Gait feature extraction and analysis method and device

Abstract

The invention provides a gait feature extraction and analysis method and device, and the method comprises the steps: obtaining an acceleration original signal collected by a three-axis accelerometer and an angular velocity original signal collected by a three-axis gyroscope in the same time period, and enabling the three-axis accelerometer and the three-axis gyroscope to be placed at preset body parts; respectively cutting the acceleration original signal and the angular velocity original signal into a plurality of signal windows, and taking the signal window in the gait section as a target signal window; extracting an acceleration gait feature, an angular velocity gait feature and a signal feature based on the target signal window; and generating a gait analysis feature matrix according to the acceleration gait feature, the angular velocity gait feature and the signal feature. According to the invention, the multi-dimensional feature data is obtained by analyzing the data acquired by the low-cost acquisition equipment, so that efficient and low-cost gait analysis is realized, and the accuracy of gait analysis is improved.

Description

Technical Field

[0001] The present invention relates to the field of biostatistics and data analysis, and in particular to a gait feature extraction and analysis method and device. Background Art

[0002] Gait analysis is an important tool for early detection of neurodegenerative diseases (such as Parkinson's disease and dementia). Traditional gait analysis methods mainly rely on three-dimensional motion capture systems or gait detection in laboratory environments. Although these methods can provide accurate data, they have many limitations. First, the equipment is expensive and difficult to promote in large-scale screening; second, because the test is usually conducted in a controlled environment, it is impossible to capture the natural gait characteristics of the subjects in daily life. Another common detection method is through specialist examinations in hospitals, but these methods are time-consuming, labor-intensive, and it is difficult to track the progression of the disease in real time.

[0003] In addition, although the existing gait analysis based on wearable devices has reduced costs and improved convenience, it has certain limitations in the extraction of gait features. It only uses simple acceleration data to analyze individual movement intensity, resulting in poor accuracy of gait analysis.

[0004] Therefore, the prior art has defects and needs to be improved and developed. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a gait feature extraction and analysis method and device in view of the above-mentioned defects of the prior art, aiming to solve the problem that the prior art only uses simple acceleration data to analyze individual movement intensity, resulting in poor accuracy of gait analysis.

[0006] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0007] A gait feature extraction and analysis method, wherein the method comprises:

[0008] Acquiring an original acceleration signal collected by a three-axis accelerometer and an original angular velocity signal collected by a three-axis gyroscope in the same period, wherein the three-axis accelerometer and the three-axis gyroscope are placed on a predetermined body part;

[0009] Cutting the acceleration original signal and the angular velocity original signal into a plurality of signal windows respectively, and taking the signal window in the gait segment as the target signal window;

[0010] Extracting acceleration gait features, angular velocity gait features and signal features based on the target signal window;

[0011] A gait analysis feature matrix is ​​generated according to the acceleration gait feature, the angular velocity gait feature and the signal feature.

[0012] In one embodiment of the present application, the acceleration original signal and the angular velocity original signal are respectively cut into a plurality of signal windows, and the signal window in the gait segment is used as the target signal window, including:

[0013] Cutting the acceleration original signal and the angular velocity original signal into a plurality of signal windows respectively;

[0014] Extracting root mean square, range, main frequency, spectrum flatness, signal entropy and mean crossing rate for each of the signal windows;

[0015] Inputting the root mean square, range, main frequency, spectrum flatness, signal entropy and mean crossing rate of each signal window into the trained random forest classification model to obtain output results;

[0016] If the output result is that the signal window is in the gait segment, the signal window is used as the target signal window.

[0017] In one embodiment of the present application, the acceleration gait feature, the angular velocity gait feature and the signal feature are extracted based on the target signal window, including:

[0018] identifying each gait cycle and sub-phase of the gait cycle in each of the target signal windows;

[0019] Based on each gait cycle and sub-stage of the gait cycle in each of the target signal windows, acceleration gait features, angular velocity gait features and signal features are extracted.

[0020] In one embodiment of the present application, identifying each gait cycle and sub-phase of the gait cycle in each of the target signal windows includes:

[0021] Based on the continuous wavelet transform algorithm, the downward peak value of the smoothed vertical acceleration is detected for each target signal window to determine the initial contact, and the upward peak value of the integrated vertical acceleration is detected for each target signal window to determine the final contact;

[0022] Determine each gait cycle and sub-phase of the gait cycle in the target signal window according to the initial contact and the final contact;

[0023] The sub-phases of the gait cycle include: support phase, swing phase, stride phase and half-stride phase.

[0024] In one embodiment of the present application, the acceleration gait characteristics include: at least one of stride duration, step length duration, swing duration, support duration, initial double support, termination double support, double support, single-foot support, step length, stride length, asymmetry of duration, asymmetry of length, walking speed and step frequency;

[0025] The angular velocity gait feature includes: at least one of average absolute angular velocity, angle change and rotation amount;

[0026] The signal characteristics include: at least one of standard deviation, coefficient of variation, clearance factor, absolute change mean, mean crossing rate, skewness, kurtosis, absolute energy, signal entropy, signal root mean square and signal range.

[0027] In one embodiment of the present application, the gait analysis feature matrix also includes: a mean and a standard deviation respectively obtained based on each of the acceleration gait feature, the angular velocity gait feature and the signal feature.

[0028] In one embodiment of the present application, after obtaining the original acceleration signal collected by the three-axis accelerometer in the same time period, the method further includes:

[0029] Calculate the signal vector amplitude based on the acceleration original signal, and convert the signal vector amplitude into;

[0030] Using a sliding window method, the signal vector amplitude is divided into a number of segmented windows;

[0031] Convert several segmented windows into a low-dimensional space;

[0032] A trained Gaussian mixture model is obtained, and a plurality of segmented windows in a low-dimensional space are clustered using the Gaussian mixture model to obtain a clustering probability corresponding to the signal vector amplitude.

[0033] The present application also provides a gait feature extraction and analysis device, wherein the device comprises:

[0034] An acquisition module, used to acquire an original acceleration signal collected by a three-axis accelerometer and an original angular velocity signal collected by a three-axis gyroscope in the same period of time, wherein the three-axis accelerometer and the three-axis gyroscope are used to be placed on a predetermined body part;

[0035] A cutting module, used for cutting the acceleration original signal and the angular velocity original signal into a plurality of signal windows respectively, and taking the signal window in the gait segment as the target signal window;

[0036] An extraction module, used for extracting acceleration gait features, angular velocity gait features and signal features based on the target signal window;

[0037] A generating module is used to generate a gait analysis feature matrix according to the acceleration gait feature, the angular velocity gait feature and the signal feature.

[0038] The present application also provides a terminal, which includes: a memory, a processor, and a gait feature extraction and analysis program stored in the memory and executable on the processor, wherein the gait feature extraction and analysis program implements the steps of the gait feature extraction and analysis method described above when executed by the processor.

[0039] The present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program can be executed to implement the steps of the gait feature extraction and analysis method as described above.

[0040] The present invention provides a gait feature extraction and analysis method and device, the method comprising: obtaining an original acceleration signal collected by a three-axis accelerometer and an original angular velocity signal collected by a three-axis gyroscope in the same period, wherein the three-axis accelerometer and the three-axis gyroscope are used to be placed on a predetermined body part; cutting the original acceleration signal and the original angular velocity signal into a plurality of signal windows respectively, and taking the signal window in the gait segment as a target signal window; extracting acceleration gait features, angular velocity gait features and signal features based on the target signal window; generating a gait analysis feature matrix according to the acceleration gait features, angular velocity gait features and signal features. The present invention obtains multi-dimensional feature data by analyzing the data collected by low-cost acquisition equipment, realizes efficient and low-cost gait analysis, and improves the accuracy of gait analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of a preferred embodiment of the gait feature extraction and analysis method of the present invention;

[0042] Figure 2 is the original acceleration signal of different activities in the present invention;

[0043] Figure 3 is a flow chart of gait data preprocessing in the present invention;

[0044] Figure 4 is a schematic diagram of the gait cycle in the present invention;

[0045] Figure 5 is the clustering ratio of exercise intensity in the training data in the present invention;

[0046] Figure 6 is the clustering ratio of the exercise intensity of the test subjects in the present invention;

[0047] Figure 7It is a flow chart of exercise intensity clustering in the present invention;

[0048] Figure 8 It is a functional principle block diagram of a preferred embodiment of the gait feature extraction and analysis device of the present invention;

[0049] Fig. 9 It is a functional principle block diagram of a preferred embodiment of the terminal in the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0051] The disadvantages of the existing technology include:

[0052] First, high cost: Traditional gait analysis system equipment is expensive and difficult to use on a large scale.

[0053] Second, environmental limitations: it is impossible to obtain natural gait data in daily free life, and there is a lack of analysis of gait behavior in real scenarios.

[0054] Third, insufficient feature extraction: Existing wearable devices mainly rely on simple acceleration data and cannot effectively extract angular velocity features and signal features, which limits the accuracy of gait analysis.

[0055] Fourth, insufficient motion intensity assessment: Existing systems fail to fully utilize higher-level features for accurate motion intensity analysis, resulting in poor performance in diverse motion scenarios.

[0056] The present invention provides a gait analysis platform based on wearable devices, which can collect and analyze gait data for a long time in a free living environment. First, the present invention can automatically extract a variety of gait features, especially angular velocity features and signal features through efficient signal processing algorithms, which provides richer data dimensions for gait abnormality detection and early diagnosis of neurodegenerative diseases. Secondly, by training an independent motion intensity model and utilizing features such as signal vector amplitude (SVM), the present invention can accurately assess the individual's motion intensity, identify the proportion of various daily activity types (such as walking, running, standing, etc.), and perform well in a variety of sports scenarios.

[0057] The present invention combines low-cost, portable wearable devices with advanced signal processing and machine learning algorithms to achieve efficient and accurate gait analysis and exercise intensity assessment. By collecting real gait data in a natural environment, it can improve the accuracy of early disease detection and provide support for large-scale screening and personalized health monitoring.

[0058] See also Figure 1 , Figure 1 Flowchart of the gait feature extraction and analysis method of the present invention. Figure 1 As shown, the gait feature extraction and analysis method described in the embodiment of the present invention includes:

[0059] Step S100: acquiring an original acceleration signal collected by a three-axis accelerometer and an original angular velocity signal collected by a three-axis gyroscope in the same period of time, wherein the three-axis accelerometer and the three-axis gyroscope are used to be placed at a predetermined body part.

[0060] Specifically, the gait feature extraction and analysis method provided in the embodiment of the present application can be implemented based on a gait analysis platform, which includes: a data acquisition module, a data preprocessing module, a gait feature extraction module, and a motion intensity clustering module. The data acquisition module is used to collect gait data using an IMU sensor; the data preprocessing module is used to perform preprocessing operations such as denoising, filtering, and signal segmentation on the collected data; the gait feature extraction module is used to extract acceleration gait features, angular velocity gait features, and signal features to generate a feature matrix; the motion intensity clustering module is used to construct daily activity clusters based on a Gaussian mixture model.

[0061] The measurement device used by the data acquisition module can be an Axivity 6 (AX6) sensor, which is an inertial measurement unit (IMU) that includes a three-axis accelerometer and a three-axis gyroscope. The three-axis accelerometer is mainly used for body movement measurement and can provide the amplitude and direction of acceleration in three-dimensional space; the three-axis gyroscope is used to measure the angular velocity of various parts of the body around a predetermined axis. In order to ensure the normal operation of the sensor, the measurement device also integrates a temperature sensor, an ambient light sensor, a real-time clock, and a lithium polymer battery. The sensor is capable of continuously recording 6-axis IMU data at a frequency of 100Hz for 7 days. The AX6 device is fixed to the participant's torso through a belt. By recording the acceleration changes when the person's heel touches the ground and leaves the ground, the IMU located on the torso can capture the gait cycle and its sub-stages.

[0062] Although the embodiments of the present application are mainly based on the AX6 device for data collection, other similar inertial measurement units (IMU) can also be used as alternative devices, as long as the inertial measurement unit provides data from a three-axis accelerometer and a three-axis gyroscope.

[0063] like Figure 1 As shown, the gait feature extraction and analysis method described in the embodiment of the present invention also includes:

[0064] Step S200: Cut the original acceleration signal and the original angular velocity signal into a plurality of signal windows respectively, and use the signal window in the gait segment as the target signal window.

[0065] In the embodiment of the present application, the step S200 specifically includes:

[0066] Step S210, cutting the acceleration original signal and the angular velocity original signal into a plurality of signal windows respectively;

[0067] Step S220, extracting the root mean square, range, main frequency, spectrum flatness, signal entropy and mean crossing rate for each signal window;

[0068] Step S230, respectively inputting the root mean square, range, main frequency, spectrum flatness, signal entropy and mean crossover rate of each signal window into the trained random forest classification model to obtain an output result;

[0069] Step S240: If the output result is that the signal window is in the gait segment, the signal window is used as a target signal window.

[0070] Specifically, Figure 2 The raw data samples of different activities recorded by AX6 are shown. Since sensor data is often interfered by high-frequency noise, the raw data first needs to be processed by applying a low-pass filter. The embodiment of the present application adopts a fourth-order digital Butterworth low-pass filter with a cutoff frequency of 20 Hz.

[0071] After denoising, the gait segment needs to be distinguished from other activities. First, the original signal is cut into 3-second windows, and features such as root mean square, range, main frequency, spectrum flatness, signal entropy, and mean crossing rate are extracted in each signal window. These signal features are then input as input data into the trained random forest classification model to determine whether the window is in a gait segment.

[0072] The embodiment of the present application preprocesses the data to facilitate the extraction of gait features.

[0073] like Figure 1 As shown, the gait feature extraction and analysis method described in the embodiment of the present invention also includes:

[0074] Step S300: extracting acceleration gait features, angular velocity gait features and signal features based on the target signal window.

[0075] In one embodiment of the present application, step S300 specifically includes:

[0076] Step S310, identifying each gait cycle and sub-phase of the gait cycle in each of the target signal windows;

[0077] Step S320: extracting acceleration gait features, angular velocity gait features and signal features based on each gait cycle and sub-stage of the gait cycle in each of the target signal windows.

[0078] The embodiment of the present application can accurately determine the start and end time of each step in the walking process by identifying each gait cycle in the target signal window, thereby providing an accurate time reference for subsequent gait feature extraction.

[0079] In the embodiment of the present application, the step S310 specifically includes:

[0080] Step S311: Based on the continuous wavelet transform algorithm, the downward peak value of the smoothed vertical acceleration is detected for each target signal window to determine the initial contact, and the upward peak value of the integrated vertical acceleration is detected for each target signal window to determine the final contact;

[0081] Step S312: determine each gait cycle and sub-phase of the gait cycle in the target signal window according to the initial contact and the final contact.

[0082] The sub-phases of the gait cycle include: support phase, swing phase, stride phase and half-stride phase.

[0083] Specifically, after extracting the gait segment, in order to further identify each gait cycle and its sub-phases within the gait segment and extract relevant gait features, it is necessary to detect the initial contact (IC) and final contact (FC) of each gait cycle. Figure 2 From the change in the amplitude of vertical acceleration, it can be seen that when the person is stationary, the accelerometer measures the inertial force of gravity; when the initial contact occurs, the mass block inside the AX6 is affected by the inertial force and experiences hypergravity, which leads to an increase in the amplitude of vertical acceleration, which is reflected in the formation of a downward peak in the graph. This provides a clue for detecting IC and FC. Therefore, the embodiment of the present application utilizes a continuous wavelet transform (CWT) algorithm, which has good robustness in detecting IC and FC. By applying the continuous wavelet transform algorithm, the IC can be obtained by detecting the downward peak of the smoothed vertical acceleration, and the FC can be obtained by detecting the upward peak of the velocity (integrated vertical acceleration). Thus, the beginning and end of the gait cycle can be identified. The overall process of gait data preprocessing is as follows. Figure 3 As shown in Figure 2, the gait cycle and its sub-phases are as follows: Figure 4 shown.

[0084] The embodiment of the present application uses continuous wavelet transform (CWT) to detect key moments of the gait cycle. Other time-frequency analysis methods, such as short-time Fourier transform (STFT) or Hilbert transform, may also be used to improve the ability to capture gait features.

[0085] like Figure 1 As shown, the gait feature extraction and analysis method described in the embodiment of the present invention also includes:

[0086] Step S400: generating a gait analysis feature matrix according to the acceleration gait feature, the angular velocity gait feature and the signal feature.

[0087] The definitions and symbols involved in feature extraction are shown in Table 1.

[0088] Table 1

[0089] symbol describe IC Initial contact (event in the gait cycle) FC Final contact (event in the gait cycle) t Time index or current time step l Effective leg length h Vertical displacement of the center of gravity T The total number of samples in a given time period <![CDATA[v t ]]> Angular velocity at time t D Gait phase collection (stride, step length, swing, stance) <![CDATA[D duration ]]> Duration of gait phase S Signal Window <![CDATA[S r ]]> RMS of the square root of the signal <![CDATA[s t ]]> Signal value at time t <![CDATA[μ S ]]> The mean of the signal S <![CDATA[σ S ]]> The standard deviation of the signal S

[0090] In an embodiment of the present application, the acceleration gait characteristics include: stride duration, step length duration, swing duration, support duration, initial double support, termination double support, double support, single-leg support, step length, stride length, asymmetry of duration, asymmetry of length, walking speed and at least one of the steps.

[0091] Specifically, once the gait cycle and its subphases are identified, the duration of each gait phase can be calculated. The duration is defined as the time difference between the start and end of the phase. This calculation relies on detecting the initial contact (IC) and final contact (FC) events, including:

[0092] Stride duration: the time from one IC to the next IC of the same foot: IC(t+2)-IC(t);

[0093] Step duration: the time from one IC to the next IC of the contralateral foot: IC(t+1)-IC(t);

[0094] Swing duration: the time from FC to the next IC: IC(t+2)-FC(t);

[0095] Stance duration: time from IC to FC: FC(t)-IC(t);

[0096] Initial double support: the time when both feet are on the ground at the beginning of support: FC(t-1)-IC(t);

[0097] Terminal double support: The time when both feet are on the ground at the end of support: FC(t)-IC(t+1);

[0098] Double support: the sum of the initial and final double support times;

[0099] Single limb support: The time when only one foot is on the ground: IC(t+1)-FC(t-1).

[0100] For the length of each gait cycle and its sub-phases, include:

[0101] Step length: The distance covered by a single step, calculated as: Where h is the vertical displacement of the center of gravity and l is the effective leg length;

[0102] Stride length: The total distance covered in one gait cycle (two steps): stride length (t) + stride length (t+1).

[0103] Asymmetry measures the difference in characteristics between two consecutive gait cycles. Asymmetry includes:

[0104] Asymmetry in duration: |x duration (t+1)-x duration (t)|;

[0105] Asymmetry in length: |x_length(t+1)-x_length(t)|, where x can be any of the above parameters.

[0106] For each gait cycle and gait cycle sub-phase speed, including:

[0107] Walking speed: calculated by dividing the stride length by the stride duration: stride length / stride duration;

[0108] Steps per minute: calculated as the number of steps taken per minute: 60 / step duration;

[0109] After extracting these spatiotemporal features, the mean and standard deviation (SD) were calculated as representative indicators to describe the overall gait characteristics and variability of the population.

[0110] The angular velocity gait feature includes at least one of an average absolute angular velocity, an angle change, and a rotation amount.

[0111] Specifically, the angular velocity feature captures the motion characteristics of gait and can use the three-axis angular velocity measured by AX6.

[0112] Mean absolute angular velocity refers to the average absolute angular velocity within a period of time, and the calculation formula is Among them, v t is the angular velocity at time t during the duration D of a particular signal phase.

[0113] Angle change refers to the cumulative angle change over time, and the calculation formula is

[0114] Rotation refers to the total rotation in one stage, that is, the absolute value of the angle change.

[0115] The signal characteristics include: at least one of standard deviation, coefficient of variation, clearance factor, absolute change mean, mean crossing rate, skewness, kurtosis, absolute energy, signal entropy, signal root mean square and signal range.

[0116] Specifically, signal features describe the characteristics of the signal generated during the gait cycle by extracting representative indicators from each signal stage. For a signal window S = {s1, s2, ..., sT}, where s t is the signal value at time t, T is the number of target signal windows, and the following features are calculated:

[0117] Standard Deviation: Standard deviation is a measure of the variation or dispersion of a set of values. It quantifies the average distance between each data point and the mean of the data set. For a signal S with T samples, the standard deviation σ S Calculated as:

[0118]

[0119] Among them, μ S is the mean of the signal S.

[0120] Coefficient of Variation: The coefficient of variation (CV) is a standardized measure of dispersion of a probability distribution or frequency distribution. It is defined as the standard deviation σ S With mean μ S Ratio of:

[0121]

[0122] Among them, σ S is the standard deviation, μ S is the mean of the signal S.

[0123] Clearance Factor: The clearance factor is used to describe the importance of impulse or burst interference in the signal. It is defined as the ratio of the absolute signal maximum to the square root of the absolute signal root mean square:

[0124]

[0125] Among them, S r The calculation formula is:

[0126] Mean of Absolute Change: Mean of Absolute Change measures the average magnitude of the change between consecutive samples in a signal. It is calculated as:

[0127]

[0128] Among them, s t is the value of the signal at time t.

[0129] Mean Cross Rate: Mean Cross Rate is the average rate at which a signal crosses its mean. It quantifies how often the sign of a signal changes. Mean Cross Rate is defined as:

[0130]

[0131] Among them, n crossing is the number of times the signal crosses its mean, and T is the number of target signal windows.

[0132] Skewness: Skewness is a measure of the asymmetry of the probability distribution of a real-valued random variable about its mean. For a signal S, the skewness is calculated as:

[0133] where μ S is the mean of the signal S.

[0134] Kurtosis: Kurtosis is a measure of the "tail extent" of the probability distribution of a real-valued random variable, describing the shape of the distribution, especially the height and sharpness of the peak in relation to the standard deviation. For a signal S, kurtosis is calculated by the following formula:

[0135]

[0136] Absolute Energy: Absolute energy is a measure that describes the overall amplitude and variability of a signal. It is calculated as the sum of the squares of the signal values:

[0137]

[0138] Among them, s tis the signal value at time t, and T is the number of target signal windows.

[0139] Signal Entropy: Signal entropy is a measure of signal uncertainty or randomness, indicating the amount of information contained in the signal. For a standardized signal S, the signal entropy H(S) is calculated as:

[0140]

[0141] Among them, p(s t ) is the value of each s in the signal S t The probability of occurrence, T is the number of different values ​​in the signal.

[0142] Signal Root Mean Square: The RMS value provides a measure of the signal amplitude:

[0143]

[0144] Signal Range: The signal range quantifies the span between the maximum and minimum values ​​of the signal, providing an indication of the signal amplitude variation. The calculation formula is:

[0145] Range(S)=max(S)-min(S).

[0146] The embodiment of the present application combines data from multiple sensors to achieve efficient and low-cost gait analysis, which is suitable for large-scale population screening.

[0147] In the embodiment of the present application, the gait analysis feature matrix also includes: a mean and a standard deviation respectively obtained based on each of the acceleration gait feature, the angular velocity gait feature and the signal feature.

[0148] Specifically, a total of 28 acceleration gait features can be extracted from the three-axis accelerometer. Since the gait cycle includes four stages, a total of 72 angular velocity gait features are extracted from the three-axis gyroscope, thus extracting 264 signal features from the three-axis accelerometer and gyroscope.

[0149] The embodiments of the present application can automatically extract angular velocity and multiple signal features, providing more dimensional information for detecting gait abnormalities and neurodegenerative diseases, especially performing well in identifying movement incoordination and abnormal gait in the early stages.

[0150] In the embodiment of the present application, after obtaining the original acceleration signal collected by the three-axis accelerometer in the same period, the method further includes:

[0151] Calculate the signal vector amplitude based on the acceleration original signal, and convert the signal vector amplitude into;

[0152] Using a sliding window method, the signal vector amplitude is divided into a number of segmented windows;

[0153] Convert several segmented windows into a low-dimensional space;

[0154] A trained Gaussian mixture model is obtained, and a plurality of segmented windows in a low-dimensional space are clustered using the Gaussian mixture model to obtain a clustering probability corresponding to the signal vector amplitude.

[0155] Specifically, the embodiment of the present application reflects the individual's exercise intensity through an autonomously trained model. The system clusters the gait data through the extracted gait features, especially key indicators such as signal vector amplitude (SVM), and Gaussian mixture model (GMM). The exercise intensity model can not only identify conventional actions such as walking and standing, but also detect high-intensity exercise behaviors such as running and going up and down stairs. This model has been optimized based on a large amount of experimental data, and can accurately assess the individual's exercise intensity and daily activity level in a free living environment, thereby providing reliable data support for early disease prevention, rehabilitation assessment, and personalized health monitoring.

[0156] We selected data from sixty testers wearing AX6 devices for seven days in the database, randomly extracted the data of each participant, and integrated them together.

[0157] Signal vector magnitude (SVM): Calculate the signal vector magnitude (SVM) to assess the intensity of the motion. The calculation formula for SVM is:

[0158]

[0159] where a x 、a y and a z are the accelerations measured along the x-axis, y-axis, and z-axis at time t, respectively.

[0160] Segmentation of the signal vector amplitude is a key technique in preprocessing, which is used to divide the accelerometer data into continuous segments (windows). A sliding window method is used to divide the large data stream into small windows to capture activity patterns more effectively. For example, a window of 100 samples corresponding to one second of accelerometer data is used for feature extraction because it can fully capture the cycle of daily activities. In order to reduce the information loss at the edge of the window, a 50% overlap between adjacent windows is considered.

[0161] After segmentation, principal component analysis (PCA) is performed. Specifically, principal component analysis is used to reduce the dimension of the data by transforming the original data into a lower dimensional space and minimizing information loss. For example, 15% of the sliding windows are randomly selected and principal component analysis is applied to calculate the cumulative proportion of the principal components. It is found that the first 11 principal components can explain 95% of the data variation.

[0162] After dimensionality reduction, Gaussian Mixture Model (GMM) is used to cluster different SVMs. GMM is a probabilistic model used to represent multiple Gaussian distributions and is usually used for clustering. Assuming that the data comes from a linear combination of multiple Gaussian distributions, its probability density function can be expressed as:

[0163]

[0164] Where K is the number of Gaussian components, π k is the mixing coefficient of the kth component, satisfying is the kth Gaussian distribution with mean μ k , the covariance matrix is ​​Σ k .

[0165] From the training data, estimate the mean μ for each model k and the covariance matrix Σ k , and obtain the following clustering probabilities, such as Figure 5 The fourth category includes jogging, squatting, climbing stairs, etc. The fifth category includes walking, etc. The sixth category includes stillness and sitting, etc. The eighth category includes waist rotation, etc.

[0166] The embodiment of the present application can accurately identify daily activities of different intensities, such as walking, running, standing, etc., through the autonomous training of the motion intensity model, and perform stably in a free living environment.

[0167] For new test objects, follow the same data processing process:

[0168] First, segment the signal vector amplitude into sliding windows to obtain a new matrix

[0169] Second, we apply principal component analysis to the test data. We use the same principal component matrix V as the training data. m , get the test data after dimension reduction: X′ ts =X ts V m ;

[0170] Third, based on the Gaussian mixture model trained on the original data, calculate each test window y i The probability of belonging to each class k.

[0171] The formula for calculating probability is:

[0172]

[0173] In calculating the posterior probability γ for each category k ik Then, the test sample y i Assigned to the class with the highest posterior probability, i.e.: argmax k γ ik ;

[0174] The proportion of each cluster in the test sample:

[0175] The data of a participant is randomly selected for clustering, and the results are as follows Figure 6 and Figure 7 As shown. In this test, the proportion of the fourth category was 2.37%, the proportion of the fifth category was 17.08%, and the proportion of the sixth category was as high as 79.64%. These results show that the participants mainly performed activities such as stillness and walking during the test, especially the proportion of the sixth category showed a high level of stillness. These data provide an important basis for subsequent motion analysis and individual behavior pattern recognition.

[0176] In addition to gait analysis, the platform of the present invention can achieve more comprehensive physical activity monitoring and health assessment by introducing other sensors (such as heart rate sensors or ECG devices). In addition to early diagnosis of neurodegenerative diseases, the present invention can be extended to scenarios such as rehabilitation training assessment, gait analysis in athlete training, and fall risk assessment. Specifically, the following expansions can be made:

[0177] First, multi-dimensional data fusion: In addition to gait data, more physiological data (such as heart rate, blood oxygen saturation, etc.) can be integrated for joint analysis in the future, thereby providing patients with more comprehensive individual health assessment and disease prevention;

[0178] Second, personalized health management: Personalized health management programs can be further developed, using long-term gait data to predict disease progression or health risks, and customizing rehabilitation training plans or exercise recommendations based on individual differences;

[0179] Third, remote health monitoring: By connecting the platform to mobile devices or cloud services, remote health monitoring can be achieved, providing real-time health data feedback, helping doctors or caregivers to intervene in time and improve the quality of life of patients;

[0180] Fourth, upgrade of smart wearable devices: The platform can be integrated into smarter wearable devices by combining artificial intelligence technology to realize automatic data collection, analysis and feedback, and provide real-time health advice.

[0181] The embodiment of the present application expands the traditional gait analysis that only relies on the data of the three-axis accelerometer by introducing angular velocity features and signal features. The fusion of such features provides more comprehensive motion information, making feature extraction richer, so as to more accurately reflect the gait dynamics. At the same time, the Gaussian mixture model is used to perform cluster analysis on the motion intensity to effectively identify and classify different motion states. Compared with traditional clustering technology, the Gaussian mixture model has better flexibility and accuracy.

[0182] In addition, combined with the function of calculating common exercise intensity indicators, this platform can more comprehensively evaluate the user's exercise performance and health status, thereby enhancing the practicality of the analysis. It also enables the platform to provide personalized exercise recommendations and enhance the user's health management experience.

[0183] In one embodiment, if Figure 8 As shown, based on the above gait feature extraction and analysis method, the present invention also provides a gait feature extraction and analysis device, including:

[0184] An acquisition module 100 is used to acquire an original acceleration signal collected by a three-axis accelerometer and an original angular velocity signal collected by a three-axis gyroscope in the same period of time, wherein the three-axis accelerometer and the three-axis gyroscope are used to be placed on a predetermined body part;

[0185] A cutting module 200, used for cutting the acceleration original signal and the angular velocity original signal into a plurality of signal windows respectively, and taking the signal window in the gait segment as the target signal window;

[0186] An extraction module 300, configured to extract acceleration gait features, angular velocity gait features and signal features based on the target signal window;

[0187] The generating module 400 is used to generate a gait analysis feature matrix according to the acceleration gait feature, the angular velocity gait feature and the signal feature.

[0188] It should be noted that the above explanations of the embodiment of the gait feature extraction and analysis method are also applicable to the gait feature extraction and analysis device of this embodiment, and will not be repeated here.

[0189] According to the gait feature extraction and analysis device proposed in the embodiment of the present application, an original acceleration signal collected by a three-axis accelerometer and an original angular velocity signal collected by a three-axis gyroscope in the same time period are obtained, and the three-axis accelerometer and the three-axis gyroscope are used to be placed on a predetermined body part; the original acceleration signal and the original angular velocity signal are respectively cut into a plurality of signal windows, and the signal window in the gait segment is used as a target signal window; based on the target signal window, acceleration gait features, angular velocity gait features and signal features are extracted; a gait analysis feature matrix is ​​generated according to the acceleration gait features, angular velocity gait features and signal features, and multi-dimensional feature data is obtained by analyzing the data collected by low-cost acquisition equipment, thereby realizing efficient and low-cost gait analysis and improving the accuracy of gait analysis.

[0190] Fig. 9 A schematic diagram of the structure of a terminal provided in an embodiment of the present application. The terminal may include:

[0191] A memory 501 , a processor 502 , and a computer program stored in the memory 501 and executable on the processor 502 .

[0192] When the processor 502 executes the program, the gait feature extraction and analysis method provided in the above embodiment is implemented.

[0193] Furthermore, the terminal further includes:

[0194] The communication interface 503 is used for communication between the memory 501 and the processor 502 .

[0195] The memory 501 is used to store computer programs that can be executed on the processor 502 .

[0196] The memory 501 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0197] If the memory 501, the processor 502 and the communication interface 503 are implemented independently, the communication interface 503, the memory 501 and the processor 502 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0198] Optionally, in a specific implementation, if the memory 501, the processor 502 and the communication interface 503 are integrated on a chip, the memory 501, the processor 502 and the communication interface 503 can communicate with each other through an internal interface.

[0199] The processor 502 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0200] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the above gait feature extraction and analysis method is implemented.

[0201] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0202] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0203] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes additional implementations, in which the order shown or discussed may not be followed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0204] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can read instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or N wirings (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically by optically scanning the paper or other medium and then editing, interpreting or otherwise processing in a suitable manner if necessary and then storing it in a computer memory.

[0205] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0206] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0207] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0208] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A gait feature extraction and analysis method, characterized in that: The method comprises: Acquiring an original acceleration signal collected by a three-axis accelerometer and an original angular velocity signal collected by a three-axis gyroscope in the same period, wherein the three-axis accelerometer and the three-axis gyroscope are placed on a predetermined body part; Cutting the acceleration original signal and the angular velocity original signal into a plurality of signal windows respectively, and taking the signal window in the gait segment as the target signal window; Extracting acceleration gait features, angular velocity gait features and signal features based on the target signal window; A gait analysis feature matrix is ​​generated according to the acceleration gait feature, the angular velocity gait feature and the signal feature.

2. The gait feature extraction and analysis method according to claim 1, characterized in that: The acceleration original signal and the angular velocity original signal are respectively cut into a plurality of signal windows, and the signal window in the gait segment is used as the target signal window, including: Cutting the acceleration original signal and the angular velocity original signal into a plurality of signal windows respectively; Extracting root mean square, range, main frequency, spectrum flatness, signal entropy and mean crossing rate for each of the signal windows; Inputting the root mean square, range, main frequency, spectrum flatness, signal entropy and mean crossing rate of each signal window into the trained random forest classification model to obtain output results; If the output result is that the signal window is in the gait segment, the signal window is used as the target signal window.

3. The gait feature extraction and analysis method according to claim 1, characterized in that: Extracting acceleration gait features, angular velocity gait features and signal features based on the target signal window includes: identifying each gait cycle and sub-phase of the gait cycle in each of the target signal windows; Based on each gait cycle and sub-stage of the gait cycle in each of the target signal windows, acceleration gait features, angular velocity gait features and signal features are extracted.

4. The gait feature extraction and analysis method according to claim 3, characterized in that: Identifying each gait cycle and sub-phase of the gait cycle in each of the target signal windows, including: Based on the continuous wavelet transform algorithm, the downward peak value of the smoothed vertical acceleration is detected for each target signal window to determine the initial contact, and the upward peak value of the integrated vertical acceleration is detected for each target signal window to determine the final contact; Determine each gait cycle and sub-phase of the gait cycle in the target signal window according to the initial contact and the final contact; The sub-phases of the gait cycle include: support phase, swing phase, stride phase and half-stride phase.

5. The gait feature extraction and analysis method according to claim 1, characterized in that: The acceleration gait characteristics include at least one of: stride duration, step length duration, swing duration, support duration, initial double support, end double support, double support, single-leg support, step length, stride length, asymmetry of duration, asymmetry of length, walking speed and step frequency; The angular velocity gait feature includes: at least one of average absolute angular velocity, angle change and rotation amount; The signal characteristics include: at least one of standard deviation, coefficient of variation, clearance factor, absolute change mean, mean crossing rate, skewness, kurtosis, absolute energy, signal entropy, signal root mean square and signal range.

6. The gait feature extraction and analysis method according to claim 5, characterized in that: The gait analysis feature matrix also includes: a mean value and a standard deviation respectively obtained based on each feature of the acceleration gait feature, the angular velocity gait feature and the signal feature.

7. The gait feature extraction and analysis method according to claim 1, characterized in that: After obtaining the original acceleration signal collected by the three-axis accelerometer in the same period, it also includes: Calculate the signal vector amplitude based on the acceleration original signal, and convert the signal vector amplitude into; Using a sliding window method, the signal vector amplitude is divided into a number of segmented windows; Convert several segmented windows into a low-dimensional space; A trained Gaussian mixture model is obtained, and a plurality of segmented windows in a low-dimensional space are clustered using the Gaussian mixture model to obtain a clustering probability corresponding to the signal vector amplitude.

8. A gait feature extraction and analysis device, characterized in that: The device comprises: An acquisition module, used to acquire an original acceleration signal collected by a three-axis accelerometer and an original angular velocity signal collected by a three-axis gyroscope in the same period of time, wherein the three-axis accelerometer and the three-axis gyroscope are used to be placed on a predetermined body part; A cutting module, used for cutting the acceleration original signal and the angular velocity original signal into a plurality of signal windows respectively, and taking the signal window in the gait segment as the target signal window; An extraction module, used for extracting acceleration gait features, angular velocity gait features and signal features based on the target signal window; A generating module is used to generate a gait analysis feature matrix according to the acceleration gait feature, the angular velocity gait feature and the signal feature.

9. A terminal, characterized in that: include: A memory, a processor, and a gait feature extraction and analysis program stored in the memory and executable on the processor, wherein the gait feature extraction and analysis program, when executed by the processor, implements the steps of the gait feature extraction and analysis method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program can be executed to implement the steps of the gait feature extraction and analysis method according to any one of claims 1 to 7.

Images