Motion posture recognition method and recognition system based on UWB positioning and tracking

Abstract

The invention relates to the technical field of motion posture recognition, in particular to a motion posture recognition method and recognition system based on UWB positioning and tracking. The method comprises the steps of obtaining UWB positioning data of a target user, collecting motion data of multiple parts of the target user, constructing a motion track, extracting time domain features from the collected motion data, establishing a motion recognition model, outputting and obtaining a motion recognition result including swing arm motion and gait motion, constructing a time synchronization fusion multi-modal correlation model, and performing motion recognition on the motion recognition result. And capturing a dynamic relationship between the motion track and the action recognition result, generating a motion posture and a track result, and transmitting the generated motion posture and track result to upper computer software for analyzing and generating dynamically displayed posture and track information. According to the method, time sequence information is fully utilized to construct a high-precision motion posture and trajectory result, and motion recognition and trajectory tracking can be carried out more accurately.

Description

Technical Field

[0001] The present invention relates to the technical field of motion posture recognition, and in particular to a motion posture recognition method and recognition system based on UWB positioning and tracking. Background Art

[0002] With the help of computer vision or wearable devices, sports posture recognition technology can accurately measure and analyze whether the athlete's movements are correct and standardized by collecting sports action data and using machine learning methods. It can also automatically recognize movements and analyze human movement postures by capturing human movement information. This technology has been widely used in rehabilitation training, health monitoring for the elderly, somatosensory games, film and television dramas, sports, and other fields, providing people with more intelligent, personalized, and convenient services and experiences.

[0003] In the current positioning tracking and motion posture recognition methods based on ultra-wideband (UWB), although the introduction of UWB positioning can provide high-precision spatial position data, it is often difficult to achieve comprehensive motion analysis by relying solely on positioning information or motion recognition data. In addition, existing running posture recognition methods usually only focus on single motion features (such as arm swing or gait), lack multi-part motion fusion and trajectory analysis, and have limited functions. At the same time, in dynamic motion scenes, how to achieve real-time fusion and efficient analysis of multimodal data is still a key challenge that determines system performance and applicability.

[0004] In the prior art, the invention patent with publication number CN110694252A discloses a wireless positioning and tracking system based on UWB positioning, and proposes to realize accurate positioning and tracking of the target object by combining one-dimensional and two-dimensional positioning technologies through the interaction of UWB signals between the base station and the target object. However, this method only focuses on the positioning and tracking of targets based on UWB signals, without visual display, and cannot pay attention to the position of the target object in real time; the invention patent with publication number CN116112868A discloses an indoor positioning device for an exhibition hall based on ultra-wideband positioning technology and its use method, and proposes to realize accurate indoor positioning through UWB base stations, synchronous base stations, computing engines and other equipment by using triangulation positioning technology. However, this method mainly focuses on accurate positioning in the exhibition hall and visualization of personnel positions, and does not involve real-time recognition and analysis of human body movements and postures, and its functional scope is relatively limited; the invention patent with publication number CN110694252A discloses a running posture detection method based on a six-axis sensor, and proposes to collect acceleration and angular velocity data through a six-axis sensor, combine gait cycle detection and eigenvalue calculation, and use a Softmax classifier to identify the running landing mode. However, this method only analyzes gait movements, cannot show the comprehensiveness of human body movements and postures, and does not pay attention to the position and posture changes of the human body during movement, lacking the comprehensive use of multi-dimensional sensor features; the invention patent with publication number CN109256187A discloses a running posture suggestion system and method based on multi-mode motion sensor data, and proposes to collect acceleration and angular velocity data through multi-mode motion sensors, analyze them in combination with historical data, generate running posture suggestions and display them on the display screen. However, this method does not combine positioning information, cannot generate motion trajectories or real-time spatial dynamic analysis, and its functional scope is relatively limited.

[0005] In summary, the existing positioning tracking and motion posture recognition methods only focus on a single data source, such as UWB positioning or sensor technology perception, and lack the comprehensive use of motion data of multiple parts of the human body, and cannot achieve comprehensive motion posture analysis and motion trajectory display. Therefore, it is necessary to combine UWB positioning technology with sensors for multi-modal information fusion for motion posture recognition. Summary of the invention

[0006] The technical problem to be solved by the embodiments of the present invention is to provide a motion posture recognition method and recognition system based on UWB positioning and tracking, so as to solve the problem that comprehensive motion posture analysis and motion trajectory display cannot be achieved in the prior art.

[0007] The present invention discloses a motion posture recognition method based on UWB positioning and tracking, comprising: Obtain the target user's location data based on UWB positioning, and collect the target user's motion data in multiple parts; Constructing a motion trajectory according to the acquired positioning data, and extracting time domain features from the collected motion data; Establishing a motion recognition model according to the extracted time domain features, inputting the collected motion data into the motion recognition model, and outputting motion recognition results including arm swinging motion and gait motion; Constructing a time-synchronized fused multimodal association model according to the motion trajectory and the action recognition result, capturing the dynamic relationship between the motion trajectory and the action recognition result from the multimodal association model, and generating motion posture and trajectory results; The generated motion posture and trajectory results are transmitted to the host computer software for analysis to obtain the dynamically displayed posture and trajectory information.

[0008] Optionally, the acquiring positioning data of the target user based on UWB positioning includes: Based on the coverage scenario of multiple UWB base stations, the time difference between the target user reaching two UWB base stations is calculated according to the UWB positioning signal sent by the device worn by the target user. The function expression for calculating the time difference is:

[0009] In the formula, Indicates the time difference, represents the time when the signal arrives at base station m, represents the time when the signal arrives at base station k; Get the speed at which the target user reaches the UWB base station, and calculate the distance between the target user and the UWB base station based on the time difference. The function expression for calculating the distance is:

[0010] In the formula, Represents the distance difference, Indicates the speed at which the target user reaches the two UWB base stations; The base station location coordinates of all UWB base stations are obtained, a set of equations is constructed according to all the base station location coordinates and the distance between the target user and the corresponding UWB base station, and the least squares method is used to solve the set of equations to obtain the positioning data, wherein the positioning data includes the user location coordinates of the target user at different times.

[0011] Optionally, constructing a motion trajectory according to the acquired positioning data includes: According to the user position coordinates at different times, filtering is performed on all the user position coordinates; The user position coordinates after filtering are sorted in time series, and a trajectory fitting algorithm is used to perform smooth difference processing between the user position coordinates corresponding to adjacent times to generate a continuous trajectory curve to construct the motion trajectory.

[0012] Optionally, establishing an action recognition model according to the extracted time domain features includes: According to the collected motion data, filtering is performed on the motion data, and signal drift correction is performed on the motion data after filtering; Based on signal frame processing, the window length and overlap rate are selected to segment the motion data after signal drift correction to obtain multiple windows containing motion cycle data; Selecting features associated with the arm swing motion and the gait motion from the motion data, and performing statistics on the selected features in the motion cycle data in the window; Calculate the time domain features of the selected features according to the statistical selected features, and extract the time domain features from each of the windows; Filtering sample data from all the extracted time-domain features, and dividing the sample data into a training set and a test set; Select a network architecture to build a basic classification model, input the time domain features in the training set into the basic classification model in batches for training, and obtain the action recognition model; The time domain features in the test set are batch-inputted into the trained action recognition model for testing, and the model parameters of the action recognition model are updated through an iterative optimization algorithm.

[0013] Optionally, constructing a time-synchronous fused multimodal association model according to the motion trajectory and the action recognition result includes: Adding timestamps to the motion trajectory and the action recognition result respectively; The sampling period of the positioning data and the sampling period of the action recognition result are calculated and obtained, and the function expressions of the sampling period of the positioning data and the action recognition result are respectively:

[0014]

[0015] In the formula, Indicates the sampling period of positioning data, represents the sampling frequency of positioning data, Indicates the sampling period of action recognition results, Indicates the sampling frequency of action recognition results; According to the consistency of the sampling period of the positioning data and the sampling period of the action recognition result, the timestamps of the motion trajectory and the action recognition result are time-aligned, and a time synchronization sequence containing multiple time points is generated in the same time window, each of which corresponds to the time-synchronized positioning data and the action recognition result. The function expression of the time synchronization sequence is:

[0016] In the formula, represents a time synchronization sequence, Indicates the nth time point; A multimodal association model of time synchronization fusion is constructed according to the positioning data and the action recognition results corresponding to all time points. The function expression of the multimodal association model is:

[0017] In the formula, It represents the multi-modal information vector generated by the fusion of the positioning data and action recognition results corresponding to the nth time point, represents the action recognition result corresponding to the nth time point, Represents the positioning data corresponding to the nth time point.

[0018] Optionally, the time aligning the timestamps of the motion trajectory and the action recognition result includes: When the sampling period of the positioning data is consistent with the sampling period of the action recognition result, the timestamps of the motion trajectory and the action recognition result are unified to the same reference time and converted into absolute timestamps for alignment; When the sampling period of the positioning data is inconsistent with the sampling period of the action recognition result, the two timestamps closest to a timestamp of the action recognition result are found in the timestamps of the motion trajectory. The function expression for finding the timestamp is:

[0019] In the formula, Indicates the i-th timestamp added to the action recognition result, represents the jth timestamp added to the motion trajectory, Indicates the j+1th timestamp added to the motion trajectory; According to the two timestamps found, the positioning data corresponding to a timestamp of the action recognition result is calculated by linear interpolation. The calculation function expression of the positioning data is:

[0020] In the formula, represents the positioning data corresponding to the i-th timestamp, represents the positioning data corresponding to the j-th timestamp; represents the positioning data corresponding to the j+1th timestamp; According to the action recognition result and the positioning data corresponding to the same timestamp, the motion trajectory and the timestamp of the action recognition result are aligned.

[0021] The present invention also discloses a recognition system, which adopts the above-mentioned motion posture recognition method based on UWB positioning and tracking, and the recognition system comprises: The motion posture tracking module is used to obtain the positioning data of the target user based on UWB positioning, and collect the motion data of multiple parts of the target user; A data processing module, used to construct a motion trajectory according to the acquired positioning data, and extract time domain features from the collected motion data; An action recognition module is used to establish an action recognition model according to the extracted time domain features, input the collected action data into the action recognition model, and output action recognition results including arm swinging actions and gait actions; A joint recognition module, used to construct a time-synchronized fused multimodal association model according to the motion trajectory and the action recognition result, capture the dynamic relationship between the motion trajectory and the action recognition result from the multimodal association model, and generate motion posture and trajectory results; The visualization module is used to transmit the generated motion posture and trajectory results to the host computer software.

[0022] Optionally, the motion posture tracking module is connected to a wearable device, which includes a wearable shell, a display unit arranged on the wearable shell, and a main control chip, a power supply unit, a UWB positioning unit and a posture sensing unit integrated in the wearable shell. The display unit, the power supply unit, the UWB positioning unit and the posture sensing unit are electrically connected to the main control chip respectively, so that the UWB positioning unit collects and sends the UWB positioning signal of the target user, and the posture sensing unit collects and sends the motion data of the wearing part of the target user.

[0023] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the motion posture recognition method based on UWB positioning and tracking are implemented.

[0024] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the motion posture recognition method based on UWB positioning and tracking are implemented.

[0025] Compared with the prior art, the motion posture recognition method and recognition system based on UWB positioning and tracking provided by the embodiments of the present invention have the following beneficial effects: By combining UWB positioning and multi-part motion sensing data, the motion trajectory is constructed based on the acquired target user positioning data. At the same time, the collected motion data is extracted in the time domain, and a motion recognition model is established based on the extracted time domain features. The collected motion data is input into the trained motion recognition model to obtain motion recognition results including arm swinging and gait movements. By constructing a multimodal association model with time synchronization fusion, the dynamic relationship between the motion trajectory and the motion recognition results is captured, and accurate motion posture and trajectory results are generated. In this way, the time series information is fully utilized to construct high-precision motion posture and trajectory results, which can more accurately perform motion recognition and trajectory tracking, and has broad application potential in the fields of sports health monitoring, behavior analysis, and intelligent interaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments, in which: Figure 1 A schematic block diagram of the steps of a motion posture recognition method based on UWB positioning and tracking provided by an embodiment of the present invention; Figure 2 A schematic diagram of a process for collecting target user location data and action data provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of the LMS adaptive filtering algorithm provided by an embodiment of the present invention; Figure 4 A schematic diagram of a window sliding for performing windowing processing on action data provided by an embodiment of the present invention; Figure 5 A schematic diagram of a convolutional neural network algorithm model provided by an embodiment of the present invention; Figure 6 A schematic diagram showing the comparison of the accuracy of the action recognition model updated by the iterative algorithm provided in the embodiment of the present invention; Figure 7 A schematic diagram showing a comparison of training losses of an action recognition model updated by an iterative algorithm according to an embodiment of the present invention; Figure 8 A schematic diagram of the structure of a wearing device provided in an embodiment of the present invention.

[0027] The reference numerals in the figures are: 1. UWB base station; 2. Computing server; 3. Host computer software; 4. Motion posture tracking module; 41. Display unit; 42. Main control chip; 43. Power supply unit; 44. UWB positioning unit; 45. Posture sensing unit. DETAILED DESCRIPTION

[0028] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. Now, in conjunction with the accompanying drawings, the preferred embodiments of the present invention are described in detail.

[0029] The present invention discloses a motion posture recognition method based on UWB positioning and tracking. Figure 1 As shown, including: S1. Acquire the positioning data of the target user based on UWB positioning, and collect the motion data of multiple parts of the target user; S2, constructing a motion trajectory based on the acquired positioning data and extracting time domain features from the collected motion data; S3, establishing a motion recognition model according to the extracted time domain features, inputting the collected motion data into the motion recognition model, and outputting motion recognition results including arm swinging motion and gait motion; S4, constructing a multimodal association model of time synchronization fusion according to the motion trajectory and action recognition results, capturing the dynamic relationship between the motion trajectory and the action recognition results from the multimodal association model, and generating motion posture and trajectory results; S5, transmitting the generated motion posture and trajectory results to the host computer software 3 for parsing and generating posture and trajectory information for dynamic display.

[0030] Through the implementation of the above-mentioned motion posture recognition method embodiment, combined with UWB positioning and multi-part motion sensing data, a motion trajectory is constructed according to the acquired target user positioning data. At the same time, the collected motion data is subjected to time domain feature extraction, a motion recognition model is established according to the extracted time domain features, and the collected motion data is input into the trained motion recognition model to obtain motion recognition results including arm swinging motion and gait motion. By constructing a multimodal association model with time synchronization fusion, the dynamic relationship between the motion trajectory and the motion recognition results is captured, and accurate motion posture and trajectory results are generated. Thus, time series information is fully utilized to construct high-precision motion posture and trajectory results, which can more accurately perform motion recognition and trajectory tracking, and has broad application potential in the fields of sports health monitoring, behavior analysis, and intelligent interaction.

[0031] Furthermore, obtaining the positioning data of the target user based on UWB positioning includes: Based on the coverage scenario of multiple UWB base stations 1, the time difference between the target user arriving at two UWB base stations 1 is calculated according to the UWB positioning signal sent by the device worn by the target user. The function expression for calculating the time difference is:

[0032] In the formula, Indicates the time difference, represents the time when the signal arrives at base station m, represents the time when the signal arrives at base station k; The speed at which the target user reaches UWB base station 1 is obtained, and the distance between the target user and UWB base station 1 is calculated based on the time difference. The function expression for calculating the distance is:

[0033] In the formula, Represents the distance difference, Indicates the speed at which the target user reaches the two UWB base stations; The base station location coordinates of all UWB base stations 1 are obtained, and a set of equations is constructed according to the position coordinates of all base stations and the distance between the target user and the corresponding UWB base station 1. The least squares method is used to solve the set of equations to obtain the positioning data, which includes the user location coordinates of the target user at different times.

[0034] Through the implementation of the above-mentioned motion gesture recognition method embodiment, as Figure 2 As shown, the wearable device worn on the upper arm and ankle of the target user is responsible for collecting the target user's arm swing and gait and other motion data, as well as obtaining the UWB positioning signal of the target user. The wearable device sends and receives UWB positioning signals, establishes communication with the UWB base station 1 arranged at a known fixed position, and uses positioning algorithms such as flight time or arrival time difference to accurately calculate the three-dimensional position coordinates of the target user in space. That is, assuming that there are n UWB base stations 1 with known positions, the position coordinates of the kth UWB base station 1 are , use the distance difference between multiple UWB base stations 1 and the target user to construct an equation group, use the least squares method to solve the equation group, and get the location coordinates of the target user , in order to complete the location tracking of target users.

[0035] Furthermore, a motion trajectory is constructed according to the acquired positioning data, including: According to the user location coordinates at different times, all user location coordinates are filtered; The user position coordinates after filtering are sorted in time series, and a trajectory fitting algorithm is used to perform smooth difference processing between the user position coordinates corresponding to adjacent times to generate a continuous trajectory curve to construct the motion trajectory.

[0036] By implementing the above-mentioned motion gesture recognition method embodiment, the user position coordinates obtained at different times are , preferably using the Kalman filter algorithm for denoising and outlier processing. Then, by sorting the positioning data containing the user's position coordinates in time series, and using the trajectory fitting algorithm to smooth the coordinate points, preferably using cubic spline interpolation to smoothly interpolate between adjacent coordinate points, a continuous trajectory curve is generated to construct the target user's motion trajectory, thereby completing the trajectory fitting.

[0037] Furthermore, establishing an action recognition model according to the extracted time domain features includes: According to the collected motion data, the motion data is filtered, and the signal drift correction is performed on the filtered motion data; Based on signal frame processing, the window length and overlap rate are selected to segment the motion data after signal drift correction to obtain multiple windows containing motion cycle data; Select features associated with the arm swing motion and the gait motion from the motion data, and perform statistics on the selected features in the motion cycle data in the window; The time domain features of the selected features are calculated based on the statistical selected features, and the time domain features are extracted from each window; Filter out sample data from all extracted time-domain features, and divide the sample data into a training set and a test set; Select a network architecture to build a basic classification model, input the time domain features in the training set into the basic classification model in batches for training, and obtain an action recognition model; The time domain features in the test set are batch-inputted into the trained action recognition model for testing, and the model parameters of the action recognition model are updated through an iterative optimization algorithm.

[0038] Through the implementation of the above-mentioned motion posture recognition method embodiment, it is preferred to use the LMS (Least Mean Squares) adaptive filtering algorithm to filter the motion data. The algorithm will filter the motion data according to the output signal. and expected signal The minimum mean square error of the filter is adaptively adjusted. , so that the filtering algorithm can adapt to the time domain changes of the random signal.

[0039] Assuming the input signal for:

[0040] The filter output signal is , is the filter weight coefficient, then:

[0041]

[0042] Where L represents the filter order, express The input signal at time, represents the vector transpose operation, represents the adaptive filter The weight coefficient, represents the index of the sum of weight coefficients, represents the transpose of the filter weight coefficient vector, Represents the last element in the filter weight coefficient vector; Assumptions is the error signal, is the expected signal, then:

[0043] Define the cost function of the filter is the error signal The mean square value of , then:

[0044] Use the steepest descent method to adjust the weight vector along the negative gradient direction of the performance surface , solving the optimal weight vector, we can get:

[0045] The weight vector The update formula is:

[0046] In the formula, , represents the step size factor, E represents the mean square value of the error signal, represents the gradient of the cost function, represents partial derivative.

[0047] As described above, the schematic diagram of the LMS adaptive filtering algorithm provided by the embodiment of the present invention is as follows: Figure 3 shown.

[0048] In addition, considering the continuity, periodicity and integrity of the motion, the window length is preferably set to N = 200 sampling data points, and the window overlap rate is set to 50%. The sliding window example of the windowing process is shown in the figure below: Figure 4As shown. The horizontal axis represents the sampling point, which refers to each time point in the time series data, and each vertical line represents a sampling moment; the vertical axis represents the value of acceleration, that is, the acceleration value of each sampling point has a corresponding height on this axis. The process of sliding the window is to first define the size of a window. For example, the window contains five sampling points, and slide the window from the beginning of the data sequence, moving one sampling point each time. After each window slides, the sampling points in the window form a new data subset. Therefore, the local features of the data sequence can be analyzed by sliding the window. Therefore, by extracting the time domain features of the action data in each window, the differential information between different arm swinging actions can be established. The time domain feature extraction is performed using a statistical method, and the time domain features of each action window data are considered. The features associated with the arm swinging action and the gait action are selected from the action data information, such as: Y-axis acceleration, X-axis angular velocity, Z-axis angular velocity, pitch angle, and roll angle are features. In the process of time domain feature extraction, time domain features such as mean, standard deviation, kurtosis, and skewness are selected for extraction. The algorithm complexity of these time domain features is low and they are easy to implement in wearable devices.

[0049] Preferably, XGBoost (Extreme Gradient Boosting), Bayes (Bayes Classifier), CNN (Convolutional Neural Network), SVM (Support Vector Machine) and other classification machine learning algorithms can be selected as basic classification models to train the training set, and the hyperparameters and model structure of the algorithm can be adjusted to optimize the model performance. The schematic diagram of the convolutional neural network algorithm model is shown in the figure. Figure 5 As shown in the figure, the collected action data is filtered, windowed and feature extracted, and then input into the convolutional neural network algorithm model for training. Among them, the input layer is designed according to the shape of the action data after filtering and other processing. If the data is two-dimensional, the input layer should accept two-dimensional input. The convolution layer extracts local features from the input data by adding one or more convolution layers, and the appropriate convolution kernel size and number can be selected. After each convolution layer, a pooling layer (such as maximum pooling) is added to reduce the spatial dimension of the feature and reduce the amount of calculation. Usually, multiple convolution layers and pooling layers are stacked to form multiple convolution blocks. After the convolution layer, one or more fully connected layers are added to learn high-level features and perform classification. The number of neurons in the fully connected layer depends on the needs of the output layer. For example, for multi-classification problems, the number of neurons in the last layer is equal to the number of categories. After the last fully connected layer, the Softmax function (normalized exponential function) is used to convert the output into a probability distribution. Thus, high-precision action recognition is achieved using the classification algorithm.

[0050] At this time, it is preferred to select 80% of the sample data as the training set and 20% as the test set, with a total of 3,600 training sets and 900 test sets for model training and testing. Commonly used iterative algorithms such as Momentum (Momentum GradientDescen, improved gradient descent method), RMSProp (Root Mean Square Propagation, adaptive learning rate optimization algorithm) and Adam (Adaptive Moment Estimation, adaptive learning rate optimization algorithm) are selected for testing respectively. The accuracy (Accuracy) and training loss (TrainingFunction) of the provided iterative algorithms are compared. Figure 5 and Figure 6 As shown. Through the above steps, a trained action recognition model is finally obtained. Among them, the division of sample data takes into account randomness and representativeness, and then divides the training set and test set to avoid the impact of data distribution deviation on model performance.

[0051] The collected motion data is input into the motion recognition model, and the output is the motion recognition results including arm swinging motion and gait motion. Motion recognition results usually include arm swinging motion and gait motion. For example, arm swinging motion is divided into excessive arm swinging, crossed arm swinging, outward wide arm swinging and standard arm swinging: excessive arm swinging when the arm swings forward too much or the reverse swing arc increases significantly; crossed arm swinging frequently on the midline of the body; outward wide arm swinging when the arm deviates outward by a large amplitude; other situations are standard arm swinging. Gait motions include full sole landing, heel landing and toe landing: full sole landing when the outside of the sole touches the ground at the same time; heel landing first when the heel lands; toe landing when the forefoot touches the ground first and the center of gravity gradually transitions to the heel. The recognition of motion results is completed by dividing the motion categories.

[0052] According to the action recognition results output by the action recognition model, the arm swing action and the gait action can be integrated to obtain twelve action categories, as shown in the action category table in Table 1: Table 1 Action category table

[0053] Furthermore, a multimodal association model of time synchronization fusion is constructed based on the motion trajectory and action recognition results, including: Add timestamps to motion trajectories and action recognition results respectively; Calculate the sampling period for obtaining positioning data and the sampling period for motion recognition results. The function expressions for the sampling period for positioning data and motion recognition results are:

[0054]

[0055] In the formula, Indicates the sampling period of positioning data, represents the sampling frequency of positioning data, Indicates the sampling period of action recognition results, Indicates the sampling frequency of action recognition results; According to the consistency of the sampling period of the positioning data and the sampling period of the action recognition result, the timestamps of the motion trajectory and the action recognition result are time-aligned, and a time synchronization sequence containing multiple time points is generated in the same time window. Each time point corresponds to the time-synchronized positioning data and action recognition result. The function expression of the time synchronization sequence is:

[0056] In the formula, represents a time synchronization sequence, Indicates the nth time point; A multimodal association model with time synchronization fusion is constructed based on the positioning data and action recognition results corresponding to all time points. The function expression of the multimodal association model is:

[0057] In the formula, It represents the multi-modal information vector generated by the fusion of the positioning data and action recognition results corresponding to the nth time point, represents the action recognition result corresponding to the nth time point, Represents the positioning data corresponding to the nth time point.

[0058] Furthermore, the timestamps of the motion trajectory and the action recognition result are time-aligned, including: When the sampling period of the positioning data is consistent with the sampling period of the action recognition result, the timestamps of the motion trajectory and the action recognition result are unified to the same reference time and converted into absolute timestamps for alignment; When the sampling period of the positioning data is inconsistent with the sampling period of the action recognition result, the two timestamps closest to a timestamp of the action recognition result are found in the timestamp of the motion trajectory. The function expression for finding the timestamp is:

[0059] In the formula, Indicates the i-th timestamp added to the action recognition result, represents the jth timestamp added to the motion trajectory, Indicates the j+1th timestamp added to the motion trajectory; According to the two timestamps found, linear interpolation is used to calculate the positioning data corresponding to a timestamp of the action recognition result. The calculation function expression of the positioning data is:

[0060] In the formula, Represents the positioning data corresponding to the i-th timestamp, Represents the positioning data corresponding to the j-th timestamp; Indicates the positioning data corresponding to the j+1th timestamp; According to the action recognition result and positioning data corresponding to the same timestamp, the timestamps of the motion trajectory and the action recognition result are aligned.

[0061] Through the implementation of the above-mentioned motion posture recognition method embodiment, a unified timestamp is added to each set of motion trajectory and action recognition results, and a linear interpolation method is used to process data with different sampling rates, so as to complete the synchronous fusion of motion trajectory and action recognition results in the same time window. And by constructing a multimodal association model based on time series, the motion trajectory and action recognition results are feature fused to capture the dynamic relationship between the motion trajectory and action recognition results, thereby combining timestamp matching and multimodal association model to generate motion posture and trajectory results.

[0062] The present invention also discloses a recognition system, which adopts the above-mentioned motion posture recognition method based on UWB positioning and tracking, and the recognition system includes: The motion posture tracking module 4 is used to obtain the positioning data of the target user based on UWB positioning, and collect the motion data of multiple parts of the target user; A data processing module is used to construct a motion trajectory based on the acquired positioning data and extract time domain features from the collected motion data; The action recognition module is used to establish an action recognition model based on the extracted time domain features, input the collected action data into the action recognition model, and output action recognition results including arm swinging actions and gait actions; The joint recognition module is used to build a multi-modal association model with time synchronization fusion based on the motion trajectory and action recognition results, capture the dynamic relationship between the motion trajectory and action recognition results from the multi-modal association model, and generate motion posture and trajectory results; The visualization module is used to transmit the generated motion posture and trajectory results to the host computer software 3, and the host computer software 3 obtains the dynamically displayed posture and trajectory information through analysis.

[0063] Through the implementation of the above recognition system embodiment, the motion posture tracking module 4, the data processing module, the action recognition module and the joint recognition module are all executed by the computing server 2, and the upper computer software 3 receives the motion posture and trajectory results from the computing server 2 through Ethernet, and dynamically displays the real-time motion posture and trajectory information in a visual way (refer to Figure 2 ).

[0064] Preferably, the computer-side host software 3 is used as a display of posture and trajectory information. Among them, the operation execution process of the host software 3 is preferably: first open the host software 3, perform network settings, including setting the network protocol, local IP (Internet Protocol) and port number, and open UDP Server (User DatagramProtocol Serve, running on a specific host, using network communication protocol to receive and respond to client requests. The host software simultaneously receives network access request frames from the motion posture tracking module 4 and the UWB base station, completes device registration, and sends network access response frames respectively. After the device registration is completed, the motion data and positioning data continuously collected by the motion posture tracking module 4 at the upper arm and ankle of the target user are received in real time, and these data are saved in the database to realize data persistence processing and provide support for the next step of feature extraction and fusion processing. Further, the host software 3 can support two modes: analysis mode and recognition mode. The analysis mode is mainly used to train classifiers and build classification models, and to perform data analysis and model optimization based on sample data screened from the collected motion data; the recognition mode is used for real-time motion recognition and trajectory tracking, that is, the system reads the collected motion data from the database and inputs it into the trained motion recognition model, and finally outputs the classification results and generates motion trajectories. In addition, the host computer software 3 can display posture and trajectory information in real time. In the visualization interface, the terminal user can dynamically observe the recognition results of the swing arm movement and gait movement, as well as the real-time drawing of the motion trajectory. Through UWB positioning data, the system can accurately draw the three-dimensional trajectory of the target user in space, and realize the comprehensive analysis of motion posture and spatial motion. It enables the host computer software 3 to receive and analyze data in real time, dynamically display posture and trajectory information, including line graphs of key parameters and motion path trajectories, and provide precise support for motion training and posture optimization.

[0065] Preferably, the data can also be transmitted to a corresponding mobile phone APP for visual presentation.

[0066] Furthermore, combined with Figure 2 and Figure 8As shown, the motion posture tracking module 4 is connected to a wearable device, which includes a wearable shell, a display unit 41 arranged on the wearable shell, and a main control chip 42, a power supply unit 43, a UWB positioning unit 44 and a posture sensing unit 45 integrated in the wearable shell. The display unit 41, the power supply unit 43, the UWB positioning unit 44 and the posture sensing unit 45 are electrically connected to the main control chip 42 respectively, so that the UWB positioning unit 44 collects and sends the UWB positioning signal of the target user, and the posture sensing unit 45 collects and sends the motion data of the wearing part of the target user.

[0067] Through the implementation of the above recognition system embodiment, the main control chip 42, as the core of the system, is responsible for coordinating the operation between the modules, and connects the UWB positioning unit 44 and the posture sensing unit 45 through the USART (Universal Synchronous and Asynchronous Receiver and Transmitter) interface. The posture sensing unit 45 collects the motion data of the target user's movement in real time, including various parameters such as acceleration and angular velocity, and the UWB positioning unit 44 is responsible for collecting and sending the positioning signal of the target user. The display unit 41 is used to display the collected motion data and processing results in real time. The power supply unit 43 is used to provide stable power to ensure that the motion posture tracking module can operate efficiently when collecting data and wirelessly transmitting. The motion posture tracking module 4 performs wireless data transmission with external devices through the UWB positioning unit 44, combines multi-mode sensor technology with UWB positioning technology, and realizes high-precision motion posture recognition and real-time positioning functions.

[0068] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the motion posture recognition method based on UWB positioning and tracking are implemented.

[0069] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the motion posture recognition method based on UWB positioning and tracking are implemented.

[0070] The present invention is described by flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to specific embodiments. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0071] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0072] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0073] It should be understood that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. For those skilled in the art, the technical solutions described in the above embodiments can be modified, or some of the technical features therein can be replaced by equivalents; and all these modifications and replacements should fall within the protection scope of the present invention.

Claims

1. A motion posture recognition method based on UWB positioning and tracking, characterized in that: The motion posture recognition method based on UWB positioning tracking includes: Obtain the target user's location data based on UWB positioning, and collect the target user's motion data in multiple parts; Constructing a motion trajectory according to the acquired positioning data, and extracting time domain features from the collected motion data; Establishing a motion recognition model according to the extracted time domain features, inputting the collected motion data into the motion recognition model, and outputting motion recognition results including arm swinging motion and gait motion; Constructing a time-synchronized fused multimodal association model according to the motion trajectory and the action recognition result, capturing the dynamic relationship between the motion trajectory and the action recognition result from the multimodal association model, and generating motion posture and trajectory results; The generated motion posture and trajectory results are transmitted to the host computer software for parsing and generating posture and trajectory information for dynamic display.

2. The motion posture recognition method based on UWB positioning tracking according to claim 1 is characterized in that: The method of obtaining the positioning data of the target user based on UWB positioning includes: Based on the coverage scenario of multiple UWB base stations, the time difference between the target user reaching two UWB base stations is calculated according to the UWB positioning signal sent by the device worn by the target user. The function expression for calculating the time difference is: In the formula, Indicates the time difference, represents the time when the signal arrives at base station m, represents the time when the signal arrives at base station k; Get the speed at which the target user reaches the UWB base station, and calculate the distance between the target user and the UWB base station based on the time difference. The function expression for calculating the distance is: In the formula, Represents the distance difference, Indicates the speed at which the target user reaches the two UWB base stations; The base station location coordinates of all UWB base stations are obtained, a set of equations is constructed according to all the base station location coordinates and the distance between the target user and the corresponding UWB base station, and the least squares method is used to solve the set of equations to obtain the positioning data, wherein the positioning data includes the user location coordinates of the target user at different times.

3. The motion posture recognition method based on UWB positioning tracking according to claim 2 is characterized in that: The step of constructing a motion trajectory according to the acquired positioning data comprises: According to the user position coordinates at different times, filtering is performed on all the user position coordinates; The user position coordinates after filtering are sorted in time series, and a trajectory fitting algorithm is used to perform smooth difference processing between the user position coordinates corresponding to adjacent times to generate a continuous trajectory curve to construct the motion trajectory.

4. The motion posture recognition method based on UWB positioning tracking according to claim 1, characterized in that: The step of establishing an action recognition model according to the extracted time domain features comprises: According to the collected motion data, filtering is performed on the motion data, and signal drift correction is performed on the motion data after filtering; Based on signal frame processing, the window length and overlap rate are selected to segment the motion data after signal drift correction to obtain multiple windows containing motion cycle data; Selecting features associated with the arm swing motion and the gait motion from the motion data, and performing statistics on the selected features in the motion cycle data in the window; Calculate the time domain features of the selected features according to the statistical selected features, and extract the time domain features from each of the windows; Filtering sample data from all the extracted time-domain features, and dividing the sample data into a training set and a test set; Select a network architecture to build a basic classification model, input the time domain features in the training set into the basic classification model in batches for training, and obtain the action recognition model; The time domain features in the test set are batch-inputted into the trained action recognition model for testing, and the model parameters of the action recognition model are updated through an iterative optimization algorithm.

5. The motion posture recognition method based on UWB positioning tracking according to claim 1, characterized in that: The step of constructing a time-synchronous fused multimodal association model according to the motion trajectory and the action recognition result includes: Adding timestamps to the motion trajectory and the action recognition result respectively; The sampling period of the positioning data and the sampling period of the action recognition result are calculated and obtained, and the function expressions of the sampling period of the positioning data and the action recognition result are respectively: In the formula, Indicates the sampling period of positioning data, represents the sampling frequency of positioning data, Indicates the sampling period of action recognition results, Indicates the sampling frequency of action recognition results; According to the consistency of the sampling period of the positioning data and the sampling period of the action recognition result, the timestamps of the motion trajectory and the action recognition result are time-aligned, and a time synchronization sequence containing multiple time points is generated in the same time window, each of which corresponds to the time-synchronized positioning data and the action recognition result. The function expression of the time synchronization sequence is: In the formula, represents a time synchronization sequence, Indicates the nth time point; A multimodal association model of time synchronization fusion is constructed according to the positioning data and the action recognition results corresponding to all time points. The function expression of the multimodal association model is: In the formula, It represents the multi-modal information vector generated by the fusion of the positioning data and action recognition results corresponding to the nth time point, represents the action recognition result corresponding to the nth time point, Indicates the positioning data corresponding to the nth time point.

6. The motion posture recognition method based on UWB positioning tracking according to claim 5 is characterized in that: The time aligning of the motion trajectory and the timestamp of the action recognition result includes: When the sampling period of the positioning data is consistent with the sampling period of the action recognition result, the timestamps of the motion trajectory and the action recognition result are unified to the same reference time and converted into absolute timestamps for alignment; When the sampling period of the positioning data is inconsistent with the sampling period of the action recognition result, the two timestamps closest to a timestamp of the action recognition result are found in the timestamps of the motion trajectory. The function expression for finding the timestamp is: In the formula, Indicates the i-th timestamp added to the action recognition result, represents the jth timestamp added to the motion trajectory, Indicates the j+1th timestamp added to the motion trajectory; According to the two timestamps found, the positioning data corresponding to a timestamp of the action recognition result is calculated by linear interpolation. The calculation function expression of the positioning data is: In the formula, represents the positioning data corresponding to the i-th timestamp, represents the positioning data corresponding to the j-th timestamp; represents the positioning data corresponding to the j+1th timestamp; According to the action recognition result and the positioning data corresponding to the same timestamp, the motion trajectory and the timestamp of the action recognition result are aligned.

7. A recognition system, using the motion posture recognition method based on UWB positioning tracking according to any one of claims 1 to 6, characterized in that: The identification system comprises: The motion posture tracking module is used to obtain the positioning data of the target user based on UWB positioning, and collect the motion data of multiple parts of the target user; A data processing module, used to construct a motion trajectory according to the acquired positioning data, and extract time domain features from the collected motion data; An action recognition module is used to establish an action recognition model according to the extracted time domain features, input the collected action data into the action recognition model, and output action recognition results including arm swinging actions and gait actions; A joint recognition module, used to construct a time-synchronized fused multimodal association model according to the motion trajectory and the action recognition result, capture the dynamic relationship between the motion trajectory and the action recognition result from the multimodal association model, and generate motion posture and trajectory results; The visualization module is used to transmit the generated motion posture and trajectory results to the host computer software.

8. The identification system according to claim 7, characterized in that: The motion posture tracking module is connected to a wearable device, which includes a wearable shell, a display unit arranged on the wearable shell, and a main control chip, a power supply unit, a UWB positioning unit and a posture sensing unit integrated in the wearable shell. The display unit, the power supply unit, the UWB positioning unit and the posture sensing unit are electrically connected to the main control chip respectively, so that the UWB positioning unit collects and sends the UWB positioning signal of the target user, and the posture sensing unit collects and sends the motion data of the wearing part of the target user.

9. 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 motion posture recognition method based on UWB positioning tracking according to any one of claims 1 to 6 are implemented.

10. 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 motion posture recognition method based on UWB positioning tracking according to any one of claims 1 to 6 are implemented.

Images