Motion track prediction method and system based on intelligent wearable device

Abstract

The invention provides a motion trail prediction method and system based on an intelligent wearable device, and relates to the field of motion trail prediction, and the method comprises the following steps: S1, detecting the motion space of a detected person and the position change information of the detected person; s2, performing feature extraction on the motion space; s3, detecting skin activity information of a plurality of leg joints of the detected person; s4, analyzing the influence degree of the environment obstruction information on the motion trail of the tested person, and outputting initial trail prediction information; s5, predicting a local state track of the tested person according to the leg joint skin activity information, and outputting local track prediction information; and S6, carrying out comprehensive analysis on the initial track prediction information and the local track prediction information, and outputting a motion track prediction result. The motion trail is comprehensively predicted by analyzing the influence degree of the environment obstruction information on the motion trail of the detected person and detecting the skin activity information of the multiple leg joints of the detected person, and the prediction accuracy of the motion trail is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of motion trajectory prediction, and in particular to a motion trajectory prediction method and system based on a smart wearable device. Background Art

[0002] With the popularity of smart wearable devices, people are paying more and more attention to their health status, and exercise has become an indispensable part of many people's daily life. However, for some sports enthusiasts and medical workers, they hope to better understand their own or their patients' exercise status in order to better formulate exercise or rehabilitation plans. Exercise trajectory prediction mainly collects users' exercise data through smart wearable devices, analyzes and processes the data, and predicts the user's future exercise trajectory. This can help users better understand their exercise status, provide personalized exercise suggestions, and provide users with more accurate health data.

[0003] In the existing motion trajectory prediction technology, the image acquisition method is usually adopted. Although it solves the motion trajectory prediction to a certain extent, there is still room for improvement in the prediction accuracy and it is greatly affected by the environment. Summary of the invention

[0004] The present invention provides a motion trajectory prediction method and system based on a smart wearable device, which is used to solve the problem that the image acquisition method is usually adopted in the prior art. Although the motion trajectory prediction is solved to a certain extent, there is still room for improvement in the prediction accuracy and the system is greatly affected by the environment.

[0005] On the one hand, the present invention provides a motion trajectory prediction method based on a smart wearable device, and the specific steps include: S1: Detect the motion space of the person being measured and the position change information of the person being measured; the position change information includes the position change state and the position unchanged state; S2: Extract features of the motion space, perform environmental obstacle analysis, and output environmental obstacle information; S3: Detect skin activity information of multiple leg joints of the person being tested, including skin stretching information, compression information and torsion information; leg joints include hip joints, knee joints and ankle joints; S4: Analyze the influence of environmental obstacles on the motion trajectory of the person being measured, and output initial trajectory prediction information; S5: predicting the local state trajectory of the person under test according to the leg joint skin activity information, and outputting the local trajectory prediction information; S6: Perform comprehensive analysis on the initial trajectory prediction information and the local trajectory prediction information, and output the motion trajectory prediction result.

[0006] According to a motion trajectory prediction method based on a smart wearable device provided by the present invention, in step S2, the specific steps of extracting features from the motion space include: S21: collecting image information of the motion space; S22: preprocessing the image information, including image scaling, normalization and image enhancement; S23: Establish an obstacle recognition model based on convolutional neural network; S24: inputting the preprocessed image information into a trained obstacle recognition model for recognition, and outputting an obstacle information recognition result; S25: Dynamically analyze the obstacle information recognition result, and record the movement speed, acceleration and movement mode of the environmental obstacle information.

[0007] According to a motion trajectory prediction method based on a smart wearable device provided by the present invention, in step S23, the specific steps of establishing an obstacle recognition model include: S231: Divide the data set in the motion database into a training set, a validation set and a test set; S232: Use cross entropy loss as the loss function of convolutional neural network; S233: Optimizing convolutional neural networks using the love evolution algorithm; S234: training an obstacle recognition model using the training set; S235: Use the test set to evaluate the model and calculate the model accuracy. When the accuracy on the verification set no longer changes, stop training to obtain a trained obstacle recognition model.

[0008] According to a motion trajectory prediction method based on a smart wearable device provided by the present invention, in step S233, the specific steps of performing hyperparameter optimization include: S2331: Generate an initial love group, where each individual in the group serves as a hyperparameter combination; S2332: Determine the fitness function of the individual; the formula is: ; In the formula, F(x) is the fitness function, N is the size of the love group, i is the index of the individual in the love group, and y i is the true label of the ith data point, To predict the results; S2333: Selection stage, combining the fitness of each individual, using the roulette method to select individuals; S2334: In the dating phase, the individuals selected in the selection phase are paired; high-quality individuals fall in love with each other to generate new individuals; the formula is: ; In the formula, C j is a new individual, j is the position of the gene locus, P1[j] and P2[j] are two parent individuals, and α is the crossover coefficient; S2335: Mutation operation, which mutates the newly generated individuals to increase diversity; the formula for mutation operation is expressed as: ; Where M j is the individual after mutation, β is the random disturbance; S2336: Evaluate the fitness of the new generation of individuals again; S2337: Retain individuals with high fitness and replace individuals with low fitness to form a new generation of population; S2338: Repeat steps S2333-S2337 until the performance of the validation set no longer improves, then stop the iteration and output the individual with the highest fitness found during the iteration as the optimal hyperparameter.

[0009] According to a motion trajectory prediction method based on a smart wearable device provided by the present invention, in step S3, the specific steps of detecting skin activity information of multiple leg joints include: S31: establishing a dot matrix on the skin areas of multiple leg joints of the person being tested; S32: Select multiple key points from the dot matrix as standard points; S33: Calculate the Euclidean distance between the standard points and the dimensionality change of the standard points; the dimensionality change detection method includes: S331: Select one standard point from the standard points of each joint as the initial origin; S332: Record the relative coordinates of the standard point relative to the initial origin when the person being measured is in an upright state as reference coordinates; S333: Calculate the actual coordinates of the standard point when the subject's legs move, and compare them with the reference coordinates to determine the dimensional change of the standard point; S34: Determine stretching information and compression information according to the degree of change of the Euclidean distance; determine torsion information according to the degree of change of the dimension.

[0010] According to a motion trajectory prediction method based on a smart wearable device provided by the present invention, in step S5, the specific steps of performing local state trajectory prediction include: S51: performing correlation matching on the skin activity information of multiple leg joints and the motion information in the motion database to determine the motion patterns corresponding to the different skin activity information; S52: Analyze changes in skin activity over time and identify repetitive movement patterns; S53: determining the change information of the movement amplitude of the local movement through the stretching information and the compression information; determining the change information of the direction of the local movement through the torsion information; S54: Integrate the motion amplitude change information and the direction change information, and output local trajectory prediction information.

[0011] According to a motion trajectory prediction method based on a smart wearable device provided by the present invention, in step S54, the specific steps of performing information integration include: S541: Determine movement amplitude change information and direction change information corresponding to different leg joints; S542: extracting the timestamps corresponding to the motion amplitude change information and the direction change information respectively, and determining the action time periods of the motion amplitude change information and the direction change information according to the timestamps; S543: Analyze the consistency of the action time periods corresponding to different leg joints, and output consistency information; S544: Analyze the motion influence between the hip joint, the knee joint, and the ankle joint according to the consistency information, and output motion state information; S545: Outputting local trajectory prediction information according to the motion state information.

[0012] According to a motion trajectory prediction method based on a smart wearable device provided by the present invention, in step S544, the specific steps of performing motion impact analysis include: S5441: extracting change information of each joint and outputting joint state change information; the joint state change information includes: extracting position change information of the knee joint relative to the hip joint and position change information of the ankle joint relative to the hip joint based on the skin activity information of the hip joint; extracting position change information of the ankle joint relative to the knee joint based on the skin activity information of the knee joint; extracting foot state information based on the skin activity information of the ankle joint; S5442: Output motion status information according to joint status change information.

[0013] According to a motion trajectory prediction method based on a smart wearable device provided by the present invention, in step S6, the specific steps of performing comprehensive analysis include: S61: when the position of the person being measured is in a state of changing, outputting initial trajectory prediction information, and adjusting the initial trajectory prediction information according to the local trajectory prediction information; when the position of the person being measured is in a state of not changing, outputting local trajectory prediction information; S62: When the person being measured leaves the motion space, the motion trajectory prediction result is output.

[0014] On the other hand, the present invention also provides a motion trajectory prediction system based on a smart wearable device, comprising: A spatial information acquisition module is used to collect environmental information; A spatial information processing module is used to extract environmental obstacle information from environmental information; An initial trajectory prediction module is used to analyze the correlation between environmental obstacle information and the person being measured and output initial trajectory prediction information; The skin activity information collection module is used to collect skin activity information of multiple leg joints of the tested person and output the joint skin activity information; A local trajectory prediction module is used to predict the local state trajectory of the subject according to the skin activity information of multiple leg joints, and output the local trajectory prediction information; The motion trajectory prediction module is used to perform comprehensive information analysis on the initial trajectory prediction information and the local trajectory prediction information, and output the motion trajectory prediction result.

[0015] The present invention provides a motion trajectory prediction method and system based on a smart wearable device. The method and system output initial trajectory prediction information by analyzing the influence of environmental obstacle information on the motion trajectory of a person under test. The method then detects skin activity information of multiple leg joints of the person under test and predicts the local state trajectory of the person under test based on the skin activity information of the leg joints, and outputs local trajectory prediction information. The two prediction methods are combined to improve the accuracy of motion trajectory prediction and reduce the influence of the environment on the prediction results.

[0016] The present invention provides a motion trajectory prediction method and system based on a smart wearable device. By establishing an obstacle recognition model based on a convolutional neural network and using an evolutionary algorithm to optimize the convolutional neural network, the model's recognition accuracy for obstacle information is improved, thereby improving the robustness of motion trajectory prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 is a flow chart of a motion trajectory prediction method based on a smart wearable device provided in Embodiment 1 of the present invention; Figure 2 It is a structural diagram of a motion trajectory prediction system based on a smart wearable device provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Combine the following Figure 1-Figure 2 The present invention describes a motion trajectory prediction method and system based on a smart wearable device.

[0021] Figure 1 It is a flow chart of a motion trajectory prediction method based on a smart wearable device provided in an embodiment of the present invention.

[0022] like Figure 1 As shown, a motion trajectory prediction method based on a smart wearable device provided by an embodiment of the present invention mainly includes the following steps: S1: Detect the motion space of the person being measured and the position change information of the person being measured. The position change information includes the position change state and the position unchanged state. When detecting the motion space of the person being measured, it is necessary to first determine the spatial range of the person being measured, which is usually detected by motion sensors, cameras, GPS and other devices. After determining the spatial range, the position change information of the person being measured is then determined.

[0023] S2: Extract features from the motion space, perform environmental obstruction analysis, and output environmental obstruction information. Features refer to some objects and their states contained in the motion space, which can be obtained through technical means such as image processing and point cloud processing. Environmental obstruction information refers to features that have a certain obstructive effect on the movement of the person being measured. It usually includes some obstacles and some movable people or objects. The specific steps for extracting features from the motion space include: S21: Collect image information of the motion space. Image data in the motion space is obtained by using image acquisition equipment, such as cameras, drones, etc. These image data contain information such as the position, shape, and color of the object.

[0024] S22: Preprocess the image information, including image scaling, normalization and image enhancement. Preprocessing is used to improve the quality of image data to make it more suitable for feature extraction and recognition. Image scaling is to adjust the image to a uniform size. Normalization is used to eliminate the intensity difference caused by factors such as lighting and shooting angle in the image and improve the contrast of the image. Image enhancement is used to highlight important features such as edges and corners in the image and improve the accuracy of feature extraction.

[0025] S23: Establish an obstacle recognition model based on convolutional neural networks. Convolutional Neural Networks (CNNs) are basically composed of convolutional layers, activation layers, pooling layers, fully connected layers, batch normalization layers, and dropout layers. The convolutional layer is the core of CNN, which extracts the features of the input data through convolution operations. Each convolutional layer consists of multiple convolution kernels, which slide on the input data to generate feature maps. The activation layer usually uses ReLU (Rectified Linear Unit) as the activation function to introduce nonlinearity and help the network learn more complex features. The pooling layer is used to reduce the dimension of the feature map and reduce the amount of calculation while maintaining important features. Common pooling operations include maximum pooling and average pooling. The fully connected layer is located at the end of the network and is used to flatten all feature maps and connect them to the fully connected layer for classification or regression tasks. Batch normalization and dropout layers are used to speed up the training process and reduce overfitting. The specific steps to establish an obstacle recognition model include: S231: Divide the data set in the motion database into a training set, a validation set, and a test set. The training set is used to train the model, the validation set is used to adjust the parameters of the model during the training process, and the test set is used to evaluate the performance of the model. The proportion of each parameter set is usually 70% for the training set, 15% for the validation set, and 15% for the test set.

[0026] S232: Use cross entropy loss as the loss function of the convolutional neural network. Cross entropy loss is used to measure the difference between the model's output and the true label. By minimizing the cross entropy loss, we can train an accurate obstacle recognition model. The formula of the function is expressed as: ; In the formula, M is the total number of samples, which indicates the degree of difference between the model prediction result and the true label. K is the number of categories, a is the sample label, b is the category label, and x is the number of samples. ab is the actual label. When sample a belongs to category b, x ab is 1 if the value is true, otherwise it is 0. is the predicted probability.

[0027] S233: Use the love evolution algorithm to optimize the convolutional neural network. The specific steps for hyperparameter optimization include: S2331: Generate an initial love group, with each individual in the group serving as a hyperparameter combination.

[0028] S2332: Determine the fitness function of the individual. The formula is: ; In the formula, F is the fitness function, N is the size of the love group, i is the index of the individual in the love group, and y iis the true label of the ith data point, For the prediction results.

[0029] S2333: Selection stage, that is, selecting individuals for reproduction based on their fitness. Combining the fitness of each individual, the roulette method is used to select individuals. The roulette formula is expressed as: ; Where P i is the probability of the i-th individual being selected, F i is the fitness of the ith individual, is the sum of the fitness of all individuals in the love group. Then generate a random number r∈[0,1]. Select individuals until the cumulative probability reaches the random number r.

[0030] S2334: In the dating phase, the individuals selected in the selection phase are paired. High-quality individuals fall in love with each other to generate new individuals. The formula is: ; Where C is the new individual, P1 and P2 are two parent individuals, α is the crossover coefficient, and α is a random number between [0,1].

[0031] S2335: Mutation operation, mutating newly generated individuals to increase diversity. The mutation stage simulates gene mutation. Each gene position of the individual is randomly changed. The formula for mutation operation is expressed as: ; Where M j is the individual after mutation. β is a random disturbance, which obeys the normal distribution. The normal distribution formula is expressed as: ; In the formula, x is a random variable, the mean μ is set to 0, and the standard deviation σ is an indicator to measure the width of the data distribution, and its value directly affects the size of the disturbance. The larger the standard deviation, the greater the disturbance and the larger the search space; the smaller the standard deviation, the smaller the disturbance and the smaller the search space.

[0032] S2336: Evaluate the fitness of the new generation of individuals again.

[0033] S2337: Retain individuals with high fitness and replace those with low fitness to form a new generation of population.

[0034] S2338: Repeat steps S2333-S2337 until the performance of the validation set no longer improves, then stop the iteration and output the individual with the highest fitness found during the iteration as the optimal hyperparameter.

[0035] S234: Use the training set to train the obstacle recognition model. Update the weights in each epoch. In one training, the model will traverse the entire training data set once, which is called an epoch. The way to update the weights is expressed as: ; In the formula, θ is the parameter, η is the learning rate, is the gradient of the loss function with respect to the parameters.

[0036] S235: Use the test set to evaluate the model and calculate the model accuracy. When the accuracy on the validation set does not change, the highest accuracy of the model training is obtained. Then stop the training and obtain the trained obstacle recognition model.

[0037] S24: Input the pre-processed image information into the trained obstacle recognition model for recognition, and output the obstacle information recognition result. The recognition result usually includes information such as the type, location, and size of the obstacle. Then, the recognition result is processed by result screening, fusion, optimization, etc. to improve the accuracy and practicality of the recognition result.

[0038] S25: Dynamically analyze the obstacle information recognition results and record the speed, acceleration and movement mode of the environmental obstacle information. After the model recognizes the obstacle information, it dynamically analyzes it by identifying the characteristic changes of the obstacle information. The analysis results include the speed, acceleration, movement mode and distance of the obstacle information from the measured person.

[0039] S3: Detect skin activity information of multiple leg joints of the person being tested, including skin stretching information, compression information, and torsion information. Flexible sensors are usually used to detect skin activity information. Leg joints include hip joints, knee joints, and ankle joints. The specific steps of detecting skin activity information of multiple leg joints include: S31: Establish a dot matrix on the skin area of ​​multiple leg joints of the person being tested. Mark some key points at the leg joints, which can represent the center of the joint or the important position of the bone. Then, establish a dot matrix around these key points, and the size and density of the dot matrix can be determined according to actual needs.

[0040] S32: Select multiple key points from the dot matrix as standard points. When selecting standard points, the range of motion and direction of motion of the joints need to be considered to ensure that the standard points can fully reflect the motion of the joints.

[0041] S33: Calculate the Euclidean distance between the standard points and the dimensionality change of the standard points. The detection method of the dimensionality change includes: S331: Select one standard point from the standard points of each joint as the initial origin.

[0042] S332: Record the relative coordinates of the standard point relative to the initial origin when the person being measured is in an upright state as reference coordinates.

[0043] S333: Calculate the actual coordinates of the standard point when the subject's legs move, and compare them with the reference coordinates to determine the dimensional change of the standard point.

[0044] S34: Determine the stretching information and the compression information according to the degree of change of the Euclidean distance. Determine the torsion information according to the degree of change of the dimension. The calculation method of the Euclidean distance is expressed as: ; In the formula, d(A, B) is the Euclidean distance between points A and B, and (c1, d1, e1) and (c2, d2, e2) are the relative coordinates of A and B. The degree of dimension change is determined based on the change in the relative coordinates between points A and B. If the coordinates in the direction of the skin normal change, it means that the dimension of the standard point has changed, that is, the skin has twisted.

[0045] S4: Analyze the influence of environmental obstruction information on the motion trajectory of the person being measured, and output the initial trajectory prediction information. The formula for analyzing the influence degree is expressed as: ; Where g(E) is the motion direction prediction function, D0 is the impact value of the obstacle in the environmental obstacle information on the motion trajectory of the person being measured; D1 is the degree of influence of the ground conditions, which is divided into influential and uninfluential, and is numerically expressed as D1∈{0,1}; h0 is the intercept, h1 and h2 are the weight coefficients of the obstacle influencing factors and the ground condition influencing factors, respectively, and ε is the error term.

[0046] S5: Predict the local state trajectory of the person under test based on the leg joint skin activity information, and output the local trajectory prediction information. The specific steps of predicting the local state trajectory include: S51: Correlation matching is performed between the skin activity information of multiple leg joints and the motion information in the motion database to determine the motion modes corresponding to the different skin activity information. The motion modes generally include walking, running, jumping, and the like.

[0047] S52: Analyze the changes of skin activity over time and identify repetitive motion patterns. Create a same motion pattern set for the repetitive motion patterns, which includes different skin activity information under the same motion pattern.

[0048] S53: Determine the change information of the movement amplitude of the local movement through the stretching information and the compression information. The movement amplitude of each joint is determined by analyzing the stretching information and the compression information of the skin. The greater the stretching or compression degree, the greater the movement amplitude of the joint. Determine the change information of the direction of the local movement through the torsion information. The skin will twist with the movement of the joint, and this torsion can accurately determine the movement direction of each joint.

[0049] S54: Integrate the motion amplitude change information and the direction change information to output local trajectory prediction information. The specific steps of information integration include: S541: Determine the movement amplitude change information and direction change information corresponding to different leg joints.

[0050] S542: extract the timestamps corresponding to the motion amplitude change information and the direction change information respectively, so as to determine the exact time when each motion amplitude change information and the direction change information occurs. And determine the action time period of the motion amplitude change information and the direction change information according to the timestamp, that is, the process time between the start and end time of the motion amplitude change information and the direction change information.

[0051] S543: Analyze the consistency of the action time periods corresponding to different leg joints, and output consistency information to determine whether there is coordinated movement or mutual influence between the joints.

[0052] S544: Analyze the motion impact between the hip joint, the knee joint, and the ankle joint according to the consistency information, and output motion state information. The specific steps of performing the motion impact analysis include: S5441: Extract the change information of each joint and output the joint state change information. The joint state change information includes: extracting the position change information of the knee joint relative to the hip joint and the position change information of the ankle joint relative to the hip joint based on the skin activity information of the hip joint. Extracting the position change information of the ankle joint relative to the knee joint based on the skin activity information of the knee joint. Extracting the state information of the foot based on the skin activity information of the ankle joint. The hip joint is a ball-and-socket joint that allows the thigh to move in multiple planes, including flexion, extension, adduction, abduction, and rotation. The knee joint mainly performs flexion and extension movements. When walking and running, the knee joint plays a key role in supporting body weight and absorbing shock. When the hip joint flexes, the knee joint also flexes accordingly; when the hip joint extends, the knee joint also extends. The ankle joint is mainly responsible for the dorsiflexion and plantar flexion of the foot. When walking and running, the plantar flexion of the ankle joint helps to push the body forward.

[0053] S5442: Output motion state information according to the joint state change information. That is, the motion state information of the leg corresponding to the motion of each joint. The motion of different joints will produce different motion effects. For example, during walking, when the hip joint on one side flexes, the knee joint and ankle joint on the same side also flex to prepare for the foot to land. Subsequently, when the hip joint extends, the knee joint and ankle joint also extend to push the body forward.

[0054] S545: Outputting local trajectory prediction information according to the motion state information. By analyzing the motion state of each joint, predicting the possible motion state changes of each joint in the next motion cycle, and then predicting the short-term motion trajectory.

[0055] S6: Perform comprehensive analysis on the initial trajectory prediction information and the local trajectory prediction information, and output the motion trajectory prediction result. The specific steps of comprehensive analysis include: S61: When the position of the person being measured is changed, the initial trajectory prediction information is output, and the initial trajectory prediction information is adjusted according to the local trajectory prediction information. When the position of the person being measured is unchanged, the local trajectory prediction information is output.

[0056] S62: When the person being measured leaves the motion space, the motion trajectory prediction result is output. The motion trajectory prediction result is a prediction of the motion trajectory of the person being measured in the next motion space, and the prediction result will be updated as the person being measured moves.

[0057] In summary, the present invention outputs initial trajectory prediction information by analyzing the influence of environmental obstacle information on the motion trajectory of the person being measured; then, by detecting the skin activity information of multiple leg joints of the person being measured, and predicting the local state trajectory of the person being measured based on the skin activity information of the leg joints, the local trajectory prediction information is output; the two prediction methods are combined to improve the accuracy of motion trajectory prediction; and reduce the influence of the environment on the prediction results. By establishing an obstacle recognition model based on a convolutional neural network and using the love evolution algorithm to optimize the convolutional neural network, the recognition accuracy of the model for obstacle information is improved, thereby improving the robustness of motion trajectory prediction.

[0058] Embodiment 2: like Figure 2 As shown, the present invention also provides a motion trajectory prediction system based on a smart wearable device, which can be implemented by using a motion trajectory prediction method based on a smart wearable device as in Example 1. The prediction system includes: The spatial information acquisition module is used to collect environmental information, including but not limited to geographic location, terrain features, obstacle locations and types, etc. This information is usually obtained through sensors.

[0059] The spatial information processing module is used to extract environmental obstacle information from environmental information. The spatial information processing module extracts environmental obstacle information based on the recognition of features in the environment by a convolutional neural network.

[0060] The initial trajectory prediction module is used to analyze the correlation between the environmental obstruction information and the person being measured and output the initial trajectory prediction information. The initial trajectory prediction module uses the environmental obstruction information obtained from the spatial information processing module to predict its possible motion trajectory. This module is based on convolutional neural networks and relative distance calculations to analyze the correlation between environmental obstruction information and the object being measured and generate a preliminary trajectory prediction.

[0061] The skin activity information acquisition module is used to collect skin activity information of multiple leg joints of the tested person and output joint skin activity information. This module monitors and records the skin activity at the leg joints of the human body through sensors attached to the human skin.

[0062] The local trajectory prediction module is used to predict the local state trajectory of the person under test based on the skin activity information of multiple leg joints and output the local trajectory prediction information. This module infers the state of human movement, such as walking, running, jumping, etc., by analyzing the skin activity information of the leg joints.

[0063] The motion trajectory prediction module is used to perform comprehensive information analysis on the initial trajectory prediction information and the local trajectory prediction information, and output the motion trajectory prediction result. When the person under test is in a state of position change, the initial trajectory prediction information is output, and the initial trajectory prediction information is adjusted according to the local trajectory prediction information. When the person under test is in a state of unchanged position, the local trajectory prediction information is output. When the person under test leaves the motion space, the motion trajectory prediction result is output. This motion trajectory prediction result is a prediction of the motion trajectory of the next motion space of the person under test, and the prediction result will be updated as the person under test moves.

[0064] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0065] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A motion trajectory prediction method based on a smart wearable device, characterized in that: The specific steps include: S1: Detecting the motion space of the person being measured and the position change information of the person being measured; the position change information includes a position change state and a position unchanged state; S2: extracting features from the motion space, performing environmental obstacle analysis, and outputting environmental obstacle information; S3: Detecting skin activity information of multiple leg joints of the subject, including skin stretching information, compression information and torsion information; the leg joints include hip joints, knee joints and ankle joints; S4: analyzing the influence of the environmental obstacle information on the motion trajectory of the person being measured, and outputting initial trajectory prediction information; S5: predicting the local state trajectory of the person under test according to the leg joint skin activity information, and outputting local trajectory prediction information; S6: Comprehensively analyzing the initial trajectory prediction information and the local trajectory prediction information, and outputting a motion trajectory prediction result.

2. The motion trajectory prediction method based on a smart wearable device according to claim 1, characterized in that: In step S2, the specific steps of extracting features from the motion space include: S21: collecting image information of the motion space; S22: preprocessing the image information, including image scaling, normalization and image enhancement; S23: Establish an obstacle recognition model based on convolutional neural network; S24: inputting the preprocessed image information into a trained obstacle recognition model for recognition, and outputting an obstacle information recognition result; S25: Dynamically analyzing the obstacle information recognition result, and recording the movement speed, acceleration and movement mode of the environmental obstacle information.

3. The motion trajectory prediction method based on a smart wearable device according to claim 2 is characterized in that: In step S23, the specific steps of establishing the obstacle identification model include: S231: Divide the data set in the motion database into a training set, a validation set and a test set; S232: Using cross entropy loss as the loss function of the convolutional neural network; S233: Optimizing the convolutional neural network using a love evolution algorithm; S234: Using the training set to train the obstacle recognition model; S235: Use the test set to evaluate the model and calculate the model accuracy. When the accuracy on the verification set no longer changes, stop training to obtain a trained obstacle recognition model.

4. The motion trajectory prediction method based on a smart wearable device according to claim 3 is characterized in that: In step S233, the specific steps of performing hyperparameter optimization include: S2331: Generate an initial love group, where each individual in the group serves as a hyperparameter combination; S2332: Determine the fitness function of the individual; the formula is expressed as: ； Where, F(x) is the fitness function, N is the size of the love group, i is the index of the individual in the love group, and y i is the true label of the ith data point, To predict the results; S2333: Selection stage, combining the fitness of each individual, using the roulette method to select individuals; S2334: In the dating stage, the individuals selected in the selection stage are paired; high-quality individuals fall in love with each other to generate new individuals; the formula is expressed as: ； In the formula, C j is the new individual, j is the position of the gene locus, P1[j] and P2[j] are two parent individuals, and α is the crossover coefficient; S2335: Mutation operation, which mutates the newly generated individuals to increase diversity; the formula for mutation operation is expressed as: ； Where M j is the individual after mutation, β is the random disturbance; S2336: Evaluate the fitness of the new generation of individuals again; S2337: Retain individuals with high fitness and replace individuals with low fitness to form a new generation of population; S2338: Repeat steps S2333-S2337 until the performance of the verification set no longer improves, then stop the iteration and output the individual with the highest fitness found during the iteration as the optimal hyperparameter.

5. The motion trajectory prediction method based on a smart wearable device according to claim 1, characterized in that: In step S3, the specific steps of detecting skin activity information of multiple leg joints include: S31: establishing a dot matrix on the skin areas of multiple leg joints of the person being tested; S32: Selecting a plurality of key points from the point matrix as standard points; S33: Calculating the Euclidean distance between the standard points and the dimensional change of the standard points; the method for detecting the dimensional change includes: S331: Select one standard point from the standard points of each joint as the initial origin; S332: Record the relative coordinates of the standard point relative to the initial origin when the person being measured is in an upright state as reference coordinates; S333: Calculate the actual coordinates of the standard point when the subject's legs move, and compare them with the reference coordinates to determine the dimensional change of the standard point; S34: Determine the stretching information and the compression information according to the degree of change of the Euclidean distance; determine the torsion information according to the degree of change of the dimension.

6. A motion trajectory prediction method based on a smart wearable device according to claim 5, characterized in that: In step S5, the specific steps of performing local state trajectory prediction include: S51: performing correlation matching on the plurality of leg joint skin activity information and the motion information in the motion database to determine the motion patterns corresponding to the different skin activity information; S52: Analyze changes in skin activity over time and identify repetitive movement patterns; S53: determining the movement amplitude change information of the local movement through the stretching information and the compression information; determining the direction change information of the local movement through the torsion information; S54: Integrate the motion amplitude change information and the direction change information, and output local trajectory prediction information.

7. The motion trajectory prediction method based on a smart wearable device according to claim 6, characterized in that: In step S54, the specific steps of information integration include: S541: Determine the movement amplitude change information and the direction change information corresponding to different leg joints; S542: extracting timestamps corresponding to the motion amplitude change information and the direction change information respectively, and determining action time periods of the motion amplitude change information and the direction change information according to the timestamps; S543: Analyze the consistency of the action time periods corresponding to different leg joints, and output consistency information; S544: Analyze the motion influence between the hip joint, the knee joint, and the ankle joint according to the consistency information, and output motion state information; S545: Outputting local trajectory prediction information according to the motion state information.

8. The motion trajectory prediction method based on a smart wearable device according to claim 6, characterized in that: In step S544, the specific steps of performing motion impact analysis include: S5441: extracting change information of each joint and outputting joint state change information; the joint state change information includes: extracting position change information of the knee joint relative to the hip joint and position change information of the ankle joint relative to the hip joint according to the skin activity information of the hip joint; extracting position change information of the ankle joint relative to the knee joint according to the skin activity information of the knee joint; extracting foot state information according to the skin activity information of the ankle joint; S5442: Output motion state information according to the joint state change information.

9. The motion trajectory prediction method based on a smart wearable device according to claim 1, characterized in that: In step S6, the specific steps of performing comprehensive analysis include: S61: when the person under test is in the position change state, output the initial trajectory prediction information, and adjust the initial trajectory prediction information according to the local trajectory prediction information; when the person under test is in the position unchanged state, output the local trajectory prediction information; S62: When the person being measured leaves the movement space, output the movement trajectory prediction result.

10. A motion trajectory prediction system based on a smart wearable device, which adopts a motion trajectory prediction method based on a smart wearable device as claimed in any one of claims 1 to 9, characterized in that: The motion trajectory prediction system comprises: A spatial information acquisition module is used to collect environmental information; A spatial information processing module, used for extracting the environmental obstacle information from the environmental information; An initial trajectory prediction module, used to analyze the correlation between the environmental obstacle information and the person being measured, and output initial trajectory prediction information; A skin activity information collection module, used to collect skin activity information of multiple leg joints of the tested person and output the joint skin activity information; A local trajectory prediction module, used to predict the local state trajectory of the subject according to the plurality of leg joint skin activity information, and output local trajectory prediction information; The motion trajectory prediction module is used to perform comprehensive information analysis on the initial trajectory prediction information and the local trajectory prediction information, and output a motion trajectory prediction result.

Images