Exoskeleton trajectory prediction method, system and equipment based on motion intention recognition

Abstract

The invention relates to the technical field of exoskeleton trajectory prediction, and discloses an exoskeleton trajectory prediction method, system and device based on motion intention recognition. The method comprises the following steps: optimizing initial time domain data of a shoulder angle and a hip joint angle when a target object moves to obtain target time domain data; screening the target time domain data based on a correlation coefficient and an interval coefficient between the shoulder angle and the hip joint angle to obtain sample data; training an LSTM model based on the sample data to obtain a trained LSTM model; according to current three-dimensional coordinate data collected in real time and the trained LSTM model, motion track information is obtained through prediction, and safety evaluation is carried out. According to the method, key frequency domain features are extracted through time domain data optimization, sample data with high relevance are screened out through the relevance coefficient and the interval coefficient, the accuracy and reliability of a data set are improved, a more accurate model can be conveniently trained, motion track information is analyzed, motion details are accurately captured, and the high-altitude operation safety is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of exoskeleton trajectory prediction, and in particular to an exoskeleton trajectory prediction method, system and device based on motion intention recognition. Background Art

[0002] By identifying human motion intentions and predicting the exoskeleton's trajectory, the system provides real-time assistance and adjustments to ensure the safety of aerial work. This approach typically involves collecting motion datasets through motion experiments. The exoskeleton is then abstracted and simplified into a multi-link structure consisting of multiple joints, creating an exoskeleton coordinate matrix. Physical exertion is then calculated, and a prediction model is trained using the motion dataset. This model is then used to predict human motion intentions. This approach incorporates physical exertion during movement, ensuring that prediction accuracy does not decrease over time and more closely reflects the actual effects of ergonomics.

[0003] The above method solves the problem that the prediction method does not take into account the physical exertion during exercise, resulting in a decrease in prediction accuracy over time. However, during operation, the collected time domain data is not processed, and there may be noise in the data that affects the overall quality. At the same time, multiple sets of time domain data are not effectively screened, resulting in the possibility of abnormalities in the constructed sample data, and the accuracy and reliability of the data set cannot be guaranteed, thus affecting the prediction accuracy. Summary of the Invention

[0004] The main purpose of the present invention is to provide an exoskeleton trajectory prediction method, system and device based on motion intention recognition, aiming to solve at least one of the above technical problems.

[0005] To achieve the above objectives, the present invention provides an exoskeleton trajectory prediction method based on motion intention recognition, comprising:

[0006] Optimizing the initial time domain data of the shoulder angle and hip joint angle of the target object during movement to obtain target time domain data;

[0007] Filtering the target time domain data based on the correlation coefficient and the interval coefficient between the shoulder angle and the hip joint angle to obtain sample data;

[0008] Training an LSTM model based on the sample data to obtain a trained LSTM model;

[0009] The motion trajectory information is obtained based on the current three-dimensional coordinate data collected in real time and the trained LSTM model prediction, and a safety assessment is performed based on the motion trajectory information.

[0010] In some embodiments, the optimizing the initial time domain data of the shoulder angle and the hip joint angle of the target object during movement to obtain target time domain data includes:

[0011] Collecting initial three-dimensional coordinate data of the target object during movement, and extracting initial time domain data of the shoulder angle and hip joint angle based on the initial three-dimensional coordinate data;

[0012] Performing Fourier transform on the initial time domain data of the shoulder angle and the hip joint angle to obtain initial frequency domain data;

[0013] Removing high-frequency components from the initial frequency domain data and performing a normalization operation to obtain processed frequency domain data;

[0014] Perform inverse Fourier transform on the processed frequency domain data to obtain target time domain data.

[0015] In some embodiments, performing Fourier transform on the initial time domain data of the shoulder angle and the hip joint angle to obtain initial frequency domain data includes:

[0016] Set the number of data points and sampling times to be extracted;

[0017] The initial time domain data of the shoulder angle and the hip joint angle are sampled respectively according to the number of data points and the number of sampling times to obtain a plurality of sampling points;

[0018] Extracting corresponding frequency domain features from each of the sampling points based on Fourier transform;

[0019] Initial frequency domain data is obtained according to the actual value and amplitude combination of the frequency domain feature.

[0020] In some embodiments, the step of filtering the target time domain data based on the correlation coefficient and the interval coefficient between the shoulder angle and the hip joint angle to obtain sample data includes:

[0021] calculating a correlation coefficient between a shoulder angle and a hip joint angle based on the target time domain data;

[0022] According to the correlation coefficient and the interval coefficient, a group in which the time domain data of the shoulder angle is ahead of the time domain data of the hip joint angle is selected;

[0023] Corresponding sample data is constructed based on the grouping in which the shoulder angle time domain data is ahead of the hip joint angle time domain data.

[0024] In some embodiments, calculating the correlation coefficient between the shoulder angle and the hip joint angle based on the target time domain data includes:

[0025] Obtaining multiple sets of shoulder angle and hip joint angle optimized time domain data according to the target time domain data;

[0026] The average values ​​of the shoulder angle and hip joint angle are calculated based on multiple sets of optimized time domain data of the shoulder angle and hip joint angle;

[0027] The differences between the shoulder angle and the hip joint angle at the same moment and the average value are counted, and the correlation coefficient between the shoulder angle and the hip joint angle is calculated based on the differences.

[0028] In some embodiments, screening out a group in which the time-domain data of the shoulder angle is ahead of the time-domain data of the hip joint angle according to the correlation coefficient and the interval coefficient includes:

[0029] Extracting all time domain data grouped as having a strong correlation degree as candidate data based on a comparison result of the correlation coefficient and a first threshold;

[0030] Based on the interval coefficient and interval analysis algorithm, the time domain data of the shoulder angle and hip joint angle in the same group of the selected data are analyzed, and the interval coefficient when the correlation coefficient is the largest is taken as the optimal interval coefficient;

[0031] The group in which the optimal interval coefficient is greater than zero is regarded as the group in which the time domain data of the shoulder angle is ahead of the time domain data of the hip joint angle.

[0032] In some embodiments, the training of the LSTM model based on the sample data to obtain the trained LSTM model includes:

[0033] Inputting the sample data into the LSTM model to analyze the shoulder angle time domain data in the sample data and output predicted hip joint angle data;

[0034] The model parameters of the LSTM model are updated according to the offset value between the predicted hip joint angle data and the actual hip joint angle data in the sample data to obtain a trained LSTM model.

[0035] In some embodiments, the motion trajectory information is obtained based on the current three-dimensional coordinate data collected in real time and the trained LSTM model prediction, and a safety assessment is performed based on the motion trajectory information, including:

[0036] Acquire current three-dimensional coordinate data, and obtain current shoulder angle time domain data according to the current three-dimensional coordinate data;

[0037] Input the current shoulder angle time domain data into the trained LSTM model for analysis to obtain current hip joint angle data;

[0038] Calculating and analyzing the motion trajectory information of the target object based on the current hip joint angle data;

[0039] The safety factor of the target object is evaluated based on the motion trajectory information and real-time sensor data.

[0040] In addition, to achieve the above-mentioned purpose, the present invention also proposes an exoskeleton trajectory prediction system based on motion intention recognition, comprising:

[0041] A data optimization module is used to optimize the initial time domain data of the shoulder angle and hip joint angle of the target object during movement to obtain target time domain data;

[0042] a sample construction module, configured to filter the target time domain data based on a correlation coefficient and an interval coefficient between the shoulder angle and the hip joint angle to obtain sample data;

[0043] A model training module is used to train an LSTM model based on the sample data to obtain a trained LSTM model;

[0044] The trajectory prediction module is used to obtain motion trajectory information based on the current three-dimensional coordinate data collected in real time and the trained LSTM model, and perform safety assessment based on the motion trajectory information.

[0045] In addition, to achieve the above-mentioned purpose, the present invention also proposes an electronic device, which includes: a memory, a processor, and an exoskeleton trajectory prediction program based on motion intention recognition stored in the memory and executable on the processor, wherein the exoskeleton trajectory prediction program based on motion intention recognition is configured to implement the exoskeleton trajectory prediction method based on motion intention recognition as described above.

[0046] The present invention provides an exoskeleton trajectory prediction method based on motion intention recognition, comprising: optimizing the initial time domain data of the shoulder angle and hip joint angle of the target object during movement to obtain target time domain data; screening the target time domain data based on the correlation coefficient and interval coefficient between the shoulder angle and the hip joint angle to obtain sample data; training an LSTM model based on the sample data to obtain a trained LSTM model; predicting motion trajectory information based on the current three-dimensional coordinate data collected in real time and the trained LSTM model, and performing a safety assessment based on the motion trajectory information. In the present invention, the optimization of time domain data facilitates the extraction of key frequency domain features, helps identify abnormal data, utilizes the correlation coefficient and interval coefficient to screen out sample data with strong correlation, improves the accuracy and reliability of the data set, and helps train a more accurate LSTM model. The motion trajectory information is predicted by combining the current three-dimensional coordinate data collected in real time with the trained LSTM model, and can accurately capture motion details, thereby dynamically analyzing the motion trajectory and timely assessing the safety factor to ensure the personal safety of workers working at heights. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of the structure of an electronic device in the hardware operating environment involved in an embodiment of the present invention;

[0048] Figure 2 This is a flow chart of an embodiment of an exoskeleton trajectory prediction method based on motion intention recognition according to the present invention;

[0049] Figure 3 This is a schematic diagram of a process for optimizing time domain data according to an embodiment of the present invention;

[0050] Figure 4 A schematic diagram of the process of constructing sample data involved in an embodiment of the present invention;

[0051] Figure 5 A schematic diagram of the process of completing model training involved in an embodiment of the present invention;

[0052] Figure 6 A schematic diagram of a process for evaluating a safety factor according to an embodiment of the present invention;

[0053] Figure 7 This is a structural block diagram of an embodiment of an exoskeleton trajectory prediction system based on motion intention recognition according to the present invention.

[0054] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0056] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0057] In addition, the descriptions of "first", "second", etc. in the present invention are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but this must be based on the ability of ordinary technicians in this field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention. It should be understood that the specific embodiments described here are only used to explain the present invention and are not used to limit the present invention.

[0058] Reference Figure 1 , Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiment of the present invention.

[0059] like Figure 1 As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the user interface 1003 may optionally include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (Wireless-Fidelity, Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM memory) or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. The memory 1005 may optionally be a storage device independent of the aforementioned processor 1001.

[0060] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation to the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0061] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and an exoskeleton trajectory prediction program based on motion intention recognition.

[0062] exist Figure 1In the electronic device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device, and the electronic device calls the exoskeleton trajectory prediction program based on motion intention recognition stored in the memory 1005 through the processor 1001 and executes the exoskeleton trajectory prediction method based on motion intention recognition provided by the embodiment of the present invention.

[0063] The present invention proposes an exoskeleton trajectory prediction method, system and device based on motion intention recognition.

[0064] The embodiment of the present invention provides an exoskeleton trajectory prediction method based on motion intention recognition, referring to Figure 2 , Figure 2 2 is a flow chart of an embodiment of an exoskeleton trajectory prediction method based on motion intention recognition according to the present invention.

[0065] like Figure 2 As shown, the exoskeleton trajectory prediction method based on motion intention recognition includes:

[0066] Step S100: Optimizing the initial time domain data of the shoulder angle and hip joint angle of the target object during movement to obtain target time domain data;

[0067] Step S200: Filtering the target time domain data based on the correlation coefficient and the interval coefficient between the shoulder angle and the hip joint angle to obtain sample data;

[0068] Step S300: training an LSTM model based on the sample data to obtain a trained LSTM model;

[0069] Step S400: obtaining motion trajectory information based on the current three-dimensional coordinate data collected in real time and the trained LSTM model prediction, and performing a safety assessment based on the motion trajectory information.

[0070] It should be noted that the execution subject in this embodiment may be an electronic device, which may be a computer device with data processing capabilities, or other devices that can achieve the same or similar functions. This embodiment does not limit this. In this embodiment, a computer device is used as an example for illustration.

[0071] In one embodiment, the initial time domain data of the shoulder angle and hip joint angle of the target object during movement are optimized to obtain target time domain data, including: collecting initial three-dimensional coordinate data of the target object during movement, and extracting initial time domain data of the shoulder angle and hip joint angle based on the initial three-dimensional coordinate data; performing Fourier transform on the initial time domain data of the shoulder angle and hip joint angle to obtain initial frequency domain data; removing high-frequency components in the initial frequency domain data and performing a normalization operation to obtain processed frequency domain data; and performing an inverse Fourier transform on the processed frequency domain data to obtain target time domain data.

[0072] In one embodiment, the initial time domain data of the shoulder angle and hip joint angle are subjected to Fourier transform to obtain initial frequency domain data, including: setting the number of extracted data points and the number of sampling times; sampling the initial time domain data of the shoulder angle and hip joint angle respectively according to the number of data points and the number of sampling times to obtain multiple sampling points; extracting corresponding frequency domain features from each of the sampling points based on Fourier transform; and obtaining initial frequency domain data based on the actual value and amplitude combination of the frequency domain features.

[0073] Specifically, optimize time domain data: collect three-dimensional coordinate data (initial three-dimensional coordinate data) of the target object, such as the left and right shoulders, hips and knees of the human body during movement, analyze the initial time domain data of the shoulder angle and the initial time domain data of the hip joint angle based on the initial three-dimensional coordinate data, use Fourier transform to extract the corresponding frequency domain features from the time domain data (initial time domain data of the shoulder angle and hip joint angle), calculate the actual value and amplitude of the frequency domain features, and combine them into frequency domain data (initial frequency domain data), use a filter to remove the high-frequency components in the initial frequency domain data, and then convert the processed frequency domain data into time domain data (target time domain data) through inverse Fourier transform, and store it in a database.

[0074] For example, Figure 3As shown, step S100 includes steps S101 to S105. Among them, step S101, after collecting multiple sets of three-dimensional coordinate data (initial three-dimensional coordinate data) of the left shoulder, right shoulder, hip and knee of the human body during exercise, the corresponding shoulder angle is calculated according to the initial three-dimensional coordinate data of the left and right shoulders during exercise, and the corresponding hip joint angle is analyzed using the initial three-dimensional coordinate data of the hip and knee during exercise. Step S102, after arranging the shoulder angle and the hip joint angle respectively in chronological order, the initial time domain data of the shoulder angle and the initial time domain data of the hip joint angle are obtained, after setting the number of data points extracted per second and the number of sampling times, the initial time domain data of the shoulder angle and the initial time domain data of the hip joint angle are sampled respectively according to the number of data points extracted per second and the number of sampling times, thereby obtaining multiple sampling points. Step S103, using Fourier transform to extract the corresponding frequency domain features from each sampling point, and calculating the actual values ​​of different frequency domain features according to the frequency domain features, the total number of samples and the number of data points extracted per second. Step S104: Determine the amplitude of the frequency domain features and combine them according to the actual values ​​and amplitudes of different frequency domain features to obtain initial frequency domain data, which includes initial frequency domain data of the shoulder angle and initial frequency domain data of the hip joint angle. After obtaining the initial frequency domain data of the shoulder angle and the initial frequency domain data of the hip joint angle, use a filter to analyze the initial frequency domain data, remove high-frequency components in the initial frequency domain data, and perform a normalization operation on the data to obtain processed frequency domain data. Step S105: Convert the processed frequency domain data into time domain data (target time domain data) through an inverse Fourier transform and transfer it to a database for storage.

[0075] Specifically, during human motion, acquisition equipment (such as a motion capture system, an inertial measurement unit, etc.) is used to acquire three-dimensional coordinate data. The three-dimensional coordinate data can accurately describe the position of these body parts in space and reflect the dynamic position changes of these key parts of the human body during motion. Based on the acquired three-dimensional coordinate data of the shoulder and hip-related parts, the values ​​of the shoulder angle and hip joint angle that change with time can be calculated by geometric methods, thereby obtaining the shoulder angle time domain data and the hip joint angle time domain data. For example, for the shoulder angle, the coordinates of the shoulder-related joint points can be selected, and the angle formula of the vector can be used to calculate the shoulder angle at different times.

[0076] Exemplarily, high-frequency components usually represent noise, interference or some unnecessary rapidly changing components in the signal. In this embodiment, a filter is used to remove high-frequency components in the frequency domain data. By removing the high-frequency components, the signal can be smoothed, highlighting the main low-frequency features, making the signal clearer and easier to analyze. The frequency domain data after filtering is converted back to time domain data (target time domain data). The target time domain data obtained by inverse transformation is a smooth signal after removing high-frequency noise interference, which can better reflect the true characteristics of the shoulder and hip joint angle changes during human movement. The entire process of this embodiment converts the original three-dimensional coordinate data of human motion into more valuable, processed time domain data through data acquisition, analysis, transformation and processing, and effectively stores it for subsequent application and research.

[0077] In one embodiment, the target time domain data is filtered based on the correlation coefficient and the interval coefficient between the shoulder angle and the hip joint angle to obtain sample data, including: calculating the correlation coefficient between the shoulder angle and the hip joint angle based on the target time domain data; filtering out the grouping in which the shoulder angle time domain data is ahead of the hip joint angle time domain data according to the correlation coefficient and the interval coefficient; and constructing corresponding sample data according to the grouping in which the shoulder angle time domain data is ahead of the hip joint angle time domain data.

[0078] In one embodiment, the correlation coefficient between the shoulder angle and the hip joint angle is calculated based on the target time domain data, including: obtaining multiple sets of optimized time domain data of shoulder angle and hip joint angle according to the target time domain data; respectively calculating the average values ​​of the shoulder angle and hip joint angle according to the multiple sets of optimized time domain data of shoulder angle and hip joint angle; counting the differences between the shoulder angle and hip joint angle and the average values ​​at the same time, and calculating the correlation coefficient between the shoulder angle and the hip joint angle according to the differences.

[0079] In one embodiment, based on the correlation coefficient and the interval coefficient, a group in which the shoulder angle time domain data is ahead of the hip joint angle time domain data is screened out, including: extracting all time domain data with a strong correlation degree as the selected data based on the comparison result of the correlation coefficient and the first threshold; analyzing the shoulder angle and hip joint angle time domain data in the same group in the selected data based on the interval coefficient and the interval analysis algorithm, and taking the interval coefficient when the correlation coefficient is the largest as the optimal interval coefficient; and taking the group in which the optimal interval coefficient is greater than zero as the group in which the shoulder angle time domain data is ahead of the hip joint angle time domain data.

[0080] Specifically, sample data was constructed: corresponding correlation coefficients were calculated based on multiple groups of shoulder angle and hip joint angle time domain data (optimized target time domain data). After extracting all time domain data grouped as strongly correlated through the correlation coefficients, the optimal interval coefficient for the grouping was determined. The grouping in which the shoulder angle time domain data was ahead of the hip joint angle time domain data was screened out based on the optimal interval coefficient. The corresponding sample data was constructed using the time domain data in the grouping, and the sample data was stored in the original data set.

[0081] For example, Figure 4 As shown, step S200 includes steps S201 to S205. Among them, step S201, according to multiple groups of shoulder angle and hip joint angle time domain data (optimized target time domain data), the average values ​​of the shoulder angle and the hip joint angle are calculated respectively, the difference between the shoulder angle and the hip joint angle at the same time and the average value is counted, and the correlation coefficient between the shoulder angle and the hip joint angle is calculated according to the difference. Step S202, set a first threshold; if the correlation coefficient is lower than or equal to the first threshold, it is determined that the shoulder angle and the hip joint angle in the current group are weakly correlated; if the correlation coefficient is higher than the first threshold, it is determined that the shoulder angle and the hip joint angle in the current group are strongly correlated. S203, after extracting all time domain data grouped as strongly correlated, set the interval coefficient, use the interval analysis algorithm to analyze the time domain data of the shoulder angle and the hip joint angle in the same group, and take the interval coefficient when the correlation coefficient is the largest as the optimal interval coefficient. Step S204: Calculate the optimal interval coefficient corresponding to each group; if the optimal interval coefficient is less than zero, the hip joint angle time domain data in the current group is determined to be ahead of the shoulder angle time domain data; if the optimal interval coefficient is equal to zero, the hip joint angle time domain data and the shoulder angle time domain data in the current group are determined to be synchronized; if the optimal interval coefficient is greater than zero, the shoulder angle time domain data in the current group is determined to be ahead of the hip joint angle time domain data. Step S205: Use the group where the shoulder angle time domain data is ahead of the hip joint angle time domain data to construct corresponding sample data, and store the sample data in the original data set.

[0082] Specifically, the difference between the angles and their average values ​​at the same moment is calculated. This means that at the same point in time, the shoulder and hip angles are subtracted from their respective average values ​​to obtain the corresponding difference, resulting in two sets of difference value sequences. Based on these differences, the correlation coefficient between the shoulder and hip angles is calculated. The correlation coefficient between two variables can be calculated using the Pearson correlation coefficient, which measures the degree of linear correlation between the two variables.

[0083] For example, the time domain data of the shoulder angle and hip joint angle in the same group are analyzed using an interval analysis algorithm. The interval analysis algorithm is specifically as follows:

[0084]

[0085] Where G(ε) represents the interval coefficient representation function, ε represents the interval coefficient, T represents the current time, α(T) represents the shoulder angle time domain data at the Tth time, β(T) represents the hip joint time domain data at the Tth time, β(T+ε) represents the hip joint time domain data at the T+εth time, and N represents the total number of time domain data.

[0086] It should be noted that the interval coefficient can be understood as an interval parameter at a time or data point. When analyzing the time domain data of the shoulder angle and hip joint angle, considering that there may be a certain time delay or phase difference between the two, different interval coefficients are set to adjust the relative position relationship between the data. Usually, a range of interval coefficient values ​​is set. After calculating the corresponding correlation coefficient for each set interval coefficient, these correlation coefficients are compared. The larger the correlation coefficient, the stronger the correlation between the shoulder angle and hip joint angle after adjustment. The interval coefficient corresponding to the maximum correlation coefficient is taken as the optimal interval coefficient. The optimal interval coefficient reflects the optimal time delay or phase difference relationship between the shoulder angle and the hip joint angle.

[0087] In one embodiment, an LSTM model is trained based on the sample data to obtain a trained LSTM model, including: inputting the sample data into the LSTM model to analyze the shoulder angle time domain data in the sample data and outputting predicted hip joint angle data; updating the model parameters of the LSTM model according to the offset value between the predicted hip joint angle data and the actual hip joint angle data in the sample data to obtain a trained LSTM model.

[0088] Specifically, to complete the model training: the original data set is divided into a preset ratio, such as 7:3, to obtain a training set and a test set, and the sample data in the training set is transferred to the LSTM model. When the number of training times reaches the preset value, the sample data in the test set is transferred to the LSTM model for analysis to obtain the predicted value, and the accuracy of the current LSTM model is calculated using the offset value between the predicted value and the actual value to obtain the trained LSTM model.

[0089] For example, Figure 5As shown, step S300 includes steps S301 to S303. In step S301, the original data set is divided according to a preset ratio, such as 7:3, to obtain a training set and a test set, the sample data in the training set is transferred to the LSTM model, and the shoulder angle time domain data in the sample data is analyzed to predict the corresponding hip joint angle data. Step S302, the predicted hip joint angle data and the actual hip joint angle data of the sample data are determined, and the offset value between the predicted hip joint angle data and the actual hip joint angle data is analyzed using a loss function to update the corresponding model parameters, and the number of training times is automatically increased by 1. Step S303, when the number of training times reaches the preset value, the sample data in the test set is transferred to the LSTM model for analysis to obtain a predicted value, and the accuracy of the current LSTM model is calculated using the offset value between the predicted value and the actual value. If the accuracy is lower than the second threshold, the next round of model training is entered. If the accuracy is higher than or equal to the second threshold, the model training is terminated.

[0090] Specifically, the loss function is used to analyze the offset between the predicted hip joint angle data and the actual hip joint angle data. The loss function is specifically:

[0091]

[0092] Where f(s) represents the loss function, m represents the length of the time domain data, t represents the time, g(s+ε) represents the actual hip joint time domain data at the s+εth moment, g′(s) represents the predicted hip joint time domain data at the sth moment, s represents the parameter, represents the weight parameter vector.

[0093] For example, LSTM (Long Short-Term Memory) is a special recurrent neural network RNN ​​that can process sequence data and can effectively solve the problem of gradient vanishing or gradient exploding when traditional RNN processes long sequences. In this embodiment, the shoulder angle time domain data is a time series data, and the LSTM model is suitable for analyzing the shoulder angle time domain data. The sample data in the training set is transferred to the LSTM model. The LSTM model learns the change pattern of the shoulder angle over time through its internal memory unit and gating mechanism, and predicts the corresponding hip joint angle data based on these patterns.

[0094] In one embodiment, motion trajectory information is obtained based on the current three-dimensional coordinate data collected in real time and the trained LSTM model prediction, and a safety assessment is performed based on the motion trajectory information, including: obtaining the current three-dimensional coordinate data, and obtaining current shoulder angle time domain data based on the current three-dimensional coordinate data; inputting the current shoulder angle time domain data into the trained LSTM model for analysis to obtain current hip joint angle data; inferring and analyzing the motion trajectory information of the target object based on the current hip joint angle data; and evaluating the safety factor of the target object based on the motion trajectory information and real-time sensing data.

[0095] Specifically, the safety factor is evaluated: after obtaining the three-dimensional coordinate data of the left and right shoulders of the current user (current three-dimensional coordinate data), the corresponding shoulder angle time domain data is calculated based on the three-dimensional coordinate data of the left and right shoulders, and the data is transferred to the trained LSTM model for analysis to obtain the hip joint angle data. The user's motion trajectory information is analyzed using the hip joint angle data, and after extracting the collected real-time data (such as real-time sensor data), the safety factor of the current user is evaluated through deep learning technology. If the safety factor is lower than the preset value, an emergency prompt message is issued.

[0096] For example, Figure 6 As shown, step S400 includes steps S401 to S403. Among them, step S401, after obtaining the three-dimensional coordinate data of the left and right shoulders of the current user (current three-dimensional coordinate data), the corresponding shoulder angle time domain data is calculated according to the three-dimensional coordinate data of the left and right shoulders, and the corresponding shoulder angle time domain data is transmitted to the trained LSTM model for analysis to obtain hip joint angle data (current hip joint angle data); S402, the three-dimensional coordinate data of the user's hip and knee in the next interval time is calculated using the hip joint angle data, and the user's motion trajectory information is analyzed according to the three-dimensional coordinate data of the user's hip and knee at the current moment and the three-dimensional coordinate data of the user's hip and knee in the next interval time; S403, after extracting the real-time data (real-time sensing data) collected by the depth camera and sensor carried outside the user's body, the safety factor of the current user is evaluated according to the user's motion trajectory information and the collected real-time sensing data through deep learning technology. If the safety factor is lower than the preset value, an emergency prompt message is issued, otherwise no operation is performed.

[0097] In one example, human motion follows certain kinematic laws, and the movement of the hip and knee is closely related to the hip joint angle. Based on the hip joint angle data, combined with a human kinematic model, the three-dimensional coordinate data of the user's hip and knee for the next time interval can be inferred. This typically takes into account factors such as the degrees of freedom of each joint, the connections between joints, and muscle movement, and uses mathematical formulas and algorithms to calculate the change in joint position at different angles. Taking a simple two-dimensional planar linkage model as an example, given the hip joint angle and the current hip and knee positions, the knee position at the new hip joint angle can be calculated using trigonometric relationships. Using the current three-dimensional coordinate data of the user's hip and knee and the inferred three-dimensional coordinate data for the next time interval, the user's motion trajectory can be analyzed. By connecting the hip and knee coordinate points at different times, a series of line segments are formed. The curve formed by these line segments represents the motion trajectory of the user's hip and knee. Motion trajectory information can include trajectory shape, length, direction, speed, and more. For example, by calculating the distance and time interval between two adjacent coordinate points at a given moment, the user's movement speed during that time period can be obtained; by observing the shape of the trajectory, it can be determined whether the user is moving in a straight line, a curve, or performing a specific action (such as turning, jumping, etc.).

[0098] In this embodiment, the time domain data is converted into frequency domain data through Fourier transform, which facilitates the extraction of key frequency domain features and helps identify abnormal data. At the same time, a filter is used to remove high-frequency components, which can effectively reduce noise and improve data quality. The correlation coefficient and the optimal interval coefficient are used to screen out grouped data to ensure that the final sample data has strong correlation, improve the accuracy and reliability of the data set, and help train a more accurate prediction model (trained LSTM model). This method calculates the current shoulder angle data through the three-dimensional coordinates of the shoulder, and combines the trained LSTM model to predict the hip joint angle data. It can accurately capture the user's movement details, thereby dynamically analyzing the movement trajectory, timely evaluating the safety factor, and ensuring the personal safety of people working at heights.

[0099] This embodiment provides an exoskeleton trajectory prediction method based on motion intention recognition, including: optimizing the initial time domain data of the shoulder angle and hip joint angle of the target object during movement to obtain target time domain data; filtering the target time domain data based on the correlation coefficient and interval coefficient between the shoulder angle and the hip joint angle to obtain sample data; training an LSTM model based on the sample data to obtain a trained LSTM model; and obtaining motion trajectory information based on the current three-dimensional coordinate data collected in real time and the trained LSTM model prediction, and performing a safety assessment based on the motion trajectory information. In this embodiment, the optimization of time domain data facilitates the extraction of key frequency domain features, helps identify abnormal data, and uses the correlation coefficient and interval coefficient to filter out sample data with strong correlation, thereby improving the accuracy and reliability of the data set and helping to train a more accurate LSTM model. The motion trajectory information is predicted by combining the current three-dimensional coordinate data collected in real time with the trained LSTM model, which can accurately capture the details of the movement, thereby dynamically analyzing the motion trajectory and timely evaluating the safety factor to ensure the personal safety of workers working at heights.

[0100] In addition, an embodiment of the present invention further proposes a storage medium on which is stored an exoskeleton trajectory prediction program based on motion intention recognition. When the exoskeleton trajectory prediction program based on motion intention recognition is executed by a processor, the steps of the exoskeleton trajectory prediction method based on motion intention recognition as described above are implemented.

[0101] Reference Figure 7 , Figure 7 This is a structural block diagram of an embodiment of an exoskeleton trajectory prediction system based on motion intention recognition according to the present invention.

[0102] like Figure 7 As shown, the exoskeleton trajectory prediction system based on motion intention recognition includes:

[0103] A data optimization module 10 is used to optimize the initial time domain data of the shoulder angle and hip joint angle of the target object during movement to obtain target time domain data;

[0104] A sample construction module 20 is configured to filter the target time domain data based on a correlation coefficient and an interval coefficient between the shoulder angle and the hip joint angle to obtain sample data;

[0105] A model training module 30 is used to train an LSTM model based on the sample data to obtain a trained LSTM model;

[0106] The trajectory prediction module 40 is used to predict motion trajectory information based on the current three-dimensional coordinate data collected in real time and the trained LSTM model, and perform safety assessment based on the motion trajectory information.

[0107] Exemplarily, the exoskeleton trajectory prediction system based on motion intention recognition can serve as a high-altitude safety protection system based on human motion intention recognition or its subsystem, including a data analysis unit, a sample generation unit, a model construction unit and an emergency prompt output unit.

[0108] Specifically, the data optimization module 10 includes a data analysis unit, which is specifically used to collect three-dimensional coordinate data of the left and right shoulders, hips and knees during human movement, analyze the shoulder angle time domain data and hip joint angle time domain data based on the three-dimensional coordinate data, use Fourier transform to extract the corresponding frequency domain features from the time domain data, and calculate the actual value and amplitude of the frequency domain features, combine them into frequency domain data, use a filter to remove the high-frequency components in the frequency domain data, and then convert the frequency domain data into time domain data through inverse Fourier transform, and store it in a database.

[0109] Specifically, the sample construction module 20 includes a sample generation unit, which is specifically used to calculate the corresponding correlation coefficients based on multiple groups of shoulder angle and hip joint angle time domain data, extract all time domain data grouped as strong correlation through the correlation coefficient, determine the optimal interval coefficient for the grouping, and screen out the group in which the shoulder angle time domain data is ahead of the hip joint angle time domain data based on the optimal interval coefficient, use the time domain data in the group to construct the corresponding sample data, and store the sample data in the original data set.

[0110] Specifically, the model training module 30 includes a model construction unit, which is specifically used to divide the original data set into a training set and a test set, and transfer the sample data in the training set to the LSTM model. When the number of training times reaches a preset value, the sample data in the test set is transferred to the LSTM model for analysis to obtain a predicted value, and the accuracy of the current LSTM model is calculated using the offset value between the predicted value and the actual value to obtain a trained LSTM model.

[0111] Specifically, the trajectory prediction module 40 includes an emergency prompt output unit, which is specifically used to obtain the three-dimensional coordinate data of the left and right shoulders of the current user, calculate the corresponding shoulder angle time domain data based on the three-dimensional coordinate data of the left and right shoulders, and transmit it to the trained LSTM model for analysis to obtain hip joint angle data. The hip joint angle data is used to analyze the user's motion trajectory information. After extracting the collected real-time data, the safety factor of the current user is evaluated through deep learning technology. If the safety factor is lower than the preset value, an emergency prompt message is issued.

[0112] In this embodiment, the time domain data is converted into frequency domain data through Fourier transform, which facilitates the extraction of key frequency domain features and helps identify abnormal data. At the same time, a filter is used to remove high-frequency components, which can effectively reduce noise and improve data quality. The correlation coefficient and the optimal interval coefficient are used to screen out grouped data to ensure that the final sample data has strong correlation, improve the accuracy and reliability of the data set, and help train a more accurate prediction model (trained LSTM model). The system calculates the current shoulder angle data through the three-dimensional coordinates of the shoulder, and combines the trained LSTM model to predict the hip joint angle data. It can accurately capture the user's movement details, thereby dynamically analyzing the movement trajectory, timely evaluating the safety factor, and ensuring the personal safety of people working at heights.

[0113] This embodiment provides an exoskeleton trajectory prediction system based on motion intention recognition. In this embodiment, time domain data optimization is used to facilitate the extraction of key frequency domain features, help identify abnormal data, and use correlation coefficients and interval coefficients to screen out sample data with strong correlation, thereby improving the accuracy and reliability of the data set and helping to train a more accurate LSTM model. By combining the current three-dimensional coordinate data collected in real time with the trained LSTM model to predict motion trajectory information, it is possible to accurately capture motion details, thereby dynamically analyzing the motion trajectory and timely evaluating the safety factor to ensure the personal safety of workers working at heights.

[0114] It should be noted that the technical details not fully described in the embodiment of the exoskeleton trajectory prediction system based on motion intention recognition can be referred to the exoskeleton trajectory prediction method based on motion intention recognition as described above provided in any embodiment of the present invention, and will not be repeated here.

[0115] It should be understood that the above is only an example and does not constitute any limitation to the technical solution of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any limitation on this.

[0116] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of it according to actual needs to achieve the purpose of the embodiment scheme, and no limitation is made here.

[0117] In addition, it should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0118] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0119] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, or of course by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as a read-only memory (ROM) / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0120] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting exoskeleton trajectory based on motion intention recognition, characterized in that: include: Optimizing the initial time domain data of the shoulder angle and hip joint angle of the target object during movement to obtain target time domain data; Filtering the target time domain data based on the correlation coefficient and the interval coefficient between the shoulder angle and the hip joint angle to obtain sample data; Training an LSTM model based on the sample data to obtain a trained LSTM model; The motion trajectory information is obtained based on the current three-dimensional coordinate data collected in real time and the trained LSTM model prediction, and a safety assessment is performed based on the motion trajectory information.

2. The method according to claim 1, wherein The step of optimizing the initial time domain data of the shoulder angle and the hip joint angle of the target object during movement to obtain target time domain data includes: Collecting initial three-dimensional coordinate data of the target object during movement, and extracting initial time domain data of the shoulder angle and hip joint angle based on the initial three-dimensional coordinate data; Performing Fourier transform on the initial time domain data of the shoulder angle and the hip joint angle to obtain initial frequency domain data; Removing high-frequency components from the initial frequency domain data and performing a normalization operation to obtain processed frequency domain data; Perform inverse Fourier transform on the processed frequency domain data to obtain target time domain data.

3. The method according to claim 2, wherein The Fourier transform is performed on the initial time domain data of the shoulder angle and the hip joint angle to obtain initial frequency domain data, including: Set the number of data points and sampling times to be extracted; The initial time domain data of the shoulder angle and the hip joint angle are sampled respectively according to the number of data points and the number of sampling times to obtain a plurality of sampling points; Extracting corresponding frequency domain features from each of the sampling points based on Fourier transform; Initial frequency domain data is obtained according to the actual value and amplitude combination of the frequency domain feature.

4. The method according to claim 1, wherein The step of screening the target time domain data based on the correlation coefficient and the interval coefficient between the shoulder angle and the hip joint angle to obtain sample data includes: calculating a correlation coefficient between a shoulder angle and a hip joint angle based on the target time domain data; According to the correlation coefficient and the interval coefficient, a group in which the time domain data of the shoulder angle is ahead of the time domain data of the hip joint angle is selected; Corresponding sample data is constructed based on the grouping in which the shoulder angle time domain data is ahead of the hip joint angle time domain data.

5. The method according to claim 4, wherein The calculating the correlation coefficient between the shoulder angle and the hip joint angle based on the target time domain data includes: Obtaining multiple sets of shoulder angle and hip joint angle optimized time domain data according to the target time domain data; The average values ​​of the shoulder angle and hip joint angle are calculated based on multiple sets of optimized time domain data of the shoulder angle and hip joint angle; The differences between the shoulder angle and the hip joint angle at the same moment and the average value are counted, and the correlation coefficient between the shoulder angle and the hip joint angle is calculated based on the differences.

6. The method according to claim 4, wherein The step of screening out a group in which the time domain data of the shoulder angle is ahead of the time domain data of the hip joint angle according to the correlation coefficient and the interval coefficient comprises: Extracting all time domain data grouped as having a strong correlation degree as candidate data based on a comparison result of the correlation coefficient and a first threshold; Based on the interval coefficient and interval analysis algorithm, the time domain data of the shoulder angle and hip joint angle in the same group of the selected data are analyzed, and the interval coefficient when the correlation coefficient is the largest is taken as the optimal interval coefficient; The group in which the optimal interval coefficient is greater than zero is regarded as the group in which the time domain data of the shoulder angle is ahead of the time domain data of the hip joint angle.

7. The method according to any one of claims 1 to 6, characterized in that The step of training the LSTM model based on the sample data to obtain the trained LSTM model includes: Inputting the sample data into the LSTM model to analyze the shoulder angle time domain data in the sample data and output predicted hip joint angle data; The model parameters of the LSTM model are updated according to the offset value between the predicted hip joint angle data and the actual hip joint angle data in the sample data to obtain a trained LSTM model.

8. The method according to claim 7, wherein The motion trajectory information is obtained based on the current three-dimensional coordinate data collected in real time and the trained LSTM model prediction, and a safety assessment is performed based on the motion trajectory information, including: Acquire current three-dimensional coordinate data, and obtain current shoulder angle time domain data according to the current three-dimensional coordinate data; Inputting the current shoulder angle time domain data into the trained LSTM model for analysis to obtain current hip joint angle data; Calculating and analyzing the motion trajectory information of the target object based on the current hip joint angle data; The safety factor of the target object is evaluated based on the motion trajectory information and real-time sensor data.

9. An exoskeleton trajectory prediction system based on motion intention recognition, characterized in that: include: A data optimization module is used to optimize the initial time domain data of the shoulder angle and hip joint angle of the target object during movement to obtain target time domain data; a sample construction module, configured to filter the target time domain data based on a correlation coefficient and an interval coefficient between the shoulder angle and the hip joint angle to obtain sample data; A model training module is used to train an LSTM model based on the sample data to obtain a trained LSTM model; The trajectory prediction module is used to predict motion trajectory information based on the current three-dimensional coordinate data collected in real time and the trained LSTM model, and perform safety assessment based on the motion trajectory information.

10. An electronic device, characterized in that: The electronic device includes: a memory, a processor, and an exoskeleton trajectory prediction program based on motion intention recognition stored in the memory and executable on the processor, wherein the exoskeleton trajectory prediction program based on motion intention recognition is configured to implement the exoskeleton trajectory prediction method based on motion intention recognition as described in any one of claims 1 to 8.

Images