Respiratory pattern classification method, system, device and medium based on residual network

Abstract

The present invention provides a breathing pattern classification method, system, device and medium based on a residual network, the method comprising: presetting a plurality of breathing patterns, and respectively collecting UWB echo signals corresponding to each breathing pattern; establishing a breathing signal data set according to the UWB echo signals of each breathing pattern, and dividing the breathing signal data set into a training set and a test set according to a preset ratio; training an improved residual network according to the training set to obtain a breathing classification model; inputting the test set into the breathing classification model for classification prediction to obtain a breathing pattern recognition result. Based on the advantages of UWB technology and deep learning, the present invention can perform simple and efficient prediction and classification of non-contact breathing patterns, while effectively improving the prediction accuracy and stability of small data volume application scenarios, so as to provide more accurate and effective abnormal breathing monitoring, thereby enhancing the application value of abnormal breathing monitoring.

Description

Technical Field

[0001] The present invention relates to the field of biomedical technology, and in particular to a breathing pattern classification method, system, computer equipment and storage medium based on a residual network. Background Art

[0002] As one of the most important parameters of the human body, breathing can often indicate whether the physiological state of the human body is normal and healthy. Respiratory movement is the movement of the respiratory muscles to maintain rhythmic activity under the action of the lung traction reflex and the respiratory center, causing the chest to expand or contract. It maintains the stability of the oxygen content in the body by inhaling oxygen and exhaling carbon dioxide. This is an automatic and rhythmic movement. By observing the breathing of the human body in different states, it is possible to detect abnormal breathing and extract physiological information of the human body. Therefore, respiratory research is of great significance in sleep stratification, clinical monitoring, feedback regulation of the cardiovascular system and many other fields.

[0003] The existing common methods for monitoring respiratory morphology include: pressure sensor method, impedance respiration measurement method, temperature sensor method and respiratory induction plethysmography method, etc. However, these methods are all contact monitoring methods, which have great limitations. For some special populations, for example, for some patients with fever, infectious diseases, skin diseases, and large-area burns, contact devices cannot be worn. In order to overcome the limitations of contact respiratory monitoring, researchers have also proposed non-contact respiratory morphology monitoring methods, but the current non-contact respiratory morphology monitoring classification based on ultra-wideband radar (UWB) technology not only requires relevant professional knowledge and labor intensity in manual feature extraction, but also has a relatively small amount of abnormal respiratory data in clinical monitoring, which affects the effect of deep learning, resulting in insufficient accuracy and stability of respiratory classification prediction and low practicality.

[0004] Therefore, there is an urgent need to provide a non-contact breathing pattern classification method that can improve the prediction accuracy and stability of small data application scenarios based on the advantages of UWB technology and deep learning. Summary of the invention

[0005] The purpose of the present invention is to provide a breathing pattern classification method based on a residual network. After converting the echo signal obtained by UWB technology into tensor data that can be used for deep learning, the improved residual network is divided into preset batches for training to obtain a stable breathing classification model, which is used to perform simple and efficient breathing pattern recognition on non-contact breathing data. It can effectively improve the prediction accuracy and stability of small data volume application scenarios, provide more accurate and effective breathing abnormality monitoring, and thus enhance the application value of breathing abnormality monitoring.

[0006] In order to achieve the above objectives, it is necessary to provide a breathing pattern classification method, system, computer device and storage medium based on a residual network in response to the above technical problems.

[0007] In a first aspect, an embodiment of the present invention provides a breathing pattern classification method based on a residual network, the method comprising the following steps:

[0008] Preset multiple breathing modes and collect UWB echo signals corresponding to each breathing mode respectively;

[0009] According to the UWB echo signals of each breathing mode, a breathing signal data set is established, and the breathing signal data set is divided into a training set and a test set according to a preset ratio; the breathing signal data set includes echo tensor data and mode labels corresponding to the UWB echo signals of each breathing mode;

[0010] According to the training set, the improved residual network is trained to obtain a breathing classification model;

[0011] The test set is input into the breathing classification model for classification prediction to obtain a breathing pattern recognition result.

[0012] Furthermore, the breathing mode includes a breath-holding mode, a chest breathing mode, an abdominal breathing mode and a mixed breathing mode;

[0013] Furthermore, the step of establishing a breathing signal data set according to the UWB echo signals of each breathing mode includes:

[0014] Perform data preprocessing on each UWB echo signal to obtain corresponding echo tensor data;

[0015] Corresponding pattern labels are added to each echo tensor data to obtain the respiratory signal data set.

[0016] Furthermore, the step of performing data preprocessing on each UWB echo signal to obtain corresponding echo tensor data includes:

[0017] Each UWB echo signal is equally spaced according to a preset interval to obtain echo segmentation data;

[0018] Standardizing the echo segmentation data to obtain echo standardized data;

[0019] The echo normalized data is converted into corresponding tensor data, and the tensor data is dimensionally expanded to obtain the echo tensor data.

[0020] Furthermore, the step of normalizing the echo segmentation data to obtain echo normalized data includes:

[0021] Selecting the real part of the echo segmentation data to obtain corresponding data to be standardized;

[0022] The data to be standardized is standardized to obtain the echo standardized data.

[0023] Furthermore, the improved residual network includes an input layer, four residual blocks and an output layer connected in sequence;

[0024] The input layer includes a two-dimensional convolution layer, a normalization layer, an activation function and a maximum pooling layer connected in sequence;

[0025] The residual block includes a first convolutional layer block and a second convolutional layer block connected in sequence; the first convolutional layer block includes a first preset number of two-dimensional convolutional layers of different sizes connected in sequence, and a corresponding plurality of normalization layers; the second convolutional layer block includes a second preset number of two-dimensional convolutional layers of different sizes connected in sequence, and a corresponding plurality of normalization layers;

[0026] The output layer includes a global maximum pooling layer, a flattening layer and a linear layer connected in sequence.

[0027] Furthermore, the step of training the improved residual network according to the training set to obtain the breathing classification model includes:

[0028] The training set is randomly divided into preset batches, and the improved residual network is trained by back propagation supervision using the Adam optimizer to obtain the respiratory classification model.

[0029] In a second aspect, an embodiment of the present invention provides a breathing pattern classification system based on a residual network, the system comprising:

[0030] A data acquisition module is used to preset multiple breathing modes and collect UWB echo signals corresponding to each breathing mode respectively;

[0031] A data processing module, for establishing a respiratory signal data set according to the UWB echo signals of each respiratory mode, and dividing the respiratory signal data set into a training set and a test set according to a preset ratio; the respiratory signal data set includes echo tensor data and mode labels corresponding to the UWB echo signals of each respiratory mode;

[0032] A model training module, used to train the improved residual network according to the training set to obtain a breathing classification model;

[0033] The pattern recognition module is used to input the test set into the breathing classification model for classification prediction to obtain a breathing pattern recognition result.

[0034] In a third aspect, an embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0035] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0036] The present application provides a method, system, computer device and storage medium for respiratory pattern classification based on residual network. Through the method, the collected UWB echo signals of each preset respiratory pattern are converted into tensor data that can be used for deep learning by corresponding data preprocessing, and the corresponding respiratory signal data set is established and divided into training set and test set according to a preset ratio. Then, the improved residual network is trained according to the training set to obtain a respiratory classification model, and the test set is input into the respiratory classification model for classification prediction to obtain a technical solution for respiratory pattern recognition results. Compared with the prior art, the respiratory pattern classification method based on residual network can, based on the advantages of UWB technology and deep learning, perform simple and efficient prediction and classification of non-contact respiratory patterns, while effectively improving the prediction accuracy and stability of small data application scenarios, so as to provide more accurate and effective respiratory abnormality monitoring, thereby enhancing the application value of respiratory abnormality monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a schematic diagram of an application scenario of respiratory pattern classification based on a residual network in an embodiment of the present invention;

[0038] Figure 2 is a flow chart of a breathing pattern classification method based on a residual network in an embodiment of the present invention;

[0039] Figure 3 is a schematic diagram of the structure of an improved residual network in an embodiment of the present invention;

[0040] Figure 4 yes Figure 3 Schematic diagram of the structure of the residual block;

[0041] Figure 5 is a structural schematic diagram of a breathing classification model obtained by training an improved residual network in an embodiment of the present invention;

[0042] Figure 6 yes Figure 3 The diagram shows the visualization results of the loss rate and accuracy of the improved residual network training;

[0043] Figure 7is a structural schematic diagram of a breathing pattern classification system based on a residual network in an embodiment of the present invention;

[0044] Figure 8 It is a diagram of the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical scheme and beneficial effects of the present application clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the embodiments described below are part of the embodiments of the present invention and are only used to illustrate the present invention, but are not used to limit the scope of the present invention. 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.

[0046] The breathing pattern classification method based on residual network provided by the present invention can be applied to Figure 1 The terminal or server shown. Among them, the terminal can be but not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the server can be implemented by an independent server or a server cluster composed of multiple servers. The present invention is based on the research of breathing pattern classification based on UWB breathing echo signals. For example, the server or terminal can collect the UWB echo signal corresponding to the preset breathing pattern according to the breathing pattern classification method based on the residual network of the present invention, and then perform pre-processing to establish the corresponding breathing signal data set, and train the improved residual network accordingly to obtain a breathing classification model, which is used to predict and classify the breathing pattern of the UWB echo signal to be identified to obtain the corresponding breathing pattern recognition result; the following embodiment will explain the breathing pattern classification method based on the residual network of the present invention in detail.

[0047] In one embodiment, Figure 2 As shown, a breathing pattern classification method based on a residual network is provided, comprising the following steps:

[0048] S11, preset multiple breathing patterns, and collect UWB echo signals corresponding to each breathing pattern respectively; wherein, ultra-wideband radar (UWB) is a technology that uses electromagnetic waves and Doppler principle to measure vital signs, and has the advantages of long detection distance, high sensitivity, and mature technology; the present invention is based on the non-contact acquisition of echo signals by ultra-wideband radar (UWB) as breathing signals for research;

[0049] The breathing mode can be adjusted according to actual application requirements. This embodiment preferably includes a breath-holding mode, a chest breathing mode, an abdominal breathing mode and a mixed breathing mode; correspondingly, the breath-holding mode can be understood as the user is in a breath-holding state, and the corresponding mode label can be set to 0; the chest breathing mode can be understood as the user is in a chest breathing state, and the corresponding mode label can be set to 1; the abdominal breathing mode can be understood as the user is in an abdominal breathing state, and the corresponding mode label can be set to 2; the mixed breathing mode can be understood as the user is in both chest breathing and abdominal breathing states, and the corresponding mode label can be set to 3; when the actual echo signal is collected, the duration and frequency of the collection in each preset state can be adjusted according to the actual application requirements and the conditions of the collection equipment. In order to ensure the effect of the subsequent deep learning network training, it is necessary to collect as many UWB signals as possible; it should be noted that the device and method for collecting UWB echo signals in this embodiment can be implemented by existing technologies, and the corresponding mode label values ​​given above are only exemplary descriptions, and the protection scope is not specifically limited.

[0050] S12. According to the UWB echo signals of each breathing mode, a breathing signal data set is established, and the breathing signal data set is divided into a training set and a test set according to a preset ratio; the breathing signal data set includes echo tensor data and mode labels corresponding to the UWB echo signals of each breathing mode; wherein, the UWB echo signal (radar data) cannot be directly used for learning and training of a deep learning network. In order to improve classification efficiency and accuracy using a deep network, the present embodiment preferably pre-processes the collected UWB echo signal into corresponding tensor data to construct a breathing signal data set for deep learning; specifically, the step of establishing a breathing signal data set according to the UWB echo signals of each breathing mode includes:

[0051] Performing data preprocessing on each UWB echo signal to obtain corresponding echo tensor data; wherein the step of performing data preprocessing on each UWB echo signal to obtain corresponding echo tensor data comprises:

[0052] Each UWB echo signal is equally spaced according to a preset spacing to obtain echo segmentation data; wherein the preset spacing can be adjusted according to actual application requirements on the basis of ensuring that the data of the breathing signal data set meets the requirements of deep learning. For example, the preset spacing is set to 10s, that is, each collected UWB echo signal is equally spaced according to a spacing of 10s while ensuring the data format, to obtain a number of echo segmentation data under a certain breathing mode, and annotate the same mode label;

[0053] The echo segmentation data is standardized to obtain echo standardized data; wherein, the standardization process can be performed by an existing annotation processing method, such as the special data preprocessing function fit_transform() in the sklearn library, which can realize the standardization process of this embodiment; in addition to the echo segmentation data, in order to improve the efficiency of data conversion processing and ensure the efficiency of subsequent pattern classification and recognition, this embodiment takes into account that the echo data is in complex form and most of the information of the echo sequence signal is retained in the real part information. Before the standardization process, the echo segmentation data is subjected to real part extraction processing, and the specific steps include: selecting the real part of the echo segmentation data to obtain the corresponding data to be standardized; standardizing the data to be standardized to obtain the echo standardized data;

[0054] The echo normalized data is converted into corresponding tensor data, and the tensor data is dimensionally expanded to obtain the echo tensor data; wherein, the tensor data can be understood as tensor format data that can be used for processing by a deep learning network, and the dimension of the obtained tensor format data is expanded to four dimensions, and the corresponding dimension size can be expressed as [the total number of echo tensor data in the respiratory signal data set, 1, the length of each echo tensor data, 1]. The number 1 used for the specific dimension expansion itself has no special meaning and can be replaced by other values ​​in principle. In order to ensure the efficiency of data processing as much as possible without affecting the data processing results, this embodiment preferably expands with the number 1 to obtain the required dimension size; it should be noted that the specific tensor data conversion method can refer to the actual prior art and will not be repeated here;

[0055] Add corresponding pattern labels to each echo tensor data to obtain the respiratory signal dataset; wherein, the echo tensor data is converted from the corresponding echo signal as described above, the pattern label of the corresponding echo tensor data is consistent with the pattern label corresponding to the original echo signal, and the echo tensor data and the pattern label corresponding to the original echo signal are connected using the DataLoader data loader in pytorch to obtain the required respiratory signal dataset.

[0056] After obtaining the breathing signal data set using the above method, the echo tensor data in the call signal data set can be randomly shuffled, and then divided into a training set and a test set according to a preset ratio; specifically, the preset ratio can in principle be set according to actual application requirements. In this embodiment, taking into account the model training effect of a small amount of data, it is preferred to set the preset ratio to 8 (training set): 2 (test set). If the amount of data that can be collected in actual applications increases, the preset ratio can be adjusted according to the amount of data, and no specific restrictions are made here.

[0057] S13, according to the training set, the improved residual network is trained to obtain a breathing classification model; the improved residual network includes: Figure 3 As shown, the input layer, 4 residual blocks and output layer are connected in sequence;

[0058] The input layer includes a two-dimensional convolutional layer, a normalization layer (BatchNorm2d), an activation function (ReLU) and a maximum pooling layer (MaxPool2d) connected in sequence;

[0059] The residual block is Figure 4 As shown, it includes a first convolutional layer block and a second convolutional layer block connected in sequence; the first convolutional layer block includes a first preset number of two-dimensional convolutional layers of different sizes connected in sequence, and a corresponding plurality of normalization layers; the second convolutional layer block includes a second preset number of two-dimensional convolutional layers of different sizes connected in sequence, and a corresponding plurality of normalization layers;

[0060] The output layer includes a global maximum pooling layer (GlobalAvgPool2d), a flattening layer (FlattenLayer) and a linear layer (Linear) connected in sequence; the number of two-dimensional convolutional layers and the size of the convolution kernel in each residual block in the above improved residual network structure can be optimized according to actual application requirements and obtained after iterative training, such as Figure 5 The stable respiratory classification model shown can be directly applied to classification prediction; it should be noted that, in principle, the training of the improved residual network can be directly trained using the training set data, but considering the limitation of the small amount of data in the application scenario, the present embodiment preferably divides the training set into small batches for training to ensure the reliability and stability of the training results; specifically, the step of training the improved residual network according to the training set to obtain the respiratory classification model includes:

[0061] The training set is randomly divided into preset batches, the improved residual network is trained with back-propagation supervision by using the Adam optimizer, and the pytorch deep learning library is used for evaluation to obtain the respiratory classification model; wherein the preset batches can be determined according to the size of the actual training set, such as training two data in each batch, dividing all training sets into corresponding multi-batch training sets, setting the learning rate to 0.0001 for learning and training, and when the model reaches the preset convergence condition, such as Figure 6 When the loss rate shown basically stops decreasing and the accuracy stops improving, the iteration is stopped to obtain a stable respiratory classification model;

[0062] This embodiment adopts an improved residual network used for training, which has a simple structure and fewer layers, and is more suitable for learning and training of small data sets. Most of the existing residual networks are used in the field of image processing. This embodiment converts the UWB signal into tensor form data similar to an image and expands the dimension, that is, converts the sequence signal into a picture for efficient feature extraction and classification. On the basis of retaining the original echo signal information, it can effectively improve the accuracy and stability of breathing pattern classification, and can achieve better classification effect than convolutional neural networks.

[0063] S14, inputting the test set into the breathing classification model for classification prediction to obtain a breathing pattern recognition result; wherein, the process of inputting the test set data into the breathing classification model for classification prediction is the same as the process of the above-mentioned model training, which will not be repeated here.

[0064] The embodiment of the present application performs corresponding data preprocessing on the collected UWB echo signals of each preset breathing pattern and converts them into tensor data that can be used for deep learning. Based on this, a corresponding breathing signal data set is established and divided into a training set and a test set according to a preset ratio. The training set is divided into preset batches to train the improved residual network to obtain a stable breathing classification model. The method for identifying breathing patterns can not only convert UWB signals into images for efficient feature extraction and classification based on the advantages of UWB technology and deep learning, and realize simple and efficient prediction and classification of non-contact breathing patterns, but also can effectively improve the prediction accuracy and stability of small data application scenarios on the basis of retaining the original echo signal information, so as to provide more accurate and effective breathing abnormality monitoring, thereby enhancing the application value of breathing abnormality monitoring.

[0065] It should be noted that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders.

[0066] In one embodiment, Figure 7 As shown, a breathing pattern classification system based on a residual network is provided, and the system comprises:

[0067] The data acquisition module 1 is used to preset multiple breathing modes and respectively collect the UWB echo signals corresponding to each breathing mode;

[0068] Data processing module 2, used to establish a breathing signal data set according to the UWB echo signals of each breathing mode, and divide the breathing signal data set into a training set and a test set according to a preset ratio; the breathing signal data set includes echo tensor data and mode labels corresponding to the UWB echo signals of each breathing mode;

[0069] A model training module 3 is used to train the improved residual network according to the training set to obtain a breathing classification model;

[0070] The pattern recognition module 4 is used to input the test set into the breathing classification model for classification prediction to obtain a breathing pattern recognition result.

[0071] For the specific definition of a breathing pattern classification system based on a residual network, please refer to the definition of a breathing pattern classification method based on a residual network in the above text, which will not be repeated here. Each module in the above-mentioned breathing pattern classification system based on a residual network can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0072] Figure 8 FIG. 1 shows an internal structure diagram of a computer device in an embodiment, and the computer device may specifically be a terminal or a server. Figure 8 As shown, the computer device includes a processor, a memory, a network interface, a display and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a breathing pattern classification method based on a residual network is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0073] It can be understood by those skilled in the art that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the present application scheme, and does not constitute a limitation on the computer device to which the present application scheme is applied. The specific computing device may include more or fewer components than shown in the figure, or combine certain components, or have the same component arrangement.

[0074] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the steps of the above method are implemented when the processor executes the computer program.

[0075] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0076] In summary, the embodiments of the present invention provide a breathing pattern classification method, system, computer device and storage medium based on a residual network. The breathing pattern classification method based on the residual network realizes the preset of multiple breathing patterns and the collection of UWB echo signals corresponding to each breathing pattern, and the collected UWB echo signals of each preset breathing pattern are converted into tensor data that can be used for deep learning by corresponding data preprocessing, and the corresponding breathing signal data set is established accordingly and divided into a training set and a test set according to a preset ratio, and the training set is divided into preset batches to train the improved residual network to obtain a stable breathing classification model, which is a technical solution for identifying breathing patterns. This method can not only convert UWB signals into images for efficient feature extraction and classification based on the advantages of UWB technology and deep learning, and realize simple and efficient prediction and classification of non-contact breathing patterns, but also can effectively improve the prediction accuracy and stability of small data application scenarios on the basis of retaining the original echo signal information, so as to provide more accurate and effective abnormal breathing monitoring, thereby enhancing the application value of abnormal breathing monitoring.

[0077] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not 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.

[0078] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A breathing pattern classification method based on residual network, characterized in that: The method comprises the following steps: Preset multiple breathing modes and collect UWB echo signals corresponding to each breathing mode respectively; According to the UWB echo signals of each breathing mode, a breathing signal data set is established, and the breathing signal data set is divided into a training set and a test set according to a preset ratio; the breathing signal data set includes echo tensor data and mode labels corresponding to the UWB echo signals of each breathing mode; According to the training set, the improved residual network is trained to obtain a breathing classification model; the improved residual network includes an input layer, 4 residual blocks and an output layer connected in sequence; the input layer includes a two-dimensional convolutional layer, a normalization layer, an activation function and a maximum pooling layer connected in sequence; the residual block includes a first convolutional layer block and a second convolutional layer block connected in sequence; the first convolutional layer block includes a first preset number of two-dimensional convolutional layers of different sizes connected in sequence, and a corresponding plurality of normalization layers; the second convolutional layer block includes a second preset number of two-dimensional convolutional layers of different sizes connected in sequence, and a corresponding plurality of normalization layers; the output layer includes a global maximum pooling layer, a flattening layer and a linear layer connected in sequence; Inputting the test set into the breathing classification model for classification prediction to obtain a breathing pattern recognition result; The step of establishing a breathing signal data set according to the UWB echo signals of each breathing mode includes: Performing data preprocessing on each UWB echo signal to obtain corresponding echo tensor data, including: performing equidistant segmentation on each UWB echo signal according to a preset interval to obtain echo segmentation data; performing standardization processing on the echo segmentation data to obtain echo standardization data; converting the echo standardization data into corresponding tensor data, and performing dimension expansion on the tensor data to obtain the echo tensor data; the dimension size of the echo tensor data is [the total number of echo tensor data in the respiratory signal data set, 1, the length of each echo tensor data, 1]; Based on the DataLoader data loader in pytorch, the corresponding pattern label is added to each echo tensor data to obtain the respiratory signal dataset.

2. The breathing pattern classification method based on residual network as claimed in claim 1, characterized in that: The breathing modes include a breath-holding mode, a chest breathing mode, an abdominal breathing mode and a mixed breathing mode.

3. The breathing pattern classification method based on residual network as claimed in claim 1, characterized in that: The step of normalizing the echo segmentation data to obtain echo normalization data comprises: Selecting the real part of the echo segmentation data to obtain corresponding data to be standardized; The data to be standardized is standardized to obtain the echo standardized data.

4. The breathing pattern classification method based on residual network as claimed in claim 1, characterized in that: The step of training the improved residual network according to the training set to obtain the breathing classification model comprises: The training set is randomly divided into preset batches, and the improved residual network is trained by back propagation supervision using the Adam optimizer to obtain the respiratory classification model.

5. A breathing pattern classification system based on residual network, characterized in that: The system comprises: A data acquisition module is used to preset multiple breathing modes and collect UWB echo signals corresponding to each breathing mode respectively; A data processing module, for establishing a respiratory signal data set according to the UWB echo signals of each respiratory mode, and dividing the respiratory signal data set into a training set and a test set according to a preset ratio; the respiratory signal data set includes echo tensor data and mode labels corresponding to the UWB echo signals of each respiratory mode; A model training module, used to train an improved residual network according to the training set to obtain a breathing classification model; the improved residual network includes an input layer, 4 residual blocks and an output layer connected in sequence; the input layer includes a two-dimensional convolutional layer, a normalization layer, an activation function and a maximum pooling layer connected in sequence; the residual block includes a first convolutional layer block and a second convolutional layer block connected in sequence; the first convolutional layer block includes a first preset number of two-dimensional convolutional layers of different sizes connected in sequence, and corresponding multiple normalization layers; the second convolutional layer block includes a second preset number of two-dimensional convolutional layers of different sizes connected in sequence, and corresponding multiple normalization layers; the output layer includes a global maximum pooling layer, a flattening layer and a linear layer connected in sequence; A pattern recognition module, used for inputting the test set into the respiratory classification model for classification prediction to obtain a respiratory pattern recognition result; Wherein, establishing a breathing signal data set according to the UWB echo signals of each breathing mode includes: Performing data preprocessing on each UWB echo signal to obtain corresponding echo tensor data, including: performing equidistant segmentation on each UWB echo signal according to a preset interval to obtain echo segmentation data; performing standardization processing on the echo segmentation data to obtain echo standardization data; converting the echo standardization data into corresponding tensor data, and performing dimension expansion on the tensor data to obtain the echo tensor data; the dimension size of the echo tensor data is [the total number of echo tensor data in the respiratory signal data set, 1, the length of each echo tensor data, 1]; Based on the DataLoader data loader in pytorch, the corresponding pattern label is added to each echo tensor data to obtain the respiratory signal dataset.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images