Motion pattern recognition method and system and related equipment

Abstract

The invention relates to a motion pattern recognition method and system and related equipment. The method comprises the following steps: acquiring multi-channel sensing data in a human body movement process through a plurality of sensors, and preprocessing the multi-channel sensing data; extracting a plurality of features from the preprocessed multi-channel sensing data, and constructing a multi-dimensional feature vector; inputting the multi-dimensional feature vector into an adaptive support vector machine classification model for motion pattern recognition, and outputting a recognition result; wherein the adaptive support vector machine classification model adopts a mixed kernel function, and model parameters are dynamically updated according to the confidence coefficient of an identification result, so that the technical problems of low identification accuracy, poor real-time performance and insufficient adaptability of a traditional motion mode identification method are solved; the technical effect of remarkably improving the human body motion mode recognition capability in a complex scene is achieved.

Description

Technical Field

[0001] This application relates to the fields of intelligent human-computer interaction and wearable computing technology, and in particular to a motion pattern recognition method, system and related equipment. Background Technology

[0002] Currently, mainstream motion pattern recognition methods mainly rely on sensors such as inertial measurement units (IMUs) to collect signals such as acceleration and angular velocity, and then combine them with machine learning algorithms for classification. Common classification algorithms include traditional machine learning methods such as decision trees, K-Nearest Neighbors (KNN), and Support Vector Machines (SVM).

[0003] However, existing technologies rely solely on data from a single sensor (such as a lumbar IMU) or a limited number of channels, failing to effectively integrate information from multiple sensors (such as a hip joint angle encoder). This results in limited feature representation capabilities and difficulty in distinguishing movement patterns with similar actions but significantly different postures (such as going up and down stairs). Traditional methods employ pre-defined fixed feature sets (such as mean, variance, and other time-domain features), lacking adaptability to different individuals and movement habits. Due to significant differences in individual body size, gait, and wearing position, fixed feature weights cannot guarantee consistent recognition performance across all users.

[0004] Therefore, traditional motion pattern recognition methods often cannot adapt well to individual user differences, and feature selection is singular and fixed, failing to make full use of the synergistic effect of multidimensional data, resulting in a significant drop in accuracy in complex environments (such as going up and down stairs), as well as insufficient real-time performance and adaptive capabilities.

[0005] There is currently no effective solution to the above problems. Summary of the Invention

[0006] This application provides a motion pattern recognition method, system, and related equipment to at least solve the technical problems of low recognition accuracy, poor real-time performance, and insufficient adaptability in traditional motion pattern recognition methods.

[0007] According to one aspect of the embodiments of this application, a motion pattern recognition method is provided, comprising: acquiring multi-channel sensing data during human motion through multiple sensors, preprocessing the multi-channel sensing data; extracting multiple features from the preprocessed multi-channel sensing data to construct a multi-dimensional feature vector; inputting the multi-dimensional feature vector into an adaptive support vector machine classification model for motion pattern recognition, and outputting a recognition result; wherein the adaptive support vector machine classification model adopts a hybrid kernel function and dynamically updates the model parameters according to the confidence level of the recognition result.

[0008] Optionally, the plurality of sensors include an inertial measurement unit and a joint angle sensor; wherein the inertial measurement unit includes a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer, used to collect acceleration, angular velocity and magnetic field strength data during human movement; the joint angle sensor is used to collect angle and angular velocity information of the hip joint or knee joint.

[0009] Optionally, the preprocessing includes at least one of the following: using a sliding window to truncate data segments, performing low-pass filtering on acceleration and angular velocity data, performing Kalman filtering on magnetic field strength data, and standardizing all channel data.

[0010] Optionally, the length of the sliding window is dynamically adjusted according to the real-time detected step frequency.

[0011] Optionally, the plurality of features include at least two of the following categories: time-domain features, signal energy features, signal distribution features, frequency-domain features, and vector angle features.

[0012] Optionally, the hybrid kernel function is a linear combination of a radial basis kernel function and a polynomial kernel function, expressed as: The radial basis kernel function is: The polynomial kernel function is: in, Indicates the weighting coefficient; , They represent the first The and the first Feature vectors of each sample; Represents kernel function parameters; Represents a constant term; This represents the order of the polynomial.

[0013] Optionally, the model parameters are dynamically updated based on the confidence level of the recognition result, including: updating the weight coefficients and bias terms, and / or updating the slack variables when the confidence level of the recognition result is lower than a preset threshold; wherein the confidence level of the recognition result is obtained based on the distance mapping from the sample to the decision hyperplane.

[0014] According to another aspect of the embodiments of this application, a motion pattern recognition system is provided, comprising: a data acquisition and preprocessing module, used to acquire multi-channel sensing data during human motion through multiple sensors and preprocess the multi-channel sensing data; a feature vector construction module, used to extract multiple features from the preprocessed multi-channel sensing data and construct a multi-dimensional feature vector; and a motion pattern recognition module, used to input the multi-dimensional feature vector into an adaptive support vector machine classification model for motion pattern recognition and output a recognition result; wherein the adaptive support vector machine classification model adopts a hybrid kernel function and dynamically updates the model parameters according to the confidence level of the recognition result.

[0015] According to another aspect of the embodiments of this application, an electronic device is provided, including: a processor, and a memory storing a program, the program including instructions that, when executed by the processor, cause the processor to perform the motion pattern recognition method described above.

[0016] According to another aspect of the present application, a non-transitory machine-readable medium storing computer instructions is provided, the computer instructions being used to cause a computer to perform the motion pattern recognition method described above.

[0017] In this embodiment, multi-channel sensor data collected from multiple sensors during human movement is preprocessed. Multiple features are extracted from the preprocessed multi-channel sensor data to construct a multi-dimensional feature vector. This multi-dimensional feature vector is then input into an adaptive support vector machine (SVM) classification model for motion pattern recognition, outputting the recognition result. The SVM classification model employs a hybrid kernel function and dynamically updates its parameters based on the confidence level of the recognition result. By fusing multi-sensor data to construct a multi-dimensional feature vector and using an adaptive support vector machine classification model with dynamic parameter updates, combined with a hybrid kernel function and a confidence-driven mechanism, efficient, accurate, and real-time recognition of various motion patterns is achieved. This solves the technical problems of low accuracy, poor real-time performance, and insufficient adaptability in traditional motion pattern recognition methods, significantly improving the ability to recognize human motion patterns in complex scenarios.

[0018] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of the motion pattern recognition method provided in the embodiments of this application; Figure 2 A schematic diagram of a motion pattern recognition system provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While some embodiments of this application are shown in the drawings, it should be understood that embodiments of this application can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the embodiments of this application. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of this application.

[0022] SVM: A supervised learning model based on statistical learning theory. It achieves classification / regression by finding the optimal separating hyperplane. Its core advantages are strong generalization ability and excellent performance in processing high-dimensional data.

[0023] IMU: A sensor device that integrates a three-axis accelerometer (measuring linear acceleration), a three-axis gyroscope (measuring angular velocity), and a three-axis magnetometer (measuring magnetic field strength) to capture the motion and attitude information of equipment in real time.

[0024] Sliding window: A method for segmenting continuous sensor data streams by using a fixed-length window to extract data segments and combining overlap rate control to balance time resolution and computational efficiency.

[0025] Feature weighting: An optimization strategy that assigns differentiated weights to different features. The weight reflects the contribution of the feature to the classification task, which can improve the algorithm's adaptability to individual differences and scene changes.

[0026] SVM kernel function: A mathematical function that implements data space mapping, mapping low-dimensional linearly inseparable data to a high-dimensional linearly separable space. Its core function is to construct the decision boundary for nonlinear classification.

[0027] Z-score standardization: a statistical data preprocessing method, the formula is... ( The mean, (where the standard deviation is the standard deviation), which can eliminate the influence of data units and make different characteristics comparable.

[0028] Slack variables ( ): Variables used in SVM to handle nonlinearly separable problems, allowing some samples to deviate from the optimal hyperplane, through a penalty coefficient. Control the tolerance for deviations from the sample.

[0029] Confidence level ( ): A metric for measuring the reliability of SVM classification results, derived from the distance mapping from the sample to the decision hyperplane, used to determine whether an algorithm parameter update is triggered.

[0030] According to one aspect of the embodiments of this application, a motion pattern recognition method is provided. Figure 1 A flowchart of the motion pattern recognition method provided in the embodiments of this application is shown below. Figure 1 As shown, the method includes the following steps: Step S102: Collect multi-channel sensor data during human movement using multiple sensors, and preprocess the multi-channel sensor data. Step S104: Extract multiple features from the preprocessed multi-channel sensor data and construct a multi-dimensional feature vector; Step S106 involves inputting the multidimensional feature vector into an adaptive support vector machine (SVM) classification model for motion pattern recognition and outputting the recognition result. The SVM classification model employs a hybrid kernel function and dynamically updates its parameters based on the confidence level of the recognition result. The recognition result includes one or more motion patterns such as stopping, walking, running, climbing stairs, and descending stairs. By introducing the adaptive SVM classification model, the algorithm adapts to different individuals, ensuring its accuracy.

[0031] In this embodiment, multi-channel sensor data collected from multiple sensors during human movement is preprocessed. Multiple features are extracted from the preprocessed multi-channel sensor data to construct a multi-dimensional feature vector. This multi-dimensional feature vector is then input into an adaptive support vector machine (SVM) classification model for motion pattern recognition, outputting the recognition result. The SVM classification model employs a hybrid kernel function and dynamically updates its parameters based on the confidence level of the recognition result. By fusing multi-sensor data to construct a multi-dimensional feature vector and using an adaptive support vector machine classification model with dynamic parameter updates, combined with a hybrid kernel function and a confidence-driven mechanism, efficient, accurate, and real-time recognition of various motion patterns is achieved. This solves the technical problems of low accuracy, poor real-time performance, and insufficient adaptability in traditional motion pattern recognition methods, significantly improving the ability to recognize human motion patterns in complex scenarios.

[0032] Optionally, by fusing data from multiple sensors, including inertial measurement units (IMUs) such as triaxial accelerometers, gyroscopes, and magnetometers, as well as hip joint angle encoders, an adaptive feature-weighted SVM classification algorithm is proposed. This algorithm can accurately identify various motion patterns such as stopping, walking, running, climbing stairs, and descending stairs. The method first preprocesses the data by using a sliding window to extract rich time-domain and frequency-domain features. Then, an adaptive weighting strategy is employed to optimize feature importance, and an improved SVM kernel function is used to achieve efficient real-time classification.

[0033] In some embodiments, the multiple sensors include an inertial measurement unit and a joint angle sensor; wherein the inertial measurement unit includes a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer, used to collect acceleration, angular velocity and magnetic field strength data during human movement; the joint angle sensor is used to collect angle and angular velocity information of the hip or knee joint.

[0034] The multi-channel sensor data includes the X, Y, and Z axis linear accelerations (unit: m / s²) output by the triaxial accelerometer, the X, Y, and Z axis angular velocities (unit: rad / s) acquired by the triaxial gyroscope, the geomagnetic field intensity components (unit: μT) acquired by the triaxial magnetometer, and the hip flexion / extension angles, knee flexion / extension angles, and their corresponding angular velocities (unit: ° / s) recorded by the joint angle sensor.

[0035] Optionally, the system includes a waist-worn IMU sensor (containing a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer) and bilateral hip joint angle encoders. The sampling frequency is 50Hz, and the collected data dimensions include, but are not limited to, three-axis acceleration (g), three-axis angular velocity (rad / s), three-axis magnetic field strength (μT), bilateral hip joint angles (°), and bilateral hip joint angular velocities (° / s), totaling 15 data channels. Data is transmitted in real-time to the embedded processing unit via an I2C bus.

[0036] In some embodiments, preprocessing includes at least one of the following: using a sliding window to truncate data segments, performing low-pass filtering on acceleration and angular velocity data, performing Kalman filtering on magnetic field strength data, and standardizing all channel data.

[0037] In this embodiment, when using a sliding window to extract data segments, the window size can be set according to the actual motion cycle characteristics to achieve continuous data coverage and effective extraction. When performing low-pass filtering on acceleration and angular velocity data, a second-order Butterworth filter with a cutoff frequency of 5Hz is selected to filter out high-frequency noise interference. When performing Kalman filtering on magnetic field strength data, the drift error in the magnetic field data is dynamically corrected by establishing state equations and observation equations to suppress environmental interference. When standardizing all channel data, the Z-score standardization method is used to convert each channel data into standard normal distribution data with a mean of 0 and a standard deviation of 1, eliminating the influence of differences in the dimensions of different physical quantities on subsequent feature extraction.

[0038] Optionally, the sliding window for data segmentation can use a fixed-length window of 2.56 seconds (containing 128 samples, corresponding to a sampling frequency of 50Hz), with an overlap rate of 50% (64 samples overlapping between adjacent windows), ensuring temporal continuity while reducing computational redundancy. A second-order Butterworth low-pass filter (cutoff frequency 20Hz) is applied to the acceleration and angular velocity data to remove high-frequency noise; a Kalman filter is applied to the magnetic field strength data to suppress environmental magnetic field interference. The Z-score method is used to standardize all channel data, as shown in the formula: ,in, The mean of the samples within the window. This represents the standard deviation of samples within the window, in order to eliminate dimensional differences in data from different sensors.

[0039] In some embodiments, the length of the sliding window is dynamically adjusted based on the real-time detected step frequency.

[0040] Optionally, the feature window size can be adjusted according to the real-time step frequency. The faster the step frequency, the smaller the window to improve response speed, and the slower the step frequency, the larger the window to ensure the integrity of the features.

[0041] in, , ( Sampling frequency, (This refers to the gait period coefficient). =Real-time cadence estimate ( ); ; , It covers the normal human cadence range.

[0042] In this embodiment, the sliding window dynamic adjustment mechanism avoids feature response delay caused by an excessively large window when the step frequency is too fast, and prevents feature information loss caused by an excessively small window when the step frequency is too slow, thus providing suitable feature data for subsequent motion pattern recognition under different motion states.

[0043] It should be noted that in this embodiment, the output frequency of the recognition result is synchronized with the sampling frequency, both being 50. .

[0044] In some embodiments, the multiple features include at least two of the following categories: time-domain features, signal energy features, signal distribution features, frequency-domain features, and vector angle features.

[0045] Temporal features can characterize the intensity and stability of gait by calculating statistical quantities such as the mean, variance, and standard deviation of the acceleration signal; signal energy features can extract the energy values ​​of the triaxial acceleration signal within a specific frequency band to reflect the differences in energy distribution under different movement modes; signal distribution features can use indices such as interquartile range and range to describe the dispersion of signal data; frequency domain features convert the time-domain signal to the frequency domain through Fourier transform to obtain information such as the dominant frequency and power spectral density of the signal to capture the periodicity of motion; vector angle features measure the change in acceleration direction during gait by calculating the cosine value of the angle between acceleration vectors at different times. These features characterize motion signals from different dimensions, comprehensively reflecting the biomechanical characteristics of the human body in various movement modes such as walking, running, and climbing stairs, providing rich and discriminative input data for subsequent pattern classification.

[0046] Optionally, for the preprocessed 15 channel data, 10 features in 5 categories are extracted: time domain features, signal energy features, signal distribution features, frequency domain features, and vector angle features, as detailed below: Time-domain features are used to reflect the overall statistical characteristics of the data, including the mean, standard deviation, median absolute deviation (MAD), maximum value, and minimum value.

[0047] mean ( ): Calculates the arithmetic mean of the samples within each channel window, reflecting the overall offset of the signal in that channel; in, For the number of samples in the window, These are the sample values ​​of the signal.

[0048] Standard deviation: Calculates the standard deviation of samples within each channel window, reflecting the overall offset of the signal in that channel; Median absolute deviation: Calculates the median absolute deviation of samples within each channel window, reflecting the overall offset of the signal in that channel; in, This represents the median of the signal samples for that channel within the window.

[0049] Maximum value ( max(x) ): Calculate the peak signal value within each channel window, reflecting the extreme values ​​of motion intensity; Minimum value ( ): Calculate the signal valley value within each channel window, reflecting the extreme value of motion intensity; Optionally, the signal energy characteristics include the signal amplitude area and energy; Signal amplitude area ( sma ): Calculates the sum and mean of the three-dimensional signal amplitudes of the accelerometer within the window, reflecting the overall motion activity; Energy: Calculates the mean of the sum of squares of the signals within each channel window; Optionally, the signal distribution characteristics include interquartile range; Interquartile range ( ): Calculate the interquartile range of signals within each channel window; in, It is the first quartile. It is the third quartile.

[0050] Optionally, the frequency domain feature is the maximum frequency index obtained by Fast Fourier Transform (FFT).

[0051] Maximum frequency index: Calculates the frequency position where the signal energy is strongest within each channel window. This value is significantly higher in running mode than in walking mode. in, For FFT coefficients, Frequency index (0-64).

[0052] The vector angle characteristic is the angle between the average body acceleration and the gravity vector.

[0053] The average angle between the vectors, the angle between the body acceleration and the gravity vector ( The body's forward lean angle changes significantly when going up or down stairs; this characteristic can effectively distinguish between different types of people. in, The angle between the average body acceleration and the gravity vector.

[0054] In some embodiments, hybrid kernel functions ( ) is derived from the radial basis kernel function ( ) and polynomial kernel function ( It is composed of linear combinations to improve the nonlinear fitting ability, and the expression is: The radial basis kernel function is: The polynomial kernel function is: in, These represent weighting coefficients, obtained through cross-validation optimization, used to balance the contributions of the two types of kernel functions. ; , They represent the first The and the first Feature vectors of each sample; Represents kernel function parameters, for example, Control the local area of ​​impact; Represents a constant term; To represent the order of a polynomial, for example , Control the order of the polynomial.

[0055] It should be noted that, As a parameter controlling the "locality" or "range of influence" of the model, it determines the extent of the influence of a single training sample. The larger the value, the narrower the "bandwidth" of the radial basis function (RBF) kernel, and the smaller its influence range; The smaller the value, the wider the "bandwidth" of the RBF core and the greater its influence.

[0056] It represents the order of the polynomial; it controls the complexity of the kernel function and the complexity of the global patterns that the model can fit. The larger the value, the higher the dimension of the mapped feature space, the more complex the model, and the more complex the feature interactions it can capture.

[0057] This represents the constant or bias term; it controls the importance of lower-order polynomials (especially monomials) in the kernel function. At the same time, it ensures that even in high-dimensional spaces, the low-dimensional information of the original features (such as first-order and second-order terms) will not be lost.

[0058] In some embodiments, the model parameters are dynamically updated based on the confidence level of the recognition result, including: updating the weight coefficients and bias terms, and / or updating the slack variables when the confidence level of the recognition result is lower than a preset threshold; wherein the confidence level of the recognition result is obtained based on the distance mapping from the sample to the decision hyperplane.

[0059] In practical applications, the distance from a sample to the decision hyperplane can be obtained by calculating the functional margin or geometric margin. For example, when using the geometric margin, a larger value indicates higher confidence in the sample classification, and vice versa. The preset threshold needs to be set in conjunction with the accuracy requirements of the specific motion pattern recognition scenario. It is generally determined through statistical analysis of historical sample data. For example, in multi-pattern recognition tasks such as walking, running, and cycling, the threshold can be initially set to 0.6, and then fine-tuned based on the recognition accuracy on the validation set. When the confidence level is lower than this threshold, the weight coefficients and bias terms can be updated using gradient descent. By minimizing the classification loss function, the parameter values ​​are adjusted to make the decision hyperplane better fit the current sample distribution characteristics. Updating slack variables is to improve the model's tolerance to noisy samples or outliers. For example, for misclassified samples, the corresponding slack variable value can be appropriately increased to balance the model's generalization ability and classification accuracy. Furthermore, the dynamic update process can be combined with real-time model training, performing parameter update operations every fixed number of sample inputs or at fixed time intervals to ensure that the model can adapt to the dynamic characteristics of motion patterns changing over time.

[0060] A confidence-driven SVM update strategy is proposed, where the confidence score is obtained by mapping the distance from the SVM decision function to the hyperplane, and is used to determine the classification reliability. In existing technologies, shallow neural networks are typically used to extract features, such as hybrid models of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTMs). However, these models suffer from complexity and high power consumption, making them difficult to deploy in embedded devices. This application's embodiment uses confidence calculation to decide whether to update the current algorithm parameters, ensuring the algorithm's real-time performance. The algorithm also boasts low power consumption, enabling its deployment in embedded devices.

[0061] The confidence level is obtained by mapping the distance from the sample to the SVM decision hyperplane; the closer the distance, the lower the confidence level (the less reliable the classification). Among them, decision boundary distance ,in For support vector coefficients, For sample labels, For hybrid kernel functions, For bias terms; sensitivity coefficient: (Default value), when hour, Trigger parameter update.

[0062] When confidence level In such cases, a lightweight update process can be initiated to avoid full retraining; for example, the update amount can be calculated by minimizing the loss function. (Weight update amount) and (Bias Update Quantity): in, (Penalty coefficient) As slack variables, (.) represents the kernel function mapping.

[0063] Updated formula: ; .

[0064] For example, for a new input sample, the slack variable (reflecting the degree to which the sample deviates from the hyperplane) is updated according to the following formula: According to another aspect of the embodiments of this application, a motion pattern recognition system is provided. Figure 2 This is a schematic diagram of the motion pattern recognition system provided in the embodiments of this application, such as... Figure 2As shown, the motion pattern recognition system includes: a data acquisition and preprocessing module 202, a feature vector construction module 204, and a motion pattern recognition module 206. The motion pattern recognition system will be described in detail below.

[0065] The data acquisition and preprocessing module 202 is used to acquire multi-channel sensor data during human movement through multiple sensors and to preprocess the multi-channel sensor data. The feature vector construction module 204 is used to extract multiple features from the preprocessed multi-channel sensor data and construct a multi-dimensional feature vector. The motion pattern recognition module 206 is used to input multi-dimensional feature vectors into an adaptive support vector machine classification model for motion pattern recognition and output recognition results. The adaptive support vector machine classification model uses a hybrid kernel function and dynamically updates the model parameters based on the confidence level of the recognition results.

[0066] In this embodiment, the system uses multiple sensors to collect multi-channel sensor data during human movement, preprocesses the multi-channel sensor data, extracts multiple features from the preprocessed multi-channel sensor data to construct a multi-dimensional feature vector, and inputs the multi-dimensional feature vector into an adaptive support vector machine (SVM) classification model for motion pattern recognition, outputting the recognition result. The SVM classification model employs a hybrid kernel function and dynamically updates the model parameters based on the confidence level of the recognition result. By fusing multi-sensor data to construct a multi-dimensional feature vector and using an adaptive support vector machine classification model with dynamic parameter updates, combined with a hybrid kernel function and a confidence-driven mechanism, the system achieves efficient, accurate, and real-time recognition of various motion patterns. This solves the technical problems of low recognition accuracy, poor real-time performance, and insufficient adaptability in traditional motion pattern recognition methods, significantly improving the ability to recognize human motion patterns in complex scenarios.

[0067] It should be noted that the data acquisition and preprocessing module 202, feature vector construction module 204 and motion pattern recognition module 206 mentioned above correspond to steps S102 to S106 in the method embodiment. The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above method embodiment.

[0068] According to another aspect of the embodiments of this application, an electronic device is provided, including: a processor, and a memory storing a program, the program including instructions that, when executed by the processor, cause the processor to perform the motion pattern recognition method described above.

[0069] According to another aspect of the embodiments of this application, a non-transitory machine-readable medium storing computer instructions is provided, the computer instructions being used to cause a computer to perform the motion pattern recognition method described above.

[0070] This application also provides a computer program product, including a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform the methods of this application embodiment.

[0071] refer to Figure 3 The present invention describes a structural block diagram of an electronic device that can serve as a server or client in embodiments of this application, which is an example of a hardware device that can be applied to various aspects of this application. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the application described and / or claimed herein.

[0072] like Figure 3 As shown, the electronic device includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 302 or a computer program loaded from a storage unit 308 into a random access memory (RAM) 303. The RAM 303 may also store various programs and data required for the operation of the electronic device. The computing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0073] Multiple components in the electronic device are connected to I / O interface 305, including: input unit 306, output unit 307, storage unit 308, and communication unit 309. Input unit 306 can be any type of device capable of inputting information into the electronic device. Input unit 306 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of the electronic device. Output unit 307 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 308 may include, but is not limited to, disks and optical discs. Communication unit 309 allows the electronic device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, and wireless communication transceivers, such as Bluetooth devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0074] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, CPUs, graphics processing units (GPUs), various special-purpose artificial intelligence (AI) computing units, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any suitable processors, controllers, microcontrollers, etc. The computing unit 301 performs the various methods and processes described above. For example, in some embodiments, the method embodiments of this application can be implemented as computer programs tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program can be loaded and / or installed on an electronic device via ROM 302 and / or communication unit 309. In some embodiments, the computing unit 301 can be configured to perform the methods described above by any other suitable means (e.g., by means of firmware).

[0075] Computer programs used to implement the methods of the embodiments of this application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0076] In the context of embodiments of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable signal medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0077] It should be noted that the term "comprising" and its variations used in the embodiments of this application are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; and the term "some embodiments" means "at least some embodiments". The modifications of "one" and "multiple" mentioned in the embodiments of this application are illustrative and not restrictive. Those skilled in the art should understand that, unless explicitly indicated otherwise in the context, they should be understood as "one or more".

[0078] The steps described in the method embodiments provided in this application can be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of protection of this application is not limited in this respect.

[0079] The term "embodiment" in this specification refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily imply the same embodiment, nor does it imply independence from or alternative to other embodiments. The various embodiments in this specification are described in a related manner, with reference to each other for similar or identical parts. In particular, for apparatus, device, and system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, and relevant details are referred to in the description of the method embodiments.

[0080] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the appended claims.

Claims

1. A motion pattern recognition method, characterized in that, include: Multi-channel sensor data is collected during human movement using multiple sensors, and the multi-channel sensor data is preprocessed. Multiple features are extracted from the preprocessed multi-channel sensor data to construct a multi-dimensional feature vector; The multidimensional feature vector is input into an adaptive support vector machine classification model for motion pattern recognition, and the recognition result is output. The adaptive support vector machine classification model uses a hybrid kernel function and dynamically updates the model parameters based on the confidence level of the recognition result.

2. The motion pattern recognition method according to claim 1, characterized in that, The plurality of sensors include an inertial measurement unit and a joint angle sensor; wherein, The inertial measurement unit includes a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer, used to collect acceleration, angular velocity, and magnetic field strength data during human movement. The joint angle sensor is used to collect angle and angular velocity information of the hip or knee joint.

3. The motion pattern recognition method according to claim 1, characterized in that, The preprocessing includes at least one of the following: using a sliding window to truncate data segments, performing low-pass filtering on acceleration and angular velocity data, performing Kalman filtering on magnetic field strength data, and standardizing all channel data.

4. The motion pattern recognition method according to claim 3, characterized in that, The length of the sliding window is dynamically adjusted according to the real-time detected step frequency.

5. The motion pattern recognition method according to claim 1, characterized in that, The multiple features include at least two of the following categories: time-domain features, signal energy features, signal distribution features, frequency-domain features, and vector angle features.

6. The motion pattern recognition method according to claim 1, characterized in that, The hybrid kernel function is a linear combination of a radial basis kernel function and a polynomial kernel function, and its expression is: The radial basis kernel function is: The polynomial kernel function is: in, Indicates the weighting coefficient; , They represent the first The and the first Feature vectors of each sample; Represents kernel function parameters; Represents a constant term; This represents the order of the polynomial.

7. The motion pattern recognition method according to any one of claims 1 to 6, characterized in that, The model parameters are dynamically updated based on the confidence level of the recognition results, including: When the confidence level of the recognition result is lower than a preset threshold, the weight coefficients and bias terms are updated, and / or the slack variables are updated; wherein, the confidence level of the recognition result is obtained based on the distance mapping from the sample to the decision hyperplane.

8. A motion pattern recognition system, characterized in that, include: The data acquisition and preprocessing module is used to acquire multi-channel sensor data during human movement through multiple sensors and to preprocess the multi-channel sensor data. The feature vector construction module is used to extract multiple features from preprocessed multi-channel sensor data and construct multi-dimensional feature vectors. The motion pattern recognition module is used to input the multidimensional feature vector into an adaptive support vector machine classification model for motion pattern recognition and output the recognition result; wherein, the adaptive support vector machine classification model adopts a hybrid kernel function and dynamically updates the model parameters according to the confidence of the recognition result.

9. An electronic device, comprising: A processor and a memory storing a program, characterized in that the program includes instructions that, when executed by the processor, cause the processor to perform the motion pattern recognition method according to any one of claims 1 to 7.

10. A non-transitory machine-readable medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the motion pattern recognition method according to any one of claims 1 to 7.

Images