Non-contact intelligent cardiopulmonary state evaluation system

Abstract

The invention discloses a non-contact intelligent cardiopulmonary state evaluation system which comprises a cardiopulmonary sign radar detector, a radar data gateway and a display control terminal. The cardiopulmonary sign radar detector is used for transmitting an ultra-wideband pulse to a to-be-evaluated training person and receiving an echo signal obtained by the to-be-evaluated training person according to the reflection of the ultra-wideband pulse; the radar data gateway is used for establishing a communication link between the cardiopulmonary sign radar detector and the display control terminal to form an internal communication network; and the display control terminal is used for receiving the echo signals of the to-be-evaluated training personnel in the preset sampling time through the radar data gateway, performing feature extraction and cardiopulmonary function analysis on the echo signals to obtain a cardiopulmonary stage recognition result and a cardiopulmonary rhythm control result, and displaying the cardiopulmonary stage recognition result and the cardiopulmonary rhythm control result. According to the system, non-contact intelligent cardiopulmonary state evaluation is realized, so that cardiopulmonary state recognition accuracy is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of cardiopulmonary status recognition, and in particular to a non-contact intelligent cardiopulmonary status assessment system. Background Art

[0002] Shooting is a static and agile endurance sport. This process requires the coordination of the shooter's nerves and organs. To maintain this neural reflex process, the body needs to be based on a stable physiological state. These physiological conditions include stable heart rate, stable breathing, and stable muscle memory.

[0003] At present, shooting training generally adopts a training mode in which the coach demonstrates and guides, and the shooters make massive contacts. The training evaluation feedback of this training mode is low in accuracy and efficiency. In addition, there are also contact measurement devices that monitor the shooter's heart rate and breathing status in real time during shooting, which are used to evaluate the shooter's psychological control ability and breathing rhythm, and thus serve as status feedback for static training such as shooting. However, current contact measurement equipment, such as electrocardiogram (ECG) sensors, blood oximeters, respiratory monitoring chest straps, smart wearables, etc., need to be worn by the test person to measure, which will affect the test person's actions in the shooting training scene, which is not conducive to the evaluation of the real cardiopulmonary status. Summary of the invention

[0004] In order to solve the above technical problems, an embodiment of the present invention provides a non-contact intelligent cardiopulmonary status assessment system, which realizes non-contact cardiopulmonary physiological characteristics detection and assessment of trainees, thereby ensuring the accuracy of the assessment results.

[0005] A first aspect of an embodiment of the present invention provides a non-contact intelligent cardiopulmonary status assessment system, the system comprising a cardiopulmonary vital sign radar detector, a radar data gateway and a display control terminal;

[0006] A cardiopulmonary sign radar detector, used for transmitting ultra-wideband pulses to the trainee to be evaluated, and receiving echo signals obtained by the trainee to be evaluated based on the reflection of the ultra-wideband pulses;

[0007] Radar data gateway, used to establish a communication link between the cardiopulmonary vital signs radar detector and the display control terminal to form an internal communication network;

[0008] The display control terminal is used to receive the echo signal of the training personnel to be evaluated within a preset sampling time through the radar data gateway, denoise the echo signal to obtain the denoised signal; extract features from the denoised information to obtain respiratory signals, heartbeat signals and micro-motion features, and use a deep learning neural network to process the respiratory signals and heartbeat signals to obtain fine-grained respiratory and heartbeat waveforms; perform cardiopulmonary function analysis on the fine-grained respiratory and heartbeat waveforms to obtain cardiopulmonary stage recognition results and cardiopulmonary rhythm control results, and display the cardiopulmonary stage recognition results and cardiopulmonary rhythm control results.

[0009] In a possible implementation of the first aspect, the display control terminal includes an intelligent cardiopulmonary status assessment device and a display device.

[0010] Among them, the intelligent cardiopulmonary status assessment device is used to receive the echo signal of the training personnel to be assessed within the preset sampling time through the radar data gateway, perform denoising on the echo signal to obtain the denoised signal; perform feature extraction on the denoised information to obtain the respiratory signal, heartbeat signal and micro-motion features, and use the deep learning neural network to process the respiratory signal and heartbeat signal to obtain fine-grained respiratory and heartbeat waveforms; perform cardiopulmonary function analysis on the fine-grained respiratory and heartbeat waveforms to obtain cardiopulmonary stage recognition results and cardiopulmonary rhythm control results;

[0011] The display device is used to display the cardiopulmonary stage recognition results and the cardiopulmonary rhythm control results on the terminal.

[0012] In a possible implementation of the first aspect, the intelligent cardiopulmonary status assessment device includes a preprocessing module, a fine-grained respiratory and heartbeat waveform recovery module, and a cardiopulmonary function analysis module.

[0013] The preprocessing module is used to receive the echo signal of the training personnel to be evaluated within a preset sampling time, and perform denoising on the echo signal to obtain a denoised signal;

[0014] The fine-grained respiratory and heartbeat waveform recovery module is used to extract features from the denoised information to obtain respiratory signals, heartbeat signals and micro-motion features, and to process the respiratory and heartbeat signals using a deep learning neural network to obtain fine-grained respiratory and heartbeat waveforms;

[0015] The cardiopulmonary function analysis module is used to perform cardiopulmonary function analysis on fine-grained breathing and heartbeat waveforms to obtain cardiopulmonary stage recognition results and cardiopulmonary rhythm control results.

[0016] In a possible implementation manner of the first aspect, the preprocessing module includes a receiving unit, an average value calculating unit, and a denoising unit.

[0017] The receiving unit is used to receive the echo signal of the training person to be evaluated within a preset sampling time to obtain a signal matrix, wherein the echo signal is obtained by transmitting an ultra-wideband pulse to the training person to be evaluated through one or more cardiopulmonary sign radar detectors and reflected by the training person to be evaluated;

[0018] The average value calculation unit is used to calculate the average value of the signal matrix in the slow time dimension based on the signal matrix, and subtract the average value from the signal matrix to obtain the first processed data;

[0019] The denoising unit is used to denoise the first processed data using a moving target indicator algorithm to obtain a denoised signal.

[0020] In a possible implementation of the first aspect, the fine-grained respiratory and heartbeat waveform recovery module includes a noise adding unit, a decomposition unit, and an average processing unit.

[0021] The noise adding unit is used to add normally distributed white noise to the denoised information to obtain a signal after adding noise;

[0022] The decomposition unit is used to perform empirical mode decomposition on the signal after adding noise to obtain the intrinsic mode component, and continue to add normal distribution white noise to the denoised information and perform empirical mode decomposition to obtain new intrinsic mode components, until the preset number of times of normal distribution white noise is added to obtain the preset number of intrinsic mode components;

[0023] The average processing unit is used to perform integrated average processing on the intrinsic modal components of a preset number of times to obtain a breathing signal and a heartbeat signal.

[0024] In a possible implementation of the first aspect, the fine-grained respiratory and heartbeat waveform recovery module further includes a micro-motion feature unit,

[0025] The micro-motion feature unit is used to obtain a time-distance map, a time-Doppler map or a range-Doppler map according to the denoised information, and use the time-distance map, the time-Doppler map or the range-Doppler map as a micro-motion feature.

[0026] In a possible implementation of the first aspect, the fine-grained respiratory and heartbeat waveform recovery module further includes a neural network processing unit.

[0027] The neural network processing unit is used to convert the respiratory signal into a hidden representation through the multi-head self-attention mechanism and the feedforward neural network in the respiratory waveform recovery network to obtain the respiratory waveform feature, and to convert the heartbeat waveform recovery network into a hidden representation through the multi-head self-attention mechanism and the feedforward neural network in the heartbeat waveform recovery network to obtain the heartbeat waveform feature. The respiratory waveform feature and the heartbeat waveform feature are processed by the decoder to obtain fine-grained respiratory and heartbeat waveforms.

[0028] In a possible implementation of the first aspect, the cardiopulmonary function analysis module includes a cardiopulmonary stage recognition unit and a cardiopulmonary rhythm control statistics unit.

[0029] The cardiopulmonary stage recognition unit is used to recognize the cardiopulmonary stage according to the fine-grained breathing and heartbeat waveforms to obtain the cardiopulmonary stage recognition result;

[0030] The cardiopulmonary rhythm control statistics unit is used to perform cardiopulmonary rhythm control statistics according to fine-grained breathing and heartbeat waveforms to obtain cardiopulmonary rhythm control results.

[0031] In a possible implementation of the first aspect, the cardiopulmonary stage recognition unit includes a deep breathing period recognition subunit, a steady breathing period subunit, a breath holding period subunit, a firing vibration period subunit, and a random jitter period subunit.

[0032] Among them, the deep breathing period recognition subunit is used to recognize the deep breathing period of the person to be trained based on the fine-grained breathing and heartbeat waveforms;

[0033] The stable breathing period subunit is used to identify the stable breathing period of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0034] The breath holding period subunit is used to identify the breath holding period of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0035] The firing vibration period subunit is used to identify the firing vibration period of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0036] The random jitter period subunit is used to identify the random jitter period of the person to be trained based on fine-grained breathing and heartbeat waveforms.

[0037] In a possible implementation of the first aspect, the cardiopulmonary rhythm control statistics unit includes a heart rate statistics subunit, a respiratory rate statistics subunit, a motion statistics subunit, and a training device vibration subunit, wherein:

[0038] The heart rate statistics subunit is used to count the heart rate of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0039] The respiratory rate statistics subunit is used to count the respiratory rate of the person to be trained based on fine-grained respiratory and heartbeat waveforms;

[0040] The motion statistics subunit is used to count the small and large movements of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0041] The training device vibration statistics subunit is used to count the vibration of the training device of the person to be trained based on fine-grained breathing and heartbeat waveforms. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 : A system block diagram of an embodiment of a non-contact intelligent cardiopulmonary status assessment system provided by the present invention;

[0043] Figure 2 : A schematic diagram of the overall architecture of an intelligent cardiopulmonary status assessment system of an embodiment of a non-contact intelligent cardiopulmonary status assessment system provided by the present invention;

[0044] Figure 3 : A schematic diagram of the radar data gateway functions and equipment of an embodiment of the non-contact intelligent cardiopulmonary status assessment system provided by the present invention;

[0045] Figure 4 : A schematic diagram of a fine-grained breathing and heartbeat waveform recovery process flow of an embodiment of a non-contact intelligent cardiopulmonary status assessment system provided by the present invention;

[0046] Figure 5 : A schematic diagram of a cardiopulmonary function feature extraction process of an embodiment of a non-contact intelligent cardiopulmonary status assessment system provided by the present invention;

[0047] Figure 6 : A schematic diagram of the structure of a cardiopulmonary function analysis module of an embodiment of a non-contact intelligent cardiopulmonary status assessment system provided by the present invention;

[0048] Figure 7 : A schematic diagram of cardiopulmonary function analysis training of an embodiment of the non-contact intelligent cardiopulmonary status assessment system provided by the present invention;

[0049] Figure 8 : A schematic diagram of the overall server architecture of an embodiment of the non-contact intelligent cardiopulmonary status assessment system provided by the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Please refer to Figure 1 , which is a flow chart of an embodiment of a non-contact intelligent cardiopulmonary status assessment system provided by an embodiment of the present invention, including a cardiopulmonary vital sign radar detector, a radar data gateway and a display control terminal;

[0052] A cardiopulmonary sign radar detector, used for transmitting ultra-wideband pulses to the trainee to be evaluated, and receiving echo signals obtained by the trainee to be evaluated based on the reflection of the ultra-wideband pulses;

[0053] Radar data gateway, used to establish a communication link between the cardiopulmonary vital signs radar detector and the display control terminal to form an internal communication network;

[0054] The display control terminal is used to receive the echo signal of the training personnel to be evaluated within a preset sampling time through the radar data gateway, denoise the echo signal to obtain the denoised signal; extract features from the denoised information to obtain respiratory signals, heartbeat signals and micro-motion features, and use a deep learning neural network to process the respiratory signals and heartbeat signals to obtain fine-grained respiratory and heartbeat waveforms; perform cardiopulmonary function analysis on the fine-grained respiratory and heartbeat waveforms to obtain cardiopulmonary stage recognition results and cardiopulmonary rhythm control results, and display the cardiopulmonary stage recognition results and cardiopulmonary rhythm control results.

[0055] In this embodiment, the system mainly consists of three parts: a cardiopulmonary vital signs radar detector, a radar data gateway, and a display processing terminal. Figure 2 As shown in the figure, the functions of each module are described as follows:

[0056] In shooting applications, the cardiopulmonary radar detector is placed next to the shooter to collect data. By emitting high-frequency electromagnetic waves and receiving micro-motion echo signals reflected by the shooter's body, high-speed signal acquisition and processing are performed to demodulate the shooter's breathing, heartbeat, micro-motion information and other necessary physiological parameters and waveforms. The detector mainly consists of a microwave transmission channel, a microwave receiving channel, a transceiver antenna, a signal processing board, a communication module, etc.

[0057] The radar data gateway is responsible for establishing communication links with multiple vital sign radar detectors and display processing terminals to form an internal communication network, uploading the data of each radar detector to the display processing terminal, and issuing control instructions from the display processing terminal to each radar detector. The device diagram of the radar data gateway is shown in the figure below. Figure 3 As shown. The display control terminal can be built by a portable computer or tablet computer, running the shooting training evaluation software. Receive, store, display and evaluate the monitoring results of each radar device corresponding to the personnel.

[0058] The system can monitor the shooter's heart rate and breathing status during shooting, and evaluate the shooter's psychological control ability and breathing rhythm. At the same time, it monitors and depicts the micro-movement rhythm of each shooter during the training process, and statistically analyzes the actual shooting performance related data, which can help adjust the standardization of the shooter's shooting movements, improve the shooter's shooting ability, and optimize the shooting training method. In the shooting training application, a vital sign radar detector is placed at each shooting position to monitor the vital signs of the personnel at that position, and transmit the monitoring results to the display processing terminal through the radar data gateway. By using pulse ultra-wideband radar and using the reflected echo of the radar irradiating the human body, the human body's non-contact breathing rhythm, cardiopulmonary stage staging, heart rate and heart rate variability and other cardiopulmonary physiological characteristics can be detected and evaluated, and the micro-movement rhythm of each shooter during the training process can be monitored and depicted, and the actual shooting performance related data can be statistically analyzed.

[0059] In some embodiments, the display control terminal includes an intelligent cardiopulmonary status assessment device and a display device.

[0060] Among them, the intelligent cardiopulmonary status assessment device is used to receive the echo signal of the training personnel to be assessed within the preset sampling time through the radar data gateway, perform denoising on the echo signal to obtain the denoised signal; perform feature extraction on the denoised information to obtain the respiratory signal, heartbeat signal and micro-motion features, and use the deep learning neural network to process the respiratory signal and heartbeat signal to obtain fine-grained respiratory and heartbeat waveforms; perform cardiopulmonary function analysis on the fine-grained respiratory and heartbeat waveforms to obtain cardiopulmonary stage recognition results and cardiopulmonary rhythm control results;

[0061] The display device is used to display the cardiopulmonary stage recognition results and the cardiopulmonary rhythm control results on the terminal.

[0062] In this embodiment, the display control terminal can be built by a portable computer or a tablet computer to run the shooting training evaluation software. For the big data of vital signs observed during the shooting process, by mining the laws of individual behavior patterns in the big data of vital signs, a correlation model between life characteristics and training results based on the big data of vital signs is constructed to form an evaluation report on the breathing rhythm control effect of the shooter. The intelligent cardiopulmonary state evaluation device receives the echo signal of the training personnel to be evaluated within a preset sampling time through the radar data gateway, performs denoising processing on the echo signal to obtain the denoised signal, performs feature extraction on the denoised information to obtain the breathing signal, heartbeat signal and micro-motion features, uses a deep learning neural network to process the breathing signal and heartbeat signal to obtain fine-grained breathing and heartbeat waveforms, performs cardiopulmonary function analysis on the fine-grained breathing and heartbeat waveforms, obtains the cardiopulmonary stage recognition results and the cardiopulmonary rhythm control results, and then displays the cardiopulmonary stage recognition results and the cardiopulmonary rhythm control results on the display device.

[0063] In some embodiments, the intelligent cardiopulmonary status assessment device includes a preprocessing module, a fine-grained respiratory and heartbeat waveform recovery module, and a cardiopulmonary function analysis module.

[0064] The preprocessing module is used to receive the echo signal of the training personnel to be evaluated within a preset sampling time, and perform denoising on the echo signal to obtain a denoised signal;

[0065] The fine-grained respiratory and heartbeat waveform recovery module is used to extract features from the denoised information to obtain respiratory signals, heartbeat signals and micro-motion features, and to process the respiratory and heartbeat signals using a deep learning neural network to obtain fine-grained respiratory and heartbeat waveforms;

[0066] The cardiopulmonary function analysis module is used to perform cardiopulmonary function analysis on fine-grained breathing and heartbeat waveforms to obtain cardiopulmonary stage recognition results and cardiopulmonary rhythm control results.

[0067] In this embodiment, for the big data of vital signs observed during the shooting process, by mining the laws of individual behavior patterns in the big data of vital signs, a correlation model between life characteristics and training performance based on the big data of vital signs is constructed to form an evaluation report on the effect of the shooting personnel's breathing rhythm control. The main process is to collect the echo data of the original pulse system ultra-wideband radar equipment, perform data preprocessing, select features, and then send them to the cardiopulmonary function analysis module to output the cardiopulmonary feature results of the shooting personnel. The data preprocessing steps are:

[0068] The ultra-wideband radar using the pulse system transmits ultra-wideband pulses to the human chest cavity through a transmitting antenna or other signal transmitting equipment capable of transmitting signals. The ultra-wideband pulse is usually a Gaussian pulse signal. Taking the first-order Gaussian pulse signal as an example, the continuous time form of a pulse starting from any time t0 can be expressed as:

[0069]

[0070] Where A is the pulse amplitude, σ is the pulse variance, W is the pulse width, T p is the pulse repetition period, that is, the pulse repetition frequency f p The reciprocal of .

[0071] The radar echo signal reflected by the human body is received by the antenna. slow The echo of the transmitted pulse is collected periodically, and this frequency is called the slow time sampling frequency. The sampled data is called a frame. Since the radar transmits pulse signals periodically, the other pulses are the periodic extension of a single pulse, and the mth pulse can be expressed as:

[0072]

[0073] Where P m(t) represents the mth pulse, t is the sampling time, f slow is the sampling frequency.

[0074] Since the human body has more than one reflection point, the amplitude and delay of the echo signal are affected by the reflection point. Suppose the amplitude modulation parameter of the nth echo signal of the mth pulse is a n , the delay is τ n , then the pulse echo is expressed as:

[0075]

[0076] Where P mn (t) represents the echo signal of the nth pulse of the mth pulse, a n is the amplitude modulation parameter, τ n For delay.

[0077] Assume there are N echoes in total and the measurement noise is n0(t). The echo signal can be expressed as the sum of different delayed echoes and measurement noise. The expression of the echo signal is:

[0078] When each frame echo signal is sampled at a sampling time T f After sampling, it is recorded in a matrix, which can be expressed as:

[0079] R m,n =R m (nT f ),m=1,2,...,M; n=1,2,...,N

[0080] Where m is the slow time dimension, n is the fast time dimension, and the interval between adjacent frames becomes the pulse repetition interval.

[0081] Then suppress the clutter in the detection environment and improve the echo signal-to-noise ratio (SNR). First, remove the DC component of the baseband in the echo signal. The method is to average in the slow time dimension, regard the average value as an estimate of the DC shunt, and subtract the average value from all data. It can be written as:

[0082] Afterwards, the moving target indicator (MTI) algorithm is used to further eliminate the static DC component and background noise, which can be expressed as:

[0083]

[0084]

[0085] Among them, C m,n is the estimated value of the background signal, is the echo matrix after slow time averaging to eliminate the DC component, α is a constant between 0 and 1, used to smooth the background signal estimation, D m,n is the echo matrix obtained after MTI processing.

[0086] The background estimate of the current signal C m,n The background estimate C at the previous moment m-1,n The background signal is linearly combined with the current signal according to the weight α. In this way, the background signal can be smoothed to better suppress noise and stationary targets.

[0087] After the echo signal is preprocessed, feature extraction is performed. The EEMD algorithm is used to suppress noise and preliminarily separate the respiratory and heartbeat signals.

[0088] It should be noted that the ensemble empirical mode decomposition method (EEMD) is an analysis method for nonlinear and non-stationary time series signals. Its basic principle is to decompose the trend components or fluctuations of different scales in the signal step by step according to the local time characteristics, and generate a series of intrinsic mode functions (IMFs). Each IMF is complete and orthogonal. EEMD improves the phenomenon that traditional EMD decomposition is prone to mode aliasing in practical applications.

[0089] The respiratory signal and heartbeat signal separated by the EEMD algorithm are input into the deep learning network model for processing. The network model consists of two parts, namely the respiratory waveform recovery network and the heartbeat waveform recovery network. The inputs of the two networks are the respiratory signal and the heartbeat signal respectively, and the network is built based on the Transformer network. The Encoder layer in the network converts the radar echo data sequence in the echo signal into a series of hidden representations through the multi-head self-attention mechanism and the feedforward neural network, and extracts the deep features of the radar echo data sequence in the echo signal. The Decoder layer uses these hidden representations to gradually generate fine-grained respiratory and heartbeat waveforms. Finally, the cardiopulmonary function analysis module is used to perform cardiopulmonary function analysis on the fine-grained respiratory and heartbeat waveforms to obtain the cardiopulmonary stage recognition results and cardiopulmonary rhythm control results.

[0090] In some embodiments, the fine-grained respiratory and heartbeat waveform recovery module includes a noise adding unit, a decomposition unit and an average processing unit.

[0091] The noise adding unit is used to add normally distributed white noise to the denoised information to obtain a signal after adding noise;

[0092] The decomposition unit is used to perform empirical mode decomposition on the signal after adding noise to obtain the intrinsic mode component, and continue to add normal distribution white noise to the denoised information and perform empirical mode decomposition to obtain new intrinsic mode components, until the preset number of times of normal distribution white noise is added to obtain the preset number of intrinsic mode components;

[0093] The average processing unit is used to perform integrated average processing on the intrinsic modal components of a preset number of times to obtain a breathing signal and a heartbeat signal.

[0094] In this embodiment, for the respiratory and heartbeat signals, the specific steps of suppressing noise and preliminarily separating the respiratory and heartbeat signals using the EEMD algorithm are as follows:

[0095] S21: adding normally distributed white noise to the echo signal;

[0096] S22: performing EMD decomposition on the signal with white noise added as a whole to obtain various IMF components;

[0097] S23: repeating steps 21 and 23 for a preset number of times, adding a new normally distributed white noise sequence each time;

[0098] S24: Perform integrated averaging on the IMF obtained each time as the final result.

[0099] It should be noted that the preset number of times can be set according to actual needs.

[0100] In some embodiments, the fine-grained respiratory and heartbeat waveform recovery module further includes a micro-motion feature unit.

[0101] The micro-motion feature unit is used to obtain a time-distance map, a time-Doppler map or a range-Doppler map according to the denoised information, and use the time-distance map, the time-Doppler map or the range-Doppler map as a micro-motion feature.

[0102] In this embodiment, for the micro-motion feature, the radar echo can be selected to obtain a time distance diagram, a time Doppler diagram or a distance Doppler diagram based on the radar echo matrix as the micro-motion feature. Taking the time Doppler diagram as an example, first, a constant false alarm rate (CFAR) detection is performed in the fast time (distance) dimension to determine the position of the human target scattering point. Then, for each signal frame, a fast Fourier transform (FFT) is performed in the slow time window where the detection point is located. Finally, the slow time dimension is scanned, and the TD matrix can be obtained by summing the spectrum near all detection points of each frame signal.

[0103] In some embodiments, the fine-grained respiratory and heartbeat waveform recovery module further includes a neural network processing unit.

[0104] The neural network processing unit is used to convert the respiratory signal into a hidden representation through the multi-head self-attention mechanism and the feedforward neural network in the respiratory waveform recovery network to obtain the respiratory waveform feature, and to convert the heartbeat waveform recovery network into a hidden representation through the multi-head self-attention mechanism and the feedforward neural network in the heartbeat waveform recovery network to obtain the heartbeat waveform feature. The respiratory waveform feature and the heartbeat waveform feature are processed by the decoder to obtain fine-grained respiratory and heartbeat waveforms.

[0105] In this embodiment, the deep learning network includes a breathing waveform recovery network and a heartbeat waveform recovery network. The network is constructed based on the Transformer network. The Encoder layer in the network converts the radar echo data sequence in the echo signal into a series of hidden representations through a multi-head self-attention mechanism and a feedforward neural network, extracts deep features in the echo signal, and the Decoder layer uses these hidden representations to gradually generate fine-grained breathing and heartbeat waveforms. Figure 4 The figure is a schematic diagram of the fine-grained breathing and heartbeat waveform recovery process flow, which can be understood as follows: the Encoder layer in the breathing waveform recovery network converts the echo data sequence into a series of hidden representations through a multi-head self-attention mechanism and a feedforward neural network, extracts the breathing features of the radar echo data sequence in the echo signal, and the decoder uses these hidden representations to obtain the breathing waveform. The Encoder layer in the heartbeat waveform recovery network converts the echo data sequence into a series of hidden representations through a multi-head self-attention mechanism and a feedforward neural network, extracts the heartbeat features of the radar echo data sequence in the echo signal, and the decoder uses these hidden representations to generate fine-grained breathing and heartbeat waveforms.

[0106] The training data set of the deep learning network uses the synchronously collected pulse ultra-wideband radar human body echo and breathing and heartbeat data. The loss function is defined as the L2 norm of the difference between the restored fine-grained breathing and heartbeat waveforms and the real breathing and heartbeat waveforms at each point, that is:

[0107]

[0108] Among them, x gt Represents the true value of the breathing and heartbeat waveforms, and x represents the restored fine-grained breathing and heartbeat waveforms.

[0109] Through the back-propagation algorithm, the gradient of the loss function is calculated and the model parameters are updated. During the training process, the validation set is used to monitor the model performance and adjust hyperparameters such as learning rate and batch size to avoid overfitting or underfitting. Through multiple rounds of iterations, until the loss function on the validation set converges or the preset number of training rounds is reached. Finally, the actual performance of the model is evaluated using the test set to ensure that the model can accurately restore fine-grained breathing and heartbeat waveforms on unseen data. Based on the restored fine-grained breathing and heartbeat waveforms, parameters such as heart rate variability can be further analyzed.

[0110] In some embodiments, the cardiopulmonary function analysis module includes a cardiopulmonary stage identification unit and a cardiopulmonary rhythm control statistics unit.

[0111] The cardiopulmonary stage recognition unit is used to recognize the cardiopulmonary stage according to the fine-grained breathing and heartbeat waveforms to obtain the cardiopulmonary stage recognition result;

[0112] The cardiopulmonary rhythm control statistics unit is used to perform cardiopulmonary rhythm control statistics according to fine-grained breathing and heartbeat waveforms to obtain cardiopulmonary rhythm control results.

[0113] In this embodiment, the selected breathing or heartbeat features and micro-motion features are sent to the cardiopulmonary function analysis module, and the extraction process is as follows: Figure 5 As shown, the module structure is as follows Figure 6 As shown in the figure, this module mainly includes a cardiopulmonary stage identification submodule and a cardiopulmonary rhythm control effect statistics submodule. According to the fine-grained breathing and heartbeat waveforms, the cardiopulmonary stage identification submodule and the cardiopulmonary rhythm control effect statistics submodule are used to perform cardiopulmonary stage identification, and the cardiopulmonary stage identification results and cardiopulmonary rhythm control results are obtained.

[0114] In some embodiments, the cardiopulmonary stage identification unit includes a deep breathing period identification subunit, a steady breathing period subunit, a breath holding period subunit, a firing vibration period subunit, and a random jitter period subunit.

[0115] Among them, the deep breathing period recognition subunit is used to recognize the deep breathing period of the person to be trained based on the fine-grained breathing and heartbeat waveforms;

[0116] The stable breathing period subunit is used to identify the stable breathing period of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0117] The breath holding period subunit is used to identify the breath holding period of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0118] The firing vibration period subunit is used to identify the firing vibration period of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0119] The random jitter period subunit is used to identify the random jitter period of the person to be trained based on fine-grained breathing and heartbeat waveforms.

[0120] In this embodiment, the cardiopulmonary stage stage recognition submodule includes recognition of deep breathing period, steady breathing period, breath holding period, firing vibration period and random jitter period. According to the fine-grained breathing and heartbeat waveforms, the cardiopulmonary stage recognition is performed using the recognition of deep breathing period, steady breathing period, breath holding period, firing vibration period and random jitter period submodules to obtain the corresponding recognition results.

[0121] In some embodiments, the cardiopulmonary rhythm control statistics unit includes a heart rate statistics subunit, a respiratory rate statistics subunit, a motion statistics subunit and a training device vibration subunit, wherein:

[0122] The heart rate statistics subunit is used to count the heart rate of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0123] The respiratory rate statistics subunit is used to count the respiratory rate of the person to be trained based on fine-grained respiratory and heartbeat waveforms;

[0124] The motion statistics subunit is used to count the small and large movements of the person to be trained based on fine-grained breathing and heartbeat waveforms;

[0125] The training device vibration statistics subunit is used to count the vibration of the training device of the person to be trained based on fine-grained breathing and heartbeat waveforms.

[0126] In this embodiment, the cardiopulmonary rhythm control effect statistics submodule includes breathing statistics, heart rate statistics, small movement statistics, large movement statistics, and gun vibration statistics. In this part, it includes human body micro-motion detection and movement statistics, heart rate estimation and heart rate statistics, breathing estimation and breathing statistics.

[0127] Cardiopulmonary function analysis training for shooters Figure 7 As shown, firstly, the correspondence between shooting signs such as cardiopulmonary characteristics, micro-motion information and shooting target performance is established, which is divided into a labeled training set and a verification set. The heart rate detection model, the respiratory rate detection model, and the relationship model between micro-motion characteristics and shooting performance are trained on the training set, and then verified on the verification set. If the model verification is feasible, the evaluation status of the shooter's feature can be output.

[0128] It should be noted that the collected data is stored in the backend server, which includes the collection terminal, application server, database server, load balancing server, etc. The server adopts the B / S mode, and the overall architecture is as follows: Figure 8 As shown in the figure, the service layer obtains the data collected by the perception layer through the network, calculates, processes and mines the data, and interacts with the application layer through an open API.

[0129] It should be noted that the non-contact intelligent cardiopulmonary status assessment system implemented by the present invention can extract the human heart, lung and micro-motion characteristics without contact. It is not limited to shooting training, but can also be applied to the fields of sports posture correction, medical monitoring and sleep monitoring.

[0130] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0131] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A non-contact intelligent cardiopulmonary status assessment system, characterized in that: Including cardiopulmonary signs radar detector, radar data gateway and display control terminal; The cardiopulmonary sign radar detector is used to transmit an ultra-wideband pulse to the trainee to be evaluated, and receive an echo signal obtained by the trainee to be evaluated based on the reflection of the ultra-wideband pulse; The radar data gateway is used to establish a communication link between the cardiopulmonary vital signs radar detector and the display control terminal to form an internal communication network; The display control terminal is used to receive the echo signal of the training personnel to be evaluated within a preset sampling time through the radar data gateway, perform denoising on the echo signal to obtain a denoised signal, perform feature extraction on the denoised information to obtain a respiratory signal, a heartbeat signal and micro-motion features, process the respiratory signal and the heartbeat signal using a deep learning neural network to obtain fine-grained respiratory and heartbeat waveforms, perform cardiopulmonary function analysis on the fine-grained respiratory and heartbeat waveforms to obtain a cardiopulmonary stage recognition result and a cardiopulmonary rhythm control result, and display the cardiopulmonary stage recognition result and the cardiopulmonary rhythm control result.

2. The non-contact intelligent cardiopulmonary status assessment system according to claim 1, characterized in that: The display control terminal includes an intelligent cardiopulmonary status assessment device and a display device. The intelligent cardiopulmonary status assessment device is used to receive the echo signal of the training personnel to be assessed within a preset sampling time through the radar data gateway, perform denoising on the echo signal to obtain a denoised signal; perform feature extraction on the denoised information to obtain a respiratory signal, a heartbeat signal and micro-motion features, and use a deep learning neural network to process the respiratory signal and the heartbeat signal to obtain fine-grained respiratory and heartbeat waveforms; perform cardiopulmonary function analysis on the fine-grained respiratory and heartbeat waveforms to obtain a cardiopulmonary stage recognition result and a cardiopulmonary rhythm control result; The display device is used to display the cardiopulmonary stage recognition result and the cardiopulmonary rhythm control result on the terminal.

3. The non-contact intelligent cardiopulmonary status assessment system according to claim 2, characterized in that: The intelligent cardiopulmonary status assessment device includes a preprocessing module, a fine-grained respiratory and heartbeat waveform recovery module, and a cardiopulmonary function analysis module. The preprocessing module is used to receive the echo signal of the training personnel to be evaluated within a preset sampling time, and perform denoising on the echo signal to obtain a denoised signal; The fine-grained respiratory and heartbeat waveform recovery module is used to extract features from the denoised information to obtain respiratory signals, heartbeat signals and micro-motion features, and to process the respiratory signals and heartbeat signals using a deep learning neural network to obtain fine-grained respiratory and heartbeat waveforms; The cardiopulmonary function analysis module is used to perform cardiopulmonary function analysis on the fine-grained breathing and heartbeat waveforms to obtain cardiopulmonary stage recognition results and cardiopulmonary rhythm control results.

4. The non-contact intelligent cardiopulmonary status assessment system according to claim 3, characterized in that: The preprocessing module includes a receiving unit, an average value calculation unit and a denoising unit. The receiving unit is used to receive the echo signal of the training person to be evaluated within a preset sampling time to obtain a signal matrix, wherein the echo signal is obtained by transmitting an ultra-wideband pulse to the training person to be evaluated through one or more cardiopulmonary sign radar detectors and reflected by the training person to be evaluated; The average value calculation unit is used to calculate the average value of the signal matrix in the slow time dimension based on the signal matrix, and subtract the average value from the signal matrix to obtain the first processed data; The denoising unit is used to denoise the first processed data using the moving target indicator algorithm to obtain a denoised signal.

5. The non-contact intelligent cardiopulmonary status assessment system according to claim 3, characterized in that: The fine-grained respiratory and heartbeat waveform recovery module includes a noise adding unit, a decomposition unit and an average processing unit. The noise adding unit is used to add normally distributed white noise to the denoised information to obtain a signal after adding noise; The decomposition unit is used to perform empirical mode decomposition on the signal after adding noise to obtain intrinsic mode components, and continue to add normally distributed white noise to the denoised information and perform empirical mode decomposition to obtain new intrinsic mode components, until the normally distributed white noise after a preset number of times is added to obtain the intrinsic mode components of the preset number; The average processing unit is used to perform integrated average processing on the intrinsic modal components of a preset number of times to obtain a breathing signal and a heartbeat signal.

6. The non-contact intelligent cardiopulmonary status assessment system according to claim 5, characterized in that: The fine-grained breathing and heartbeat waveform recovery module also includes a micro-motion feature unit, The micro-motion feature unit is used to obtain a time-distance map, a time-Doppler map or a range-Doppler map according to the denoised information, and use the time-distance map, the time-Doppler map or the range-Doppler map as a micro-motion feature.

7. The non-contact intelligent cardiopulmonary status assessment system according to claim 5, characterized in that: The fine-grained respiratory and heartbeat waveform recovery module also includes a neural network processing unit, The neural network processing unit is used to convert the respiratory signal into a hidden representation through a multi-head self-attention mechanism and a feedforward neural network in a respiratory waveform recovery network to obtain a respiratory waveform feature, and to convert the heartbeat waveform recovery network into a hidden representation through a multi-head self-attention mechanism and a feedforward neural network in a heartbeat waveform recovery network to obtain a heartbeat waveform feature, and to process the respiratory waveform feature and the heartbeat waveform feature using a decoder to obtain fine-grained respiratory and heartbeat waveforms.

8. The non-contact intelligent cardiopulmonary status assessment system according to claim 3, characterized in that: The cardiopulmonary function analysis module includes a cardiopulmonary stage recognition unit and a cardiopulmonary rhythm control statistics unit. The cardiopulmonary stage recognition unit is used to perform cardiopulmonary stage recognition according to the fine-grained breathing and heartbeat waveforms to obtain a cardiopulmonary stage recognition result; The cardiopulmonary rhythm control statistics unit is used to perform cardiopulmonary rhythm control statistics according to the fine-grained breathing and heartbeat waveforms to obtain a cardiopulmonary rhythm control result.

9. The non-contact intelligent cardiopulmonary status assessment system according to claim 8, characterized in that: The cardiopulmonary stage recognition unit includes a deep breathing period recognition subunit, a steady breathing period subunit, a breath holding period subunit, a firing vibration period subunit and a random jitter period subunit Wherein, the deep breathing period identification subunit is used to identify the deep breathing period of the person to be trained according to the fine-grained breathing and heartbeat waveforms; The stable breathing period subunit is used to identify the stable breathing period of the person to be trained according to the fine-grained breathing and heartbeat waveforms; The breath-holding period subunit is used to identify the breath-holding period of the person to be trained according to the fine-grained breathing and heartbeat waveforms; The firing vibration period subunit is used to identify the firing vibration period of the person to be trained according to the fine-grained breathing and heartbeat waveforms; The random jitter period subunit is used to identify the random jitter period of the person to be trained according to the fine-grained breathing and heartbeat waveforms.

10. The non-contact intelligent cardiopulmonary status assessment system according to claim 8, characterized in that: The cardiopulmonary rhythm control statistics unit includes a heart rate statistics subunit, a respiratory rate statistics subunit, a motion statistics subunit and a training device vibration subunit, wherein: The heart rate statistics subunit is used to count the heart rate of the person to be trained according to the fine-grained breathing and heartbeat waveforms; The respiratory rate statistics subunit is used to count the respiratory rate of the person to be trained according to the fine-grained respiratory and heartbeat waveforms; The motion statistics subunit is used for counting the small and large movements of the person to be trained according to the fine-grained breathing and heartbeat waveforms; The training device vibration statistics subunit is used for counting the vibration of the training device of the person to be trained according to the fine-grained breathing and heartbeat waveforms.

Images