment, the threshold value of the decision threshold parameter can also be dynamically determined. Considering the differences between different individuals, the body movement index in a static state also varies among individuals. Therefore, the threshold value of the decision threshold parameter can be dynamically determined by combining the individual's own body movement index. Optionally, determining that the body movement index meets the preset conditions includes: using the current time as the end time of the historical preset second duration, determining the threshold value of the decision threshold parameter at the current time based on the body movement index corresponding to each time point within the historical preset second duration; if the body movement index does not exceed the threshold value, then the body movement index is determined to meet the preset conditions. The historical preset second duration can be a preset duration, but the time period corresponding to the historical preset second duration will dynamically change with the current time.

[0051] In this embodiment, the current time can be used as the end time of the historical preset second duration, and the start time can be calculated backward from the current time based on the historical preset second duration. To improve calculation efficiency, a sliding window approach can also be used, with the historical preset second duration as the length of each sliding window. When determining the threshold value of the decision threshold parameter at the current time, the current time is used as the end time of the sliding window, obtaining each time within the historical preset second duration. This allows the threshold value of the decision threshold parameter at the current time to be determined based on the body movement index at each time within the historical preset second duration. It should be noted that the method for determining the body movement index at each time is the same as the method for determining the body movement index at the current time in step S130, and will not be repeated here.

[0052] Furthermore, the motion indexes corresponding to each moment within the preset second time period can be summed to obtain an intermediate value. This intermediate value is then divided by the preset second time period, and the quotient is multiplied by a pre-set detection parameter to obtain the threshold value of the decision threshold parameter at the current moment. See also... Figure 2 By comparing the body movement index (BMI) with a threshold value, suspected changes in sleep posture can be detected. If the BMI does not exceed the threshold value, it is determined that the BMI meets the preset conditions, and the target object is in a relatively static state, i.e., a stable state. If the BMI does not exceed the threshold value, it is determined that the BMI does not meet the preset conditions, and a suspected change in sleep posture is detected. If the BMI at the current moment meets the preset conditions, a multi-dimensional imaging of the stable sleep posture can be performed using the signal matrix at the current moment to generate a spectrum. The corresponding sleep posture recognition result can then be identified based on the spectrum.

[0053] For example, determine the decision threshold parameter at the current time. The threshold value below It can be represented as: Where δ is the detection coefficient, and T2 represents the historical preset second duration. τ2 is the starting time within T2; Let T2 be the cutoff time, i.e., the current time; M(t) is the body motion index at time t. If the body motion index at the current time... Greater than or equal to the decision threshold If the body movement index does not meet the preset conditions, it is considered that the body has changed its sleeping posture; otherwise, it is considered that the body is in a relatively static state, that is, the body movement index meets the preset conditions.

[0054] The technical solution provided in this embodiment dynamically determines the threshold value of the decision threshold parameter by combining the body movement index of the target object at the current moment and before the current moment, thereby improving the accuracy of body movement state detection, filtering out signals that do not meet the preset conditions, that is, filtering out signals that indicate changes in the sleeping posture of the human body, and successfully separating the sleep time of the human body in a static state, so as to improve the accuracy of sleep posture detection and improve detection efficiency for multi-dimensional imaging of the sleeping posture in a static state.

[0055] Optionally, the method for generating the spectrum to be used based on the signal matrix to be used at the current time can be: performing angle estimation on the signal matrix to be used to obtain a first spectrum; performing velocity estimation on the signal matrix to be used to obtain a second spectrum; and using the first spectrum and the second spectrum as the spectrum to be used.

[0056] Specifically, after isolating the sleep moment of a human body in a static state, multi-dimensional imaging can be performed on the signal matrix to be used at this moment. For example, angle measurement methods can be used to estimate the angle of the signals in the signal matrix to obtain the first spectrum; velocity measurement methods can be used to estimate the velocity of the signals in the signal matrix to obtain the second spectrum. Optionally, angle measurement methods can be Angle-FFT, Capon algorithm, and MUSIC algorithm, etc., for the clutter-suppressed signal. Two-dimensional angle estimation is performed to obtain a range-azimuth spectrum M1[r,θ] as the first spectrum, where θ∈[-Θ,Θ], θ represents the azimuth angle value, and Θ represents the maximum azimuth angle range. Velocity measurement can be performed using the Doppler-FFT (Doppler-Fast Fourier Transform) algorithm, in the slow time dimension... A Doppler-FFT is performed to obtain a distance-Doppler spectrum M2[r,υ] as the second spectrum, where υ∈[-Υ,Υ], υ represents the Doppler frequency, and Υ represents the maximum Doppler frequency range. The technical solution provided in this embodiment improves the accuracy of posture recognition by performing multi-dimensional imaging of the signal matrix to be used by the human body in a static state, enabling posture recognition through the multi-dimensional imaging spectrum.

[0057] S150. Based on the spectrum to be used, determine the recumbent posture recognition result corresponding to the target object.

[0058] In this embodiment, features can be extracted from the spectrum to be used, such as shape, texture, frequency distribution, etc. The lying posture features of the target object at the current time can be extracted, and then the lying posture of the target object can be estimated based on the lying posture features to obtain the lying posture recognition result.

[0059] The technical solution of this embodiment processes the radar echo signal matrix reflected by the target object at the current moment to obtain the signal matrix to be used. Then, by combining the signal matrix to be used at the current moment and the historical moments before the current moment, the body movement index of the target object at the current moment is determined. This enables the body movement index to accurately reflect the changes in the sleeping posture of the target object. The body movement index that does not meet the preset conditions is filtered out, and the moment when the body movement index meets the preset conditions is extracted. This moment is when the target object is in a relatively static state. The signal matrix to be used at the moment when the target object is in a relatively static state is used to generate the spectrum to be used, and the posture recognition result corresponding to the target object is determined. This achieves high recognition efficiency while improving the accuracy of sleep posture recognition, thus meeting the user's needs.

[0060] Example 2

[0061] Figure 3 This is a flowchart of a sleep posture recognition method according to Embodiment 2 of the present invention. Based on the foregoing embodiments, the signal matrix to be used includes signal matrices under at least two antenna channels, and S130 can be further refined. For specific implementation details, please refer to the technical solution of this embodiment. Technical terms that are the same as or corresponding to those in the above embodiments will not be repeated here.

[0062] like Figure 3 As shown, the method specifically includes the following steps:

[0063] S210. Obtain the radar echo signal matrix reflected by the target object at at least one moment.

[0064] S220. For each time moment, process the radar echo signal matrix at the current time moment to obtain the signal matrix to be used.

[0065] S230, Use the current time as the end time of the first preset historical duration.

[0066] S240. The signal matrices of at least two antenna channels to be used at the same time are fused to obtain the signal sequence to be processed at the corresponding time.

[0067] In this embodiment, various forms of interference in the space environment, such as electronic interference and natural interference, are considered. These interferences affect radar signal quality, leading to low identification accuracy. To effectively solve this problem and achieve accurate multi-target tracking, radar echo signals are received through multiple antenna channels. That is, the radar echo signal matrix reflected from the target object at the current moment can be obtained from multiple antenna channels. By processing the radar echo signal matrix of each antenna channel separately, multiple signal matrices from different antenna channels can be obtained. This allows for the merging of signals received from multiple antennas, which not only increases the strength of the effective signal and reduces noise interference, improving the signal-to-noise ratio, but also improves the angular and range resolution of the radar system. This results in more accurate target location and identification, enhancing the radar's detection performance and anti-interference capabilities. Furthermore, when some antenna channels fail, other normally functioning antenna channels can continue to receive and process signals, ensuring the normal operation of the radar system.

[0068] Specifically, the signal matrices of at least two antenna channels at the same time can be averaged across channels and fused into a single signal sequence for that time. For example, signals from the same range cell index under the same frequency-modulated continuous wave index in the three channels corresponding to time 1 can be fused together. After averaging, we get As the signal within the distance unit under the frequency-modulated continuous wave, this signal is the signal in the signal sequence to be processed. Accordingly, the signal sequence to be processed corresponding to each moment within the preset first time period is obtained.

[0069] S250. Perform incoherent accumulation processing on the signal sequence to be processed to obtain the signal sequence to be applied.

[0070] In this embodiment, a suitable accumulation factor can be selected to perform incoherent accumulation processing on each signal in the signal sequence to be processed, and the signal sequence formed by the signals after incoherent accumulation processing is used as the signal sequence to be applied.

[0071] S260. Process the signal sequence to be applied to obtain the body trunk energy value.

[0072] Among them, the body trunk energy value can refer to the energy or strength possessed by the body trunk (such as the trunk including the chest, back, waist, abdomen, etc.).

[0073] In this embodiment, the effective range within the space of the radar-transmitted signal can be identified. For example, the distance units containing consecutive signal energy values ​​above a preset threshold in the signal sequence to be applied can be used as the distance units within the effective range to obtain the effective range; or, the area where the object carrying the target (such as a bed) is located can be used as the effective range. All signal energy values ​​within the effective range in the signal sequence to be applied are accumulated to obtain the body trunk energy value at that moment. Correspondingly, the body trunk energy value at each moment within a historical preset first time period can be obtained.

[0074] S270. Based on the body trunk energy value at each moment within the historical preset first time period, determine the body movement index of the target object at the current moment.

[0075] In this embodiment, the body movement index of the target object at the current moment can be determined based on the energy values ​​of each body trunk. This current moment is the cutoff time within a preset first time period. For example, it can be used for multidimensional signals after clutter suppression. The average signal sequence is averaged between channels, and an appropriate accumulation factor α (e.g., 0 < α < 1) is selected to perform incoherent accumulation on the averaged signal sequence. The incoherently accumulated signal sequence is then accumulated within an effective distance to obtain the body trunk energy value at the current moment. By combining the body trunk energy values ​​at the current moment and those prior to the current moment, the body movement index of the target object at the current moment is determined.

[0076] It should be noted that, considering the relatively static state of the human body during sleep, changes in sleeping posture are only possible during significant movements such as turning over or getting up at night. Therefore, to effectively identify sleeping postures, the signals from the static and dynamic states during sleep can be separated. By extracting the signals from the static state, the separation between sleeping postures can be achieved, reducing the computational load for event recognition while improving recognition accuracy.

[0077] In this embodiment, the body movement index of the target object at the current moment is determined based on the body trunk energy value at each moment within a historical preset first duration. This includes: determining the minimum body trunk energy value among all moments within the historical preset first duration; determining the difference between each body trunk energy value and the minimum body trunk energy value; and determining the body movement index of the target object at the current moment based on each difference and the historical preset first duration.

[0078] Specifically, the minimum body trunk energy value can be selected from the body trunk energy values ​​corresponding to each moment within a preset first time period. The difference between each body trunk energy value and the minimum body trunk energy value is calculated, and all differences are summed to obtain a sum. This sum is then divided by the preset first time period, and the quotient can be used as the target object's body movement index at the current moment. For example, the target object at the current moment... The lower body movement index It can be represented as:

[0079] Among them, T1 is the historical preset first duration; τ1 is the starting time within T1; R(t) represents the cutoff time within T1, i.e., the current time; R(t) represents the body trunk energy value at time t; R(t) min This represents the minimum value of R(t) within T1, i.e., the minimum energy value of the body trunk.

[0080] S280. If the body motion index meets the preset conditions, generate the spectrum to be used based on the signal matrix to be used at the current time.

[0081] S290. Based on the spectrum to be used, determine the recumbent posture recognition result corresponding to the target object.

[0082] The technical solution provided in this embodiment improves signal quality and reduces computational complexity by performing a noncoherent linear transformation on the signal fused with multiple antenna channels through an accumulation factor. This increases processing speed. At the same time, by combining the energy values ​​of each body part of the target object within the effective distance, the body movement index of the target object at the current moment is determined, thus ensuring the accuracy of body movement detection. This ensures that recumbent posture recognition can be performed when the target object is in a relatively static state, thereby improving the accuracy of recumbent posture recognition.

[0083] Example 3

[0084] Figure 4 This is a flowchart of a sleep posture recognition method according to Embodiment 3 of the present invention. Based on the foregoing embodiments, S150 is further refined. For specific implementation details, please refer to the technical solution of this embodiment. Technical terms that are the same as or corresponding to those in the above embodiments will not be repeated here.

[0085] like Figure 4 As shown, the method specifically includes the following steps:

[0086] S310. Obtain at least one radar echo signal matrix reflected by the target object at any given time.

[0087] S320. For each time moment, process the radar echo signal matrix at the current time moment to obtain the signal matrix to be used.

[0088] S330. Using the current moment as the end time of the first preset historical duration, determine the body movement index of the target object at the current moment based on the signal matrix to be used at each moment within the first preset historical duration.

[0089] S340. If the body motion index meets the preset conditions, generate the spectrum to be used based on the signal matrix to be used at the current time.

[0090] S350. Extract spatial morphological features from the first spectrum to obtain the recumbent spatial morphological features under each morphological index.

[0091] The morphological indicators include at least one of the following: bed fit, lateral positioning measurement, and overall body tilt angle. Bed fit refers to the degree of fit between the bed and the human body during sleep. Lateral positioning measurement refers to certain measurements of the human body when lying on one's side, such as height, weight, shoulder width, and hip width. Overall body tilt angle refers to the overall tilt angle of the human body relative to the horizontal plane or the direction of gravity.

[0092] It's important to note that different body postures during sleep present different spatial forms. For example, when lying still, breathing causes subtle movements of the chest, abdomen, arms, and neck. These subtle movements share some commonalities among different individuals. For instance, when lying on one's side (including left and right sides), the body is less in contact with the bed, and chest and abdominal movements are affected by lateral pressure. The neck and arms are also more pronounced in this position compared to non-side-lying positions. When not lying on one's side (including supine and prone positions), the body is in close contact with the bed, and the area and range of motion are larger and more pronounced than in side-lying positions. While non-side-lying positions share some common characteristics, significant differences also exist. For example, in supine positions, the expansion and contraction of the lungs result in significant and large-amplitude breathing movements of the chest and abdomen; while in prone positions, the movement of the chest and abdomen is compressed by the bed surface, and breathing causes less movement of the entire back. (Continue to see...) Figure 2 The spatial morphological features of human sleeping posture are extracted from the first spectrum, and the spatial morphological features of sleeping posture under different morphological indicators are extracted to enable sleep posture recognition through the spatial morphological features of sleeping posture under different morphological indicators.

[0093] Optionally, the method for extracting spatial morphological features from the first spectrum to obtain the supine spatial morphological features of the bed frame fit can be as follows: The first spectrum is extracted using a bed frame fit calculation formula to determine the supine spatial morphological features of the bed frame fit; wherein, the bed frame fit calculation formula can be expressed as:

[0094]

[0095] Where M1[r,θ] represents the first spectrum, F1 represents the supine spatial morphological characteristics of bed body fit, θ∈[-Θ,Θ], θ represents the azimuth angle value, Θ represents the maximum azimuth angle range, and r represents the distance.

[0096] Optionally, the method for extracting spatial morphological features from the first spectrogram to obtain the recumbent spatial morphological features in lateral recumbency measurement can be as follows: The first spectrogram is extracted using a lateral recumbency measurement calculation formula to determine the recumbent spatial morphological features in lateral recumbency measurement; wherein, the lateral recumbency measurement calculation formula can be expressed as:

[0097]

[0098] Where G1 represents the contour sequence extracted from the first spectrogram M1[r,θ], and |·| represents the length of the sequence.

[0099] M1[r,θ] represents a non-zero logical matrix, and the angle sequence is...

[0100] Optionally, the method for extracting spatial morphological features from the first spectrogram to obtain the supine spatial morphological features of the global body tilt angle can be: using the global body tilt angle calculation formula to extract spatial morphological features from the first spectrogram to determine the supine spatial morphological features of the global body tilt angle; wherein, the global body tilt angle calculation formula can be expressed as:

[0101] F4 = (index) max -G2(1))-(G2(N)-index max )

[0102]

[0103] Where index max =argmax(G2(n)) represents the maximum index of the angle sequence, x represents the x-coordinate sequence of the major axis of the spectrum, y represents the y-coordinate sequence of the major axis of the spectrum, cov(x,x) represents the variance of x, and cov(x,y) represents the covariance of x and y.

[0104] S360. Perform Doppler feature extraction on the second spectrum to obtain the micro-Doppler features for each local index.

[0105] The local indicators include at least one of the neck, chest, abdomen, and head. It should be noted that during sleep, human vital activities exhibit differences in the amplitude of movement in different parts of the body, which are reflected in the second spectrum. Therefore, the second spectrum can be used to extract local micro-Doppler features of human vital activities during sleep, extracting micro-Doppler features under different local indicators, enabling sleep posture recognition based on these micro-Doppler features.

[0106] Optionally, the Doppler feature extraction of the second spectrum to obtain the micro-Doppler features of the neck can be performed by: using the neck micro-Doppler feature calculation formula to extract the Doppler features of the second spectrum to obtain the neck micro-Doppler features; wherein, the neck micro-Doppler feature calculation formula can be expressed as:

[0107]

[0108] Where M2[r,υ] represents the second spectrum, G4(r)=∑ υ M2[r,υ], r=1,2,…,R;υ∈[-Υ,Υ],υ represents the Doppler frequency,Υ represents the maximum Doppler frequency range; R2 and R3 represent the second and third non-zero distance units in the second spectrum, respectively.

[0109] Optionally, the Doppler feature extraction of the second spectrum to obtain the micro-Doppler features of the chest and abdomen intensity can be performed by: using the formula for calculating the micro-Doppler features of the chest and abdomen intensity to extract the Doppler features of the second spectrum; wherein, the formula for calculating the micro-Doppler features of the chest and abdomen intensity can be expressed as:

[0110]

[0111] in, and These represent the second and third peak values ​​of sequence G4(r), respectively.

[0112] Optionally, the method for extracting Doppler features from the second spectrum to obtain the head micro-Doppler features can be as follows: The second spectrum can be extracted using a formula for calculating head micro-Doppler features; whereby the formula for calculating head micro-Doppler features can be expressed as:

[0113] in, Let M2[r,υ] be a non-zero logical matrix, and G be the non-zero logical matrix. 5m This represents the maximum value of G5, ε is the decision threshold, and ||·|0 represents the calculation of the 0-norm.

[0114] G 6m This represents the maximum value of G6.

[0115] It should be noted that S350 to S360 can be executed sequentially or in parallel. The specific execution order is not limited. The above order is only the order in which the technical solutions in each step are explained, not the execution order of each step.

[0116] S370. Based on the spatial morphological features of the lying posture and / or micro-Doppler features, determine the lying posture recognition result corresponding to the target object.

[0117] In this embodiment, all supine spatial morphological features can be used to determine the supine posture recognition result corresponding to the target object; or, all micro-Doppler features can be used to determine the supine posture recognition result corresponding to the target object. Alternatively, see [link to relevant documentation]. Figure 2 By fusing spatial morphological features and micro-Doppler features of the supine posture, the corresponding supine posture recognition result is determined based on the feature information of the fused two types of features. To ensure the accuracy of recognition, a secondary detection can be performed on the supine posture recognition results of the target object at at least two time points to determine the final supine posture recognition result. The body movement index of the target object at these at least two time points must meet preset conditions. For example, the supine posture recognition results in the first T seconds before the body movement index meets the preset conditions can be statistically analyzed, and the supine posture recognition result with the largest proportion can be taken as the final supine posture recognition result of the target object.

[0118] Optionally, the sleeping posture recognition result can be one of the following: left lateral, right lateral, or non-lateral sleeping posture classification results. In this embodiment, left lateral, right lateral, and non-lateral classification is a three-class classification problem. A bootstrap aggregation (Bagging) algorithm can be used, where labels are added to features including spatial morphological features and micro-Doppler features, which are then used as input data for model training to obtain a trained classification model. Furthermore, the spatial morphological features and micro-Doppler features of the target object at the current moment can be input into the trained classification model to obtain the sleeping posture recognition result corresponding to the target object. The advantage of using the bootstrap aggregation algorithm is that it helps reduce the model's variance, thereby improving the overall generalization ability. Its basic strategy is to generate multiple subsets by randomly sampling the training dataset with replacement, then train multiple independent basic models, and finally combine the predictions of these models. By leveraging the advantages of feature fusion, the model can be trained using input data containing two types of features with complementary information, which can improve the accuracy of model recognition. The final trained model can then be exported for left lateral, right lateral, and non-lateral sleeping posture recognition.

[0119] It should be noted that after completing the classification and recognition of left lateral, right lateral, and non-lateral sleeping positions, a secondary classification is required for non-lateral sleeping positions to determine whether they are supine or prone. Considering the differences between supine and prone sleeping positions, which are reflected in various indicators such as the overall image contour and energy intensity and micro-Doppler range of each part in the first and second spectra, multi-dimensional feature sequences can be extracted from the first and second spectra for supine and prone sleeping position classification and recognition, ensuring the accuracy of position recognition.

[0120] Based on the above technical solutions, the sleep posture recognition method further includes: when the posture recognition result is not side-lying, performing feature extraction on the spectrum to be used to obtain the feature sequence under each supine / prone evaluation index; processing the feature sequence based on a long short-term memory network model to obtain the target posture of the target object when it is not side-lying.

[0121] The evaluation metrics for supine and prone positions include at least one of contour, energy intensity distribution, and Doppler width; the target lying position includes supine or prone. It should be noted that supine and prone identification is a binary classification problem, and the extracted features are in sequence form. Therefore, a Long Short-Term Memory (LSTM) network model is used to complete this classification task. Its basic strategy is to effectively capture and process long-term dependencies by introducing memory units, gating mechanisms, and time-step unrolling techniques, making it a model suitable for sequence data. Specifically, the extracted multidimensional image contour sequence and intensity distribution sequence feature set S can be labeled and used as input data. The LSTM model is trained using the input data, and the trained LSTM network model is exported for classification and recognition of non-lateral lying events.

[0122] See also Figure 2 The system can determine whether the sleeping posture recognition result is non-lateral. If the result is non-lateral, multi-dimensional image feature sequences are extracted from the spectrum to be used, obtaining feature sequences for each supine / prone evaluation index. For example, the extracted feature sequences include the contour sequence, intensity distribution sequence, and Doppler width sequence of each spectrum. The contour sequence describes the lateral contour width of each body part in the image, the intensity distribution sequence describes the Doppler intensity and energy amplitude of each body part, and the Doppler width sequence describes the width of the Doppler range occupied by each body part. Furthermore, all feature sequences can be processed based on a pre-trained long short-term memory network model to obtain the target sleeping posture of the object in a non-lateral position. To ensure recognition accuracy, secondary detection can be performed based on the target sleeping posture of the object at at least two time points to determine the final target sleeping posture. For example, the target sleeping postures in the T seconds before the body movement index meets preset conditions can be statistically analyzed, and the target sleeping posture with the largest proportion can be taken as the final target sleeping posture of the object, achieving classification and recognition of four sleeping postures: supine, prone, left lateral, and right lateral.

[0123] The technical solution of this embodiment extracts spatial morphological features from the first spectrum to obtain the spatial morphological features of the lying posture under each morphological index. At the same time, it extracts Doppler features from the second spectrum to obtain the micro-Doppler features under each local index. By extracting sleep lying posture features from the spatial morphology and local sleep vital activities, the richness of sleep lying posture features is improved, while the accuracy of sleep lying posture recognition is also improved.

[0124] Example 4

[0125] Figure 5 This is a schematic diagram of a sleep posture recognition device according to Embodiment 4 of the present invention. Figure 5 As shown, the device includes: a signal acquisition module 410, a signal matrix determination module 420, a body movement index determination module 430, a spectrum determination module 440, and a posture recognition result determination module 450.

[0126] The system includes a signal acquisition module 410 for acquiring at least one radar echo signal matrix reflected by the target object at any given time; a signal matrix determination module 420 for processing the radar echo signal matrix at the current time to obtain a signal matrix to be used; a body movement index determination module 430 for determining the body movement index of the target object at the current time, using the current time as the cutoff time of a historical preset first time period, based on the signal matrix to be used at each time within the historical preset first time period; a spectrum determination module 440 for generating a spectrum to be used based on the signal matrix to be used at the current time, provided that the body movement index meets preset conditions; wherein the spectrum to be used includes a first spectrum and a second spectrum; the first spectrum includes the relative distance and azimuth information between the target object and the radar; the second spectrum includes the relative distance and velocity information between the target object and the radar; and a prone position recognition result determination module 450 for determining the prone position recognition result corresponding to the target object based on the spectrum to be used.

[0127] The technical solution of this embodiment processes the radar echo signal matrix reflected by the target object at the current moment to obtain a signal matrix to be used. Then, by combining the signal matrices to be used at the current moment and historical moments before the current moment, the body movement index of the target object at the current moment is determined. This enables the body movement index to accurately reflect the changes in the target object's sleeping posture. Objects with body movement indices that do not meet preset conditions are filtered out, and the moments when the body movement index meets the preset conditions are extracted. These moments represent the moments when the target object is in a relatively static state. Based on the signal matrix to be used at the moments when the target object is in a relatively static state, a spectrum to be used is generated to determine the target object's sleeping posture recognition result. This achieves high efficiency in recognition while improving the accuracy of sleeping posture recognition, thus meeting user needs. It solves the problem of low accuracy in sleeping posture recognition caused by analyzing radar signals to determine object speed, angle, and other information for posture recognition.

[0128] Optionally, the signal matrix determination module 420 includes: a discrete echo signal matrix determination unit, used to preprocess the radar echo signal matrix at the current time to obtain a discrete echo signal matrix; a frequency domain signal matrix determination unit, used to perform fast Fourier transform processing on the discrete echo signal matrix to obtain a frequency domain signal matrix; and a signal matrix determination unit, used to perform clutter suppression processing on the frequency domain signal matrix to obtain a signal matrix to be used.

[0129] Optionally, the signal matrix to be used includes signal matrices of at least two antenna channels. The body movement index determination module 430 includes: a signal sequence determination unit for fusing the signal matrices of at least two antenna channels at the same time to obtain a signal sequence to be processed at the corresponding time; a signal sequence determination unit for performing incoherent accumulation processing on the signal sequence to be processed to obtain a signal sequence to be applied; a body trunk energy value determination unit for processing the signal sequence to be applied to obtain a body trunk energy value; and a body movement index determination unit for determining the body movement index of the target object at the current time based on the body trunk energy value at each time within the historical preset first time period.

[0130] Based on the above-mentioned device, optionally, the body movement index determination unit includes: a minimum body trunk energy value determination unit, used to determine the minimum body trunk energy value in all moments within the historical preset first duration; a difference determination unit, used to determine the difference between each of the body trunk energy values ​​and the minimum body trunk energy value; and a body movement index determination unit, used to determine the body movement index of the target object at the current moment based on each of the differences and the historical preset first duration.

[0131] Optionally, based on the above-mentioned device, the preset conditions include a decision threshold parameter. The spectrum determination module 440 includes: a threshold value determination unit, used to determine the threshold value of the decision threshold parameter at the current time based on the body movement index corresponding to each moment within the historical preset second duration, using the current time as the cutoff time of the historical preset second duration; and a condition judgment unit, used to determine that the body movement index meets the preset conditions if the body movement index does not exceed the threshold value.

[0132] Optionally, the spectrum determination module 440 further includes a first spectrum determination unit, a second spectrum determination unit, and a spectrum to be used determination unit. The first spectrum determination unit is used to perform angle estimation on the signal matrix to be used at the current time to obtain a first spectrum; the second spectrum determination unit is used to perform velocity estimation on the signal matrix to be used at the current time to obtain a second spectrum; and the spectrum to be used determination unit is used to use the first spectrum and the second spectrum as the spectrum to be used.

[0133] Based on the above-mentioned device, optionally, the posture recognition result determination module 450 includes: a posture spatial morphology feature determination unit, used to extract spatial morphology features from the first spectral image to obtain posture spatial morphology features under each morphological index; the morphological index includes at least one of bed fit, lateral lying measurement, and global body tilt angle; a micro-Doppler feature determination unit, used to extract Doppler features from the second spectral image to obtain micro-Doppler features under each local index; the local index includes at least one of neck, chest, abdomen, and head; and a posture recognition result determination unit, used to determine the posture recognition result corresponding to the target object based on the posture spatial morphology features and / or the micro-Doppler features.

[0134] Optionally, based on the above-mentioned device, the device further includes: a feature sequence determination unit, used to extract features from the spectrum to be used when the recumbent posture recognition result is non-lateral recumbent, to obtain a feature sequence under each pitch evaluation index; wherein the pitch evaluation index includes at least one of contour, energy intensity distribution and Doppler width; and a target recumbent posture determination unit, used to process the feature sequence based on a long short-term memory network model to obtain the target recumbent posture of the target object when it is not lateral recumbent; wherein the target recumbent posture includes supine or prone.

[0135] The sleep posture recognition device provided in the embodiments of the present invention can execute the sleep posture recognition method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0136] Example 5

[0137] Figure 6This is a schematic diagram of the structure of an electronic device implementing the sleep posture recognition method of this invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0138] like Figure 6 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer programs stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14. Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard or mouse; an output unit 17, such as various types of displays or speakers; a storage unit 18, such as a disk or optical disk; and a communication unit 19, such as a network card, modem, or wireless transceiver. Communication unit 19 allows electronic device 10 to exchange information / data with other devices via computer networks such as the Internet and / or various telecommunications networks. Processor 11 can be a variety of general-purpose and / or dedicated processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as sleep posture recognition methods.

[0139] In some embodiments, the sleep posture recognition method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the sleep posture recognition method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the sleep posture recognition method by any other suitable means (e.g., by means of firmware).

[0140] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0141] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, partially on a machine and partially on a remote machine as a standalone software package, or entirely on a remote machine or server. In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. Computer-readable storage media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0142] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user; and a keyboard and pointing device through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback; and input from the user can be received in any form. The systems and techniques described herein can be implemented in computing systems that include back-end components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include front-end components, or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected via digital data communication (e.g., a communication network) of any form or medium. Examples of communication networks include: local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet. The computing system can include clients and servers. Clients and servers are generally geographically distant from each other and typically interact via a communication network. Client-server relationships are created by computer programs running on respective computers and having client-server relationships with each other. A server can be a cloud server, also known as a cloud computing server or cloud host. It is a host product in the cloud computing service system, which solves the problems of high management difficulty and weak business scalability in traditional physical hosts and VPS services.

[0143] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0144] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for recognizing sleep posture, characterized in that, include: Obtain the radar echo signal matrix reflected by the target object at at least one moment; For each time moment, the radar echo signal matrix at the current time is processed to obtain the signal matrix to be used; Using the current moment as the cutoff moment of a historical preset first duration, and based on the signal matrix to be used at each moment within the historical preset first duration, the body movement index of the target object at the current moment is determined, including: the signal matrix to be used includes a signal matrix to be used from at least two antenna channels; the signal matrices to be used from at least two antenna channels at the same moment are fused to obtain a signal sequence to be processed at the corresponding moment; the signal sequence to be processed is subjected to incoherent accumulation processing to obtain a signal sequence to be applied; the signal sequence to be applied is processed to obtain a body trunk energy value; and the body movement index of the target object at the current moment is determined based on the body trunk energy value at each moment within the historical preset first duration. The body trunk energy value refers to the energy possessed by the body trunk. By identifying the effective distance within the space of the radar transmission signal, the energy values ​​of all signals within the effective distance in the signal sequence to be applied are accumulated to obtain the body trunk energy value at that moment. When the body motion index meets preset conditions, a spectrum to be used is generated based on the signal matrix to be used at the current moment; wherein, the spectrum to be used includes a first spectrum and a second spectrum; the first spectrum includes the relative distance and azimuth information between the target object and the radar; the second spectrum includes the relative distance and velocity information between the target object and the radar; Based on the spectrum to be used, the method for determining the recumbent posture recognition result corresponding to the target object includes: extracting spatial morphological features from the first spectrum to obtain recumbent spatial morphological features under each morphological index; the morphological index includes at least one of bed fit, lateral recumbent measure, and global body tilt angle; extracting Doppler features from the second spectrum to obtain micro-Doppler features under each local index; the local index includes at least one of neck, chest, abdomen, and head; and fusing the recumbent spatial morphological features and micro-Doppler features, and determining the recumbent posture recognition result corresponding to the target object based on the feature information of the fused two types of features.

2. The method according to claim 1, characterized in that, The process of processing the radar echo signal matrix at the current moment to obtain the signal matrix to be used includes: The radar echo signal matrix at the current moment is preprocessed to obtain the discrete echo signal matrix; The discrete echo signal matrix is ​​subjected to a fast Fourier transform to obtain a frequency domain signal matrix; The frequency domain signal matrix is ​​subjected to clutter suppression processing to obtain the signal matrix to be used.

3. The method according to claim 1, characterized in that, The determination of the target object's body movement index at the current moment based on the body trunk energy value at each moment within the historical preset first time period includes: Determine the minimum body trunk energy value at all times within the historical preset first duration; The difference between each of the stated body trunk energy values ​​and the minimum body trunk energy value is determined respectively; Based on the differences and the historical preset first duration, the body movement index of the target object at the current moment is determined.

4. The method according to claim 1, characterized in that, The preset conditions include a decision threshold parameter, which determines whether the body movement index meets the preset conditions, including: Using the current moment as the end point of the historical preset second duration, and based on the body movement index corresponding to each moment within the historical preset second duration, the threshold value of the decision threshold parameter at the current moment is determined; If the body movement index does not exceed the threshold value, then the body movement index is determined to meet the preset conditions.

5. The method according to claim 1, characterized in that, The process of generating the spectrum to be used based on the signal matrix to be used at the current time includes: Angle estimation is performed on the signal matrix to be used at the current moment to obtain the first spectrum; The velocity of the signal matrix to be used at the current moment is estimated to obtain the second spectrum. The first and second spectra are used as the spectra to be used.

6. The method according to claim 1, characterized in that, Also includes: When the recumbent posture recognition result is not lateral, feature extraction is performed on the spectrum to be used to obtain the feature sequence under each pitch evaluation index; wherein, the pitch evaluation index includes at least one of contour, energy intensity distribution and Doppler width; The feature sequence is processed based on a long short-term memory network model to obtain the target lying posture of the target object in the non-lateral lying position; wherein, the target lying posture includes supine or prone.

7. A sleep posture recognition device, characterized in that, include: The signal acquisition module is used to acquire the radar echo signal matrix reflected by the target object at at least one moment; The module for determining the signal matrix to be used is used to process the radar echo signal matrix at the current time for each time moment to obtain the signal matrix to be used. The body movement index determination module is used to determine the body movement index of the target object at the current time, based on the signal matrix to be used at each time within the historical preset first time, with the current time as the cutoff time of the historical preset first time. The signal matrix to be used includes signal matrices for at least two antenna channels. The body movement index determination module is specifically used to fuse the signal matrices for at least two antenna channels at the same time to obtain a signal sequence to be processed at the corresponding time; to perform incoherent accumulation processing on the signal sequence to be processed to obtain a signal sequence to be applied; to process the signal sequence to be applied to obtain a body trunk energy value; and to determine the body movement index of the target object at the current time based on the body trunk energy value at each time within the historical preset first time period. The body trunk energy value refers to the energy possessed by the body trunk. By identifying the effective distance within the space of the radar transmission signal, the energy values ​​of all signals within the effective distance in the signal sequence to be applied are accumulated to obtain the body trunk energy value at that moment. The spectrum determination module is used to generate a spectrum to be used based on the signal matrix to be used at the current moment, provided that the body motion index meets preset conditions; wherein, the spectrum to be used includes a first spectrum and a second spectrum; the first spectrum includes the relative distance and azimuth information between the target object and the radar; the second spectrum includes the relative distance and velocity information between the target object and the radar; The posture recognition result determination module is used to determine the posture recognition result corresponding to the target object based on the spectrum to be used; The posture recognition result determination module is specifically used to extract spatial morphological features from the first spectral image to obtain the posture spatial morphological features under each morphological index; the morphological index includes at least one of bed fit, lateral lying measurement, and global body tilt angle; to extract Doppler features from the second spectral image to obtain micro-Doppler features under each local index; the local index includes at least one of neck, chest, abdomen, and head; and to determine the posture recognition result corresponding to the target object based on the feature information of the fused two types of features by performing feature fusion on the posture spatial morphological features and micro-Doppler features.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the sleep posture recognition method according to any one of claims 1-6.

Images