A smart portable gait detection device and testing method

Abstract

This application belongs to the field of biomedical engineering technology, specifically relating to an intelligent portable gait detection device and testing method, comprising a plantar acquisition module, a dorsum acquisition module, an ankle acquisition module, and a data processing module. The plantar acquisition module is used to analyze pressure simulation data during walking, while the dorsum and ankle acquisition modules are used to collect acceleration, angular acceleration, and ankle joint angles during walking. This application utilizes plantar pressure sensors and dorsum and ankle inertial sensors, combined with algorithms, to form a simple, reliable, and low-cost gait monitoring system. This system can comprehensively analyze a patient's gait parameters, and by combining the obtained gait parameters with medical knowledge, accurately determine gait abnormalities, predict the degree of abnormality, and perform long-term gait tracking, allowing the device to be directly integrated into the monitoring of patients' daily lives.

Description

Technical Field

[0001] This invention belongs to the field of biomedical engineering technology, and more specifically, relates to an intelligent portable gait detection device and testing method. Background Technology

[0002] In an era where people's living standards are improving, more and more people are paying attention to health issues. Many health problems can cause motor dysfunction. If related symptoms can be detected in time, it will help control the deterioration of the disease and play a role in preventing problems before they arise.

[0003] For example, among the elderly, the age group most prone to health problems in society, there are many common ailments, with musculoskeletal dysfunction caused by diabetes, flat feet, and Parkinson's disease being particularly prominent. Musculoskeletal dysfunction often leads to gait abnormalities; if these conditions are detected early and treated effectively, they can be well managed. Groups that experience high levels of physical exertion, such as athletes, may develop chronic injuries due to intense training. Timely detection and recovery can help prevent serious injuries and extend their careers.

[0004] To help address people's health issues, devices with detection capabilities have become fundamental. Devices on the market for detecting human movement and posture include inertial capture systems and optical capture systems. These systems are highly specialized, expensive, and complex to operate, and are currently mostly purchased by laboratories for research purposes, making it difficult to apply them to individual patient self-monitoring. Besides the systems mentioned above, many technologies also achieve gait detection through wearable sensors. Multimodal sensors can perform gait recognition while avoiding the insufficient accuracy problems caused by single sensors, and also measure a richer variety of data. However, the functionality of these systems is largely limited to data collection, lacking effective and comprehensive data analysis and processing, with insufficient human-computer interaction, and their complex structure also makes them difficult to implement in home settings. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by proposing an intelligent portable gait detection device and testing method. Utilizing plantar pressure sensors and inertial sensors for the dorsum of the foot and ankle, combined with artificial intelligence algorithms, a simple, reliable, and low-cost gait monitoring system is established. This system enables comprehensive analysis of a patient's gait parameters. By combining these parameters with medical knowledge, it accurately identifies gait abnormalities, predicts the degree of abnormality, and allows for long-term gait tracking, enabling the device to be directly integrated into the monitoring of patients' daily lives.

[0006] To achieve the above objectives, the present invention provides an intelligent portable gait detection device, comprising a plantar acquisition module, a dorsum acquisition module, an ankle acquisition module, and a data processing module; characterized in that the plantar acquisition module includes multiple plantar pressure sensors and an insole, wherein the multiple plantar pressure sensors are embedded in the bottom of the insole and connected to the ankle acquisition module via signal lines;

[0007] The dorsalis foot acquisition module includes a first inertial sensor, a first printed circuit board, and a first strap. The first printed circuit board is equipped with the first inertial sensor and is connected to the foot and ankle acquisition module via a signal line.

[0008] The foot and ankle acquisition module includes a second inertial sensor, a second printed circuit board, a second strap, an integrated circuit chip, and a Bluetooth module. The second printed circuit board is equipped with the second inertial sensor, the integrated circuit chip, and the Bluetooth module. The integrated circuit chip is communicatively connected to the second inertial sensor and the Bluetooth module, respectively.

[0009] The plantar pressure sensor is used to collect simulated plantar pressure data during walking and to analyze the shift of the center of gravity and the symmetry of the left and right feet during walking; the first inertial sensor is used to collect the acceleration and angle of the instep during walking; the second inertial sensor is used to collect the acceleration and angle of the ankle during walking; the integrated circuit chip is used to read the data collected by multiple plantar pressure sensors and the second inertial sensor; and the Bluetooth module is used to input the data processed by the integrated circuit chip into the data processing module.

[0010] The data processing module includes a computer, which actively searches for and establishes a connection with the Bluetooth module of the foot and ankle acquisition module, receives data from the plantar acquisition module, dorsum acquisition module, and foot and ankle acquisition module via the Bluetooth module of the foot and ankle acquisition module, and processes and analyzes the data.

[0011] Furthermore, multiple plantar pressure sensors are distributed longitudinally along the sole of the foot at the front, middle, and rear positions of the insole, with each sensor having the same distance from the outer contour of the insole.

[0012] The present invention also provides a testing method for an intelligent portable gait detection device. Utilizing the aforementioned intelligent portable gait detection device, the testing method includes:

[0013] Step 1: Place the sole acquisition module inside the shoe, and fix the instep acquisition module and ankle acquisition module to the instep and ankle respectively. After the subject has put on the shoes, they enter a walking state and data collection begins.

[0014] Step 2: The sensors of the plantar acquisition module, dorsum acquisition module, and ankle acquisition module collect walking data during the experimental period. The data processing module actively searches for the Bluetooth module of the ankle acquisition module and inputs the walking data into the data processing module through the Bluetooth module.

[0015] Step 3: The data processing module summarizes the data input from the plantar acquisition module, dorsum acquisition module, and ankle acquisition module and establishes an original dataset. It then filters each data point in the original dataset using a filtering function to remove interfering data and obtain a new dataset. Finally, it standardizes the new dataset.

[0016] Step 4: Establish the outlier function. Input each point in the new dataset after standardization in Step 3 into the outlier function analysis. Compare the average local reachability of all points within the k-th distance neighborhood of the currently analyzed data point with the local reachability density of that data point. If the ratio is greater than a threshold, then remove the data point. The outlier function is determined by the following formula:

[0017]

[0018] In the formula: p represents any data point currently being analyzed in the new dataset; o represents the data points surrounding point p; N k (p) represents the k-th distance neighborhood of data point p, that is, the set of all points within the k-th distance of point p, including points at the k-th distance; lrd k w(o) represents the k-th local reachability density of data point o; w(o) represents the weight, and the closer point o is to point p, the greater its weight.

[0019] Step 5: Based on Step 4, the data processing module predicts and evaluates the processed data, establishes the plantar pressure center trajectory line, and compares it with the pressure center trajectory line of normal people through the DTWS algorithm to perform similarity analysis of the plantar pressure center trajectory line. The obtained similarity value is between 0 and 1. The larger the value, the closer the plantar pressure distribution of the test subject is to the normal gait.

[0020] Step 6: Extract the angle and acceleration data of the test subject's walking process from the standardized dataset in Step 3, and establish a data graph with time as the horizontal axis and acceleration and angle as the vertical axis. The graph shows the changes in acceleration and joint angles during the gait cycle, and compares them with the data range of normal people to preliminarily determine whether there are any abnormalities in the test subject's gait.

[0021] Step 7: Establish a logistic regression probability model. Extract the maximum, minimum, and average values ​​of the ankle joint's inversion / valgus and dorsiflexion angles from the data in Step 6. Establish a matrix N with the extracted values ​​and list the corresponding disease types as vectors M. Substitute N and M into the logistic regression probability model to output the disease type and predicted probability. The higher the predicted probability of the corresponding symptom, the higher the severity of the corresponding symptom.

[0022] Furthermore, step 5 specifically includes:

[0023] Step 5.1: Establish a planar coordinate system in the sole plane and determine the coordinates of each pressure sensor. The location of the plantar pressure center is obtained by weighted averaging the pressure values ​​at corresponding coordinates; coordinates and In this context, 'i' represents the pressure sensor serial number; 'i' represents the i-th pressure sensor; 'x' represents the 'i'-th pressure sensor. i This represents the lateral coordinate of the i-th pressure sensor; y i F represents the longitudinal coordinate of the i-th pressure sensor; i This indicates the pressure magnitude of the i-th pressure sensor;

[0024] Step 5.2: Based on the planar coordinate system established in Step 5.1, during the process of a single foot contacting the ground, from the heel striking the ground to the toe leaving the ground, the center of pressure on the sole of the foot is continuously solved and connected in sequence to obtain the trajectory line of the center of pressure on the sole of the foot. The distribution pattern of pressure is determined by the trajectory line of the center of pressure.

[0025] Step 5.3: Compare the pressure center trajectory lines of normal individuals using the DTWS algorithm to perform curve similarity analysis. The DTWS (Dynamic Time Warping Similarity) algorithm includes the DTW algorithm and the Similarity algorithm.

[0026] Furthermore, the DTW algorithm in step 5.3 is as follows:

[0027] The coordinates of the pressure center trajectory are stored in a sequence Q. The test subject's sequence Q is compared and analyzed with the pre-calibrated normal gait sequence C. The length of the Q sequence is adjusted to be the same as that of the C sequence using interpolation. The length of the C sequence is set to n, then C = c1, c2, c3, ..., c i c n Q = q1, q2, q3, ..., q j , ..., q n ; in the sequence c i The coordinates of the element storage point are q j The coordinates of the element storage point are

[0028] Furthermore, the elements in sequence Q and sequence C are used to construct an n*n matrix. The minimum cumulative total distance ED(Q,C) from point (1,1) to point (n,n) in the matrix is:

[0029] ED(Q,C)=γ(n,n) (9)

[0030] The smaller the minimum cumulative total distance ED(Q,C), the higher the similarity of the pressure center trajectory lines of the corresponding sequences Q and C, and the more normal the plantar pressure distribution.

[0031] Furthermore, the Similarity algorithm in step 5.3 is as follows:

[0032] Store the total distance formula (9) and the corresponding similarity in the Q sequence and C sequence. The similarity calculation formula is:

[0033] Similarity=a·exp[b*ED(Q,C)] (10)

[0034] In the formula: a and b represent coefficients; ED(Q,C) represents the total distance calculated by DTW;

[0035] The similarity value is between [0, 1]. The larger the value, the more the plantar pressure distribution of the test subject is close to the normal gait.

[0036] Furthermore, the original dataset in step 3 includes plantar pressure values, dorsiflexion and angle values, and ankle acceleration and angle values.

[0037] Further, the ratio of the average local reachability of all points in the k-th neighborhood of each data point p to the local reachability of the outlier is calculated using formula (3) in step 4. If the ratio is greater than 1, it means that the density of data point p is less than the density of its surrounding points, and the data point p is recorded as an outlier and removed. If the ratio is less than or equal to 1, it means that the density of data point p is greater than the density of its surrounding points, and the data point p currently participating in the calculation is retained.

[0038] Furthermore, the methods for determining abnormal gait in step 6 include:

[0039] Fifth, analyze the angle change curve to determine whether the inversion and eversion angles of the ankle joint are normal, that is, whether the toes are pointing excessively inward or outward, and detect problems such as inversion or eversion of the foot. If the ankle joint should have no inversion and eversion angle values ​​of 0, then it is determined that the foot has no inversion or eversion.

[0040] Sixth, by observing the range of changes in the ankle dorsiflexion angle, it can be determined whether there are functional impairments such as insufficient strength in the ankle joint during walking; if the dorsiflexion angle is 0-20 degrees and the toe flexion angle is 0-30 degrees, and the extreme values ​​deviate too much from this normal range, it can be determined that there is a problem with insufficient strength in the ankle joint.

[0041] Seventh, by analyzing acceleration, the magnitude of force on the foot can be reflected to a certain extent. Observing whether there are large fluctuations in the acceleration curve can serve as a supplement to the measurement results of the pressure sensor. If the acceleration curve changes significantly, it indicates that there is a problem with the smoothness of force application in the foot.

[0042] Eighth, compare the data curves of both feet to determine whether the gait symmetry is normal; if there is a large difference in the shape of the curves, it is determined that there is a symptom of gait asymmetry in both feet.

[0043] Furthermore, the logistic regression probability model function in step 7 is determined by the following formula:

[0044]

[0045] The logistic regression probability model has a value range of (0, 1), which is suitable for representing prediction probabilities.

[0046] During the data fitting process, a multiple linear regression equation y is constructed using the logistic regression probability model f(x):

[0047] y = β1x1 + β2x2 + ... + β n x n (11)

[0048] Let x = y for f(x) in formula (9), and substitute formula (12) into formula (11) to obtain an independent variable with x1, x2, ..., x n The multivariate function f(x1, x2, ..., x) n ):

[0049]

[0050] Each parameter is taken as an independent variable x1, x2, ..., xn. n Input formula (13) and output the predicted probability value of the corresponding symptom.

[0051] The beneficial effects of this invention are:

[0052] First, this invention uses a plantar pressure sensor combined with a dorsum of the foot inertial sensor and an ankle inertial sensor for data collection. Six plantar pressure sensors are distributed at different locations on the front, back, and middle of the foot. These six sensors collect simulated pressure data during walking, reflecting the pressure distribution on the sole. The collected data is uploaded to a computer via Bluetooth serial communication. After weighted averaging of the pressure values ​​along the coordinates, a plantar pressure centerline change curve is obtained. Using the plantar pressure centerline, further analysis of center of gravity shift, lateral symmetry, etc., during walking can be performed. Combined with medical diagnosis, this analysis allows for preliminary gait classification of patients. To achieve more comprehensive and accurate detection, inertial sensors located on the dorsum of the foot and ankle are added. These sensors collect acceleration, angular acceleration, and ankle angle during walking. This data is combined with the data collected by the pressure sensors, plotted, and visualized. Machine learning models are then used for processing and analysis, avoiding the limitations of single sensors and achieving better detection results.

[0053] Secondly, in the preferred implementation, this invention embeds six pressure sensors into a thin-film insole, with corresponding wiring led out from the heel. Two inertial sensors are fixed in corresponding positions with cable ties. The chip and the inertial sensors at the ankle are located on the same cable tie, and the chip is soldered onto a printed circuit board. The wires of each sensor are also soldered onto the printed circuit board and connected to the chip. The chip is then connected to the HC-06 Bluetooth module, transmitting the data collected by the sensors to the computer via Bluetooth communication. In the control program embedded in the chip, since both the inertial sensors and Bluetooth need to be connected to the chip's TX and RX pins, two additional SoftwareSerial objects are created to simulate the TX and RX pins. Then, the simulated values ​​from the input pins of each sensor are read and transmitted to the computer via Bluetooth communication. In the computer, the pybluez library functions of the Python language are used for Bluetooth control. The read data is added to different lists for cleaning and sorting, and then classified and written into text files for storage, completing the data acquisition.

[0054] Third, in the preferred implementation, the testing method of the present invention establishes a plantar pressure center trajectory line in step 5, and compares it with the pressure center trajectory line of normal people using the DTWS algorithm to perform curve similarity analysis. This can accurately analyze the difference between the plantar pressure distribution of the test subject and the normal gait. When the sample size for certain diseases is large enough, the plantar pressure center trajectory lines of these diseases can be identified. Then, a similarity analysis is performed between the new sample and the plantar pressure center trajectory lines of various diseases, thereby classifying the new sample into the disease with the highest similarity, achieving the function of distinguishing diseases.

[0055] Fourth, in the preferred implementation, the testing method of the present invention compares the angle and acceleration data of the test subject's walking process with normal gait data in step 6 from multiple aspects. The comparative analysis is comprehensive. The established data graph can realize data visualization, which helps to observe the division of each stage in the gait and the smoothness between each cycle.

[0056] Fifth, in the preferred implementation, in step 7 of the testing method of the present invention, by establishing a logistic regression probability model, the maximum, minimum and average values ​​of the ankle joint inversion / valgus angle and dorsiflexion angle are extracted from the data in step 6. The extracted values ​​are used to establish a matrix N, and the corresponding disease types are listed as vectors M. N and M are substituted into the logistic regression probability model to output the disease type and predicted probability. The higher the predicted probability of the corresponding symptom, the higher the severity of the corresponding symptom. The test subject is further analyzed in combination with the symptoms, and the analysis structure is more accurate and the accuracy is higher.

[0057] Sixth, in a preferred implementation, the test method of the present invention, in step 3, is more accurate in data analysis by removing interference data because the inertial sensor may be affected by drift and jitter during measurement. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the wearable intelligent portable gait detection device according to Embodiment 1 of the present invention;

[0059] Figure 2 This is a schematic diagram showing the distribution of the foot acquisition module of the intelligent portable gait detection device according to Embodiment 1 of the present invention;

[0060] Figure 3 This is an overall flowchart of the testing method for the intelligent portable gait detection device according to Embodiment 2 of the present invention;

[0061] Figure 4 This is a flowchart illustrating the method for establishing the plantar pressure center trajectory line in the test method of the intelligent portable gait detection device according to Embodiment 2 of the present invention;

[0062] Figure 5 This is a schematic diagram of Embodiment 2 of the present invention, which uses interpolation to adjust the length of the Q sequence to the same length as the C sequence;

[0063] Figure 6 This refers to the joint angle change process during the normal gait cycle in step 6 of embodiment 2 of the present invention.

[0064] Figure 7 This refers to step 6 of embodiment 2 of the present invention, which describes the process of joint angle changes during the state cycle of a Parkinson's patient.

[0065] Among them, 1-foot sole acquisition module; 2-foot dorsum acquisition module; 3-foot and ankle acquisition module. Detailed Implementation

[0066] To enable those skilled in the art to better understand the technical solutions of this application, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0067] The directional terms such as above, below, left, right, front, and back used in this application are based on the positional relationships shown in the attached drawings. Different attached drawings may result in different positional relationships, therefore they should not be interpreted as limitations on the scope of protection.

[0068] In this application, the terms "installation," "connection," "interlocking," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, an integral connection, a mechanical connection, an electrical connection, or a connection that allows communication between components. They can also refer to a direct connection or an indirect connection through an intermediate medium. They can refer to the internal connection of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in this application can be understood according to the specific circumstances.

[0069] Example 1:

[0070] Refer to the instruction manual appendix Figure 1-2 A smart portable gait detection device includes a sole acquisition module 1, a dorsum acquisition module 2, an ankle acquisition module 3, and a data processing module.

[0071] The plantar pressure acquisition module 1 includes multiple plantar pressure sensors and an insole. The plantar pressure sensors are embedded in the bottom of the insole and connected to the foot and ankle acquisition module 3 via signal lines. They are used to collect simulated plantar pressure data during walking, reflecting the pressure distribution on the sole of the foot and analyzing the shift of the center of gravity and the symmetry of the left and right feet during walking. The multiple plantar pressure sensors are distributed longitudinally along the sole of the foot at the front, middle, and rear positions of the insole, with each sensor having the same distance from the outer edge of the insole.

[0072] by Figure 2 Taking the state shown as an example, in this embodiment, the foot acquisition module 1 has 6 foot pressure sensors, namely a1-a6, a1, a2, a3, a4, a5 and a6 are distributed at intervals. a1 is located at the base of the big toe phalanx, a4 is located at the base of the little toe metatarsal bone, a3 and a6 are located at the heel bone, a2 is located between a1 and a3, and a5 is located between a4 and a6.

[0073] The instep acquisition module 2 includes a first inertial sensor, a first printed circuit board, and a first strap. The first printed circuit board is equipped with the first inertial sensor and is connected to the ankle acquisition module 3 via a signal line to acquire acceleration during walking. The first printed circuit board is built into the first strap. When worn, the first printed circuit board is placed on the instep and secured tightly around the instep and sole by the first strap.

[0074] The foot and ankle acquisition module 3 includes a second inertial sensor, a second printed circuit board, a second strap, an integrated circuit chip, and a Bluetooth module. The second printed circuit board is equipped with the second inertial sensor, the integrated circuit chip, and the Bluetooth module. The integrated circuit chip is communicatively connected to the second inertial sensor and the Bluetooth module, respectively.

[0075] Specifically, the plantar pressure sensor and the first printed circuit board are connected to the integrated circuit chip of the foot and ankle acquisition module 3 via signal lines. The integrated circuit chip is used to read data collected by multiple plantar pressure sensors and the second inertial sensor, and the Bluetooth module is used to input the data processed by the integrated circuit chip into the data processing module.

[0076] The dorsum of the foot acquisition module 2 and the ankle acquisition module 3 are used to acquire acceleration, angular acceleration and ankle joint angle during walking.

[0077] Specifically, the Bluetooth module of the foot and ankle acquisition module 3 is model HC-06. Both the first and second inertial sensors are six-axis attitude sensors.

[0078] The data processing module includes a computer, which actively searches for and establishes a connection with the Bluetooth module of the foot and ankle acquisition module 3, receives data from the plantar acquisition module 1, the dorsum acquisition module 2, and the foot and ankle acquisition module 3 via the Bluetooth module of the foot and ankle acquisition module 3, and processes and analyzes the data.

[0079] The structure described in this embodiment is easy to wear and has a simple design. It can monitor, process, and analyze the gait of the test subject in real time, enabling self-testing in daily life and objective assessment in medical diagnosis. It provides great convenience to people in need and has a large application potential.

[0080] Example 2:

[0081] Refer to the instruction manual appendix Figure 3 This embodiment uses the intelligent portable gait detection device from Embodiment 1 for testing. The testing method includes:

[0082] Step 1: Place the sole acquisition module 1 inside the shoe, and fix the instep acquisition module 2 and ankle acquisition module 3 to the instep and ankle respectively. After the subject puts on the shoes, they enter a walking state and begin data collection.

[0083] Step 2: The sensors of the plantar acquisition module 1, the dorsum acquisition module 2, and the ankle acquisition module 3 collect walking data during the experimental period. The data processing module actively searches for the Bluetooth module of the ankle acquisition module 3 and inputs the walking data into the data processing module through the Bluetooth module.

[0084] The experimental time in this embodiment is the time it takes for the test subject to walk 5m in a straight line.

[0085] Step 3: The data processing module summarizes the data input from the plantar acquisition module 1, dorsum acquisition module 2, and ankle acquisition module 3 and establishes an original dataset. It then filters each data point in the original dataset using a filtering function to remove interfering data and obtain a new dataset. Finally, it standardizes the new dataset.

[0086] Specifically, the original dataset contains plantar pressure values, dorsiflexion and angle values, and ankle acceleration and angle values.

[0087] It should be noted that due to sensor drift, environmental factors, or electromagnetic interference, interference data may occur during the data acquisition process. To ensure the accuracy of data analysis, it is necessary to filter out the interference data.

[0088] Step 3 specifically includes:

[0089] Step 3.1: The data processing module summarizes the data input from the plantar acquisition module 1, dorsum acquisition module 2, and ankle acquisition module 3 and establishes the original dataset D, denoted as D0. X in the original dataset D n X represents the data from the plantar pressure sensor in the nth dataset; n (n) The n in parentheses represents the nth plantar pressure sensor; A n (1),V n (1) represents the acceleration and angle data during walking collected by the first inertial sensor of the foot dorsalis acquisition module in the nth dataset; A n (2),V n (2) represents the acceleration and angle data collected by the second inertial sensor of the ankle acquisition module in the nth dataset.

[0090] It should be noted that n data points represent the total number of data points obtained according to the sampling frequency within the experimental time.

[0091] Taking plantar pressure sensors as an example, the dataset collected n times by n plantar pressure sensors is as follows:

[0092]

[0093] This embodiment has 6 plantar pressure sensors. Therefore, the dataset collected n times by the 6 plantar pressure sensors is as follows:

[0094]

[0095] Step 3.2: Using Python, read the data collected by the plantar pressure sensor, the first inertial sensor, and the second inertial sensor from the original dataset. Establish a filtering function and apply a weighted average filtering algorithm to each data point in the original dataset to obtain a new dataset D', where D' is denoted as... Establish a standardization function to standardize the filtered dataset D'.

[0096] Specifically, the filter function is determined by the following formula:

[0097]

[0098] In the formula: y(n) represents the filtered sensor data; x(n) represents the nth sampled data of the sensor; x(n-1) and x(n+1) are the data adjacent to x(n) to the left and right respectively; the natural numbers multiplied by x(n-1), x(n) and x(n+1) represent the weights, the closer to the current value, the higher the weight; 2 is the weight of x(n); 1 is the weight of x(n-1) and x(n+1).

[0099] The filtered data is then standardized using Z-Score standardization, determined by the following formula:

[0100]

[0101] In the formula: x old Indicates the value before filtering; x new σ represents the filtered value; μ represents the average value of the data before filtering; σ represents the standard deviation of the data before filtering.

[0102] Step 4: Establish the outlier function. Input each point in the new dataset after standardization in Step 3 into the outlier function analysis. Compare the average local reachability of all points within the k-th distance neighborhood of the currently analyzed data point with the local reachability density of that data point. If the ratio is greater than a threshold, then remove the data point. The outlier function is determined by the following formula:

[0103]

[0104] In the formula: p represents any data point currently being analyzed in the new dataset; o represents the data points surrounding point p; N k (p) represents the k-th distance neighborhood of data point p, that is, the set of all points within the k-th distance of point p, including points at the k-th distance; lrd k (o) represents the k-th local reachability density of data point o; w(o) represents the weight, and the closer point o is to point p, the greater its weight.

[0105] It should be noted that the k-th distance represents the distance radiating outwards from the current data point p as the center, until it covers the k-th neighboring point.

[0106] Step 4 specifically includes:

[0107] Step 4.1: First, calculate the k-th reachable distance from the currently analyzed data point p to data point o in the new dataset, using reach_dist. k (o,p) represents:

[0108] reach_dist k (o,p)=max{d k (o),d(o,p)} (4)

[0109] In the formula: d k (o) represents the k-th distance from data point o; d(o,p) represents the distance from data point o to data point p.

[0110] Step 4.2: via reach_dist in step 4.1 k (o,p), calculate the k-th local reachability density lrd of data point p. k (p), lrd k (p) is determined by the following formula:

[0111]

[0112] In the formula: reach_dist k (o,p) represents the k-th reachable distance from the currently analyzed data point p to the data point o; N k (p) represents the k-th distance neighborhood of point p.

[0113] The k-th locally reachable density lrd of data point p k (p) characterizes the density of point p. The higher the density of point p with its surrounding points, the smaller the reachable distance of each point is likely to be its respective k-th distance, lrd. k The larger the value of (p), the lower the density of point p with its surrounding points, and the more likely the reachable distance of each point is to be the actual distance between the two points. k The smaller the (p) value.

[0114] Step 4.3: Based on Step 4.2, calculate the k-th local outlier of the data point p being analyzed. The outlier function at this point is the primary function, denoted as LOF. k (p), LOF k (p) is determined by the following formula:

[0115]

[0116] In the formula: lrd k (p) represents the k-th local reachability density of data point p; N k (p) represents the k-th distance neighborhood of data point p; lrd k (o) represents the k-th local reachability density of data point o.

[0117] Since the signals collected during gait detection are often periodic, and when the sampling frequency is constant, the greater the rate of change of the data point values, the smaller the density of surrounding points, and vice versa. Therefore, the distribution density of data points in different time periods will also show a certain variation pattern with the period. If a fixed threshold is set, the points to be removed may all fall in the same time period of each period. Therefore, it is necessary to optimize formula (6), which becomes formula (3) after optimization.

[0118] Step 4.4: Calculate the ratio of the average local reachability of all points in the k-th neighborhood of each data point p to the local reachability of the outlier using formula (3). If the ratio is greater than 1, it means that the density of data point p is less than the density of its surrounding points. The data point p is recorded as an outlier and removed. If the ratio is less than or equal to 1, it means that the density of data point p is greater than the density of its surrounding points. The data point p currently participating in the calculation is retained.

[0119] Step 5: Based on Step 4, the data processing module predicts and evaluates the processed data, establishes the plantar pressure center trajectory line, and compares it with the pressure center trajectory line of normal people through the DTWS algorithm to perform similarity analysis of the plantar pressure center trajectory line. The obtained similarity value is between 0 and 1. The larger the value, the closer the plantar pressure distribution of the test subject is to the normal gait.

[0120] Step 5 specifically includes:

[0121] Step 5.1: Establish a planar coordinate system in the sole plane and determine the coordinates of each pressure sensor. The location of the plantar pressure center is obtained by weighted averaging the pressure values ​​at the corresponding coordinates.

[0122] coordinate and In this context, 'i' represents the pressure sensor serial number; 'x' represents the sensor number. i This represents the lateral coordinate of the i-th pressure sensor; y i F represents the longitudinal coordinate of the i-th pressure sensor; i This represents the pressure magnitude of the i-th pressure sensor.

[0123] Step 5.2: Based on the planar coordinate system established in Step 5.1, during the process of a single foot contacting the ground, from the heel striking the ground to the toe leaving the ground, the center of pressure on the sole of the foot is continuously solved and connected in sequence to obtain the trajectory line of the center of pressure on the sole of the foot. The distribution pattern of pressure is determined by the trajectory line of the center of pressure.

[0124] Refer to the instruction manual appendix Figure 4 The process of obtaining the plantar pressure center trajectory line is as follows: First, the data processing module acquires the current plantar pressure sensor data and calculates the current pressure center based on the pressure value and coordinates of each data point. The pressure center is used to determine whether the foot is off the ground. If the foot is off the ground, the pressure values ​​of both left and right sensors are zero. The center coordinates obtained in this step are saved, and the points are plotted and connected to obtain the pressure center line. If the pressure values ​​of some sensors are not zero, it means that the foot is not completely off the ground, and the process proceeds directly to the next moment to acquire the pressure sensor data for the next moment.

[0125] When gait abnormalities are present, the pressure center trajectory obtained from the test will differ from that of a normal person. For example, in patients with foot inversion or foot eversion, the pressure center trajectory will be significantly deviated, and the pressure distribution will tend to be unilateral.

[0126] Step 5.3: Compare the pressure center trajectory lines of normal individuals using the DTWS algorithm to perform curve similarity analysis. The DTWS (Dynamic Time Warping Similarity) algorithm includes the DTW algorithm and the Similarity algorithm.

[0127] The DTW algorithm works as follows: The coordinates of the points on the pressure center trajectory are stored in a sequence Q. The sequence Q of the test subject is compared and analyzed with the pre-calibrated sequence C of normal gait. The length of sequence Q is adjusted to be the same as that of sequence C using interpolation. The length of sequence C is set to n, then C = c1, c2, c3, ..., c i c n Q = q1, q2, q3, ..., q j , ..., q n ; in the sequence c i The coordinates of the element storage point are q j The coordinates of the element storage point are

[0128] Furthermore, the elements in sequence Q and sequence C are used to construct an n*n matrix. The matrix (i, j) is used to store the data point c. i and q j The distance d(i,j) is determined by the following formula:

[0129]

[0130] It should be noted that the purpose of the DTW algorithm is to find a path (warping path) from (1,1) to (n,n) in the matrix with the minimum cost. This is achieved by iteratively searching for the minimum value until the minimum cumulative distance is found. Therefore, the minimum cumulative distance from coordinate point (1,1) to coordinate point (i,j) in the matrix is ​​determined by the following formula:

[0131] γ(i,j)=d(i,j)+min{γ(i,j),γ(i,j),γ(i,j)} (8)

[0132] The minimum cumulative total distance ED(Q,C) from point (1,1) to point (n,n) in the matrix is:

[0133] ED(Q,C)=γ(n,n) (9)

[0134] The smaller the minimum cumulative total distance ED(Q,C), the higher the similarity of the pressure center trajectory lines of the corresponding sequences Q and C, and the more normal the plantar pressure distribution.

[0135] Specifically, with Figure 5 Taking the example shown, we will explain how to adjust the length of the Q sequence to be the same as the length of the C sequence using interpolation:

[0136] Assuming the known coordinates (x0, y0) and (x1, y1) of the trajectory line of the pressure center, we want to obtain the value of x at a certain position on the straight line within the interval [x0, x1]. Based on... Figure 5 As shown, the results are as follows:

[0137]

[0138] Since the value of x is known, the value of y can be obtained using the above formula:

[0139]

[0140] The Similarity algorithm is as follows: Since the total distance ED(Q,C) alone cannot provide a clear and intuitive representation, the optimize.curve_fit function in the scipy library is used to store the total distance formula (9) and the corresponding similarity in the Q and C sequences. The similarity calculation formula Similarity is determined by the following formula:

[0141] Similarity=a·exp[b*ED(Q,C)] (10)

[0142] In the formula: a and b represent coefficients; ED(Q,C) represents the total distance calculated by DTW.

[0143] The similarity value is between [0, 1]. The larger the value, the more the plantar pressure distribution of the test subject is close to the normal gait.

[0144] Furthermore, if the sample size for certain diseases is large enough, the plantar pressure center trajectory lines for these diseases can be identified. Then, a similarity analysis is performed between the new samples and the plantar pressure center trajectory lines of various diseases, thereby classifying the new samples into the disease category with the highest similarity, thus achieving the function of disease differentiation.

[0145] Step 6: Extract the angle and acceleration data of the test subject's walking process from the standardized dataset in Step 3, and establish a data graph with time as the horizontal axis and acceleration and angle as the vertical axis. The graph is used to obtain the change process of acceleration and joint angle in the gait cycle, and is compared with the data range of normal people to preliminarily determine whether there is any abnormality in the test subject's gait.

[0146] Step 6: Create a data graph with time as the horizontal axis and acceleration and angle as the vertical axes. Visualize the data through the data processing module. This helps to observe the division of each stage in gait and the smoothness between each cycle.

[0147] Methods for determining whether a test subject's gait is abnormal include:

[0148] First, analyze the angle change curve to determine whether the inversion and eversion angles of the ankle joint are normal, that is, whether the toes are pointing excessively inward or outward, and detect problems such as inversion or eversion of the foot. If the ankle joint should have no inversion and eversion angle values ​​of 0, then it is determined that the foot has no inversion or eversion.

[0149] Secondly, by observing the range of changes in the ankle dorsiflexion angle, it can be determined whether there are functional impairments such as insufficient joint strength during walking. If the dorsiflexion angle is 0-20 degrees and the toe flexion angle is 0-30 degrees, and the extreme values ​​deviate too much from this normal range, it can be determined that there is a problem of insufficient joint strength.

[0150] Third, by analyzing acceleration, the magnitude of force on the foot can be reflected to some extent. Observing whether there are large fluctuations in the acceleration curve can supplement the measurement results of the pressure sensor. If the acceleration curve changes significantly, it can be judged that there is a problem with the smoothness of force exertion on the foot.

[0151] Fourth, compare the data curves of both feet to determine whether the gait symmetry is normal; if there is a large difference in the shape of the curves, it is determined that there is a symptom of gait asymmetry in both feet.

[0152] Refer to the instruction manual appendix Figure 6-7 , Figure 6 This describes the changes in joint angles during a normal gait cycle. Figure 7This describes the changes in joint angles during the symptom cycle of a Parkinson's patient. Figure 6 and Figure 7 The horizontal axis represents the percentage of a single gait, with 0% indicating the moment the leg lifts during the gait and 100% indicating the moment the leg lifts during the next step. The vertical axis represents the angle of dorsiflexion (positive values) and toe flexion (negative values). By comparing... Figure 6 and Figure 7 It can be seen that when the patient completes 80% of a single gait, the joint angle fluctuates significantly, indicating a lack of strength.

[0153] Step 7: Establish a logistic regression probability model. Extract the maximum, minimum, and average values ​​of the ankle joint's inversion / valgus and dorsiflexion angles from the data in Step 6. Establish a matrix N from the extracted values ​​and list the corresponding disease types as vectors M. Substitute N and M into the logistic regression probability model.

[0154] The above steps constitute the model training process. For each diagnosed patient, the extracted data includes the maximum, minimum, and average values ​​of the ankle joint's inversion / valgus and dorsiflexion angles, as well as the disease type. For multiple patients providing data for training, the above data sets are combined into a matrix N, and the disease types are combined into a vector M.

[0155] After the model training is completed, for new subjects, the maximum, minimum, and average values ​​of the ankle inversion / valgus and dorsiflexion angles are extracted from the data in step 6 and used as parameters x1 to x2. n When input into a logistic regression probability model, it outputs the type of disease and the predicted probability. The higher the predicted probability of a corresponding symptom, the higher the severity of the corresponding symptom.

[0156] Specifically, the logistic regression probability model function is determined by the following formula:

[0157]

[0158] The logistic regression probability model has a value range of (0, 1), which is suitable for representing prediction probabilities.

[0159] During the data fitting process, a multiple linear regression equation y is constructed using the logistic regression probability model f(x):

[0160] y = β1x1 + β2x2 + ... + β n x n (12)

[0161] Let x = y for f(x) in formula (11), and substitute formula (12) into formula (11) to obtain an independent variable with x1, x2, ..., x n The multivariate function f(x1, x2, ..., x) n ):

[0162]

[0163] For a model that has already been fitted, each parameter is used as an independent variable x1, x2, ..., xn. n By inputting formula (13), the predicted probability value of the corresponding symptom will be output.

[0164] For multi-class classification problems, the logistic regression probability model is used to construct multiple multiple linear regression equations. After calculation, the equation with the highest probability value is taken as the classification result.

[0165] The above descriptions are merely embodiments of the present invention, and common knowledge regarding specific structures and characteristics of the solutions is not described in detail here. It will be apparent to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the present invention is defined by the appended claims rather than the foregoing description. Therefore, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An intelligent portable gait detection device, comprising a plantar acquisition module, a dorsum of the foot acquisition module, an ankle acquisition module, and a data processing module; characterized in that, The foot acquisition module includes an insole and multiple foot pressure sensors embedded in the bottom of the insole at the front, middle and rear positions. The multiple foot pressure sensors are connected to the foot and ankle acquisition module via signal lines. The dorsalis foot acquisition module includes a first inertial sensor, a first printed circuit board, and a first strap. The first printed circuit board is equipped with the first inertial sensor and is connected to the foot and ankle acquisition module via a signal line. The foot and ankle acquisition module includes a second inertial sensor, a second printed circuit board, a second strap, an integrated circuit chip, and a Bluetooth module. The second printed circuit board is equipped with the second inertial sensor, the integrated circuit chip, and the Bluetooth module. The integrated circuit chip is communicatively connected to the second inertial sensor and the Bluetooth module, respectively. The plantar pressure sensor is used to collect simulated plantar pressure data during walking and to analyze the shift of the center of gravity and the symmetry of the left and right feet during walking; the first inertial sensor is used to collect the acceleration and angle of the instep during walking; the second inertial sensor is used to collect the acceleration and angle of the ankle during walking; the integrated circuit chip is used to read the data collected by multiple plantar pressure sensors and the second inertial sensor; and the Bluetooth module is used to input the data processed by the integrated circuit chip into the data processing module. The data processing module includes a computer, which actively searches for and establishes a connection with the Bluetooth module of the foot and ankle acquisition module, receives data from the plantar acquisition module, dorsum acquisition module, and foot and ankle acquisition module via the Bluetooth module of the foot and ankle acquisition module, and processes and analyzes the data. The data processing module summarizes the input data, removes interfering data to obtain a new dataset, and standardizes the new dataset. An outlier factor function is established. Each point in the standardized new dataset is input into the outlier factor function analysis. The average local reachability of all points within the k-th distance neighborhood of the currently analyzed data point is compared with the local reachability density of that data point. If the ratio is greater than a threshold, the data point is removed. The outlier factor function is determined by the following formula: ； In the formula: p represents any data point currently being analyzed in the new dataset; o represents the data points surrounding point p; N k (p) represents the k-th distance neighborhood of data point p, that is, the set of all points within the k-th distance of point p, including points at the k-th distance; This represents the k-th local reachability density of data point o; w(o) represents the weight; the closer point o is to point p, the greater its weight. The data processing module predicts and evaluates the processed data, establishes the plantar pressure center trajectory line, and compares it with the pressure center trajectory line of normal people through the DTWS algorithm to perform similarity analysis of the plantar pressure center trajectory line; it extracts the angle and acceleration data of the test subjects' walking process from the standardized dataset, and obtains the process of acceleration and joint angle changes in the gait cycle through illustration to preliminarily determine whether there are any abnormalities in the test subjects' gait. The maximum, minimum, and average values ​​of the ankle joint's inversion / valgus and dorsiflexion angles are extracted. The extracted values ​​are then substituted into a logistic regression probability model to output the disease type and predicted probability.

2. The intelligent portable gait detection device according to claim 1, characterized in that, Each sensor has the same distance from the outer contour of the insole.

3. The intelligent portable gait detection device according to claim 1, characterized in that, The data processing module is configured such that the similarity analysis result of the plantar pressure center trajectory line is a similarity value, the similarity value ranges from 0 to 1, and the larger the similarity value, the more the plantar pressure distribution of the test subject is close to the normal gait; Furthermore, the maximum, minimum, and average values ​​of the extracted ankle inversion / valgus and dorsiflexion angles are used to construct a matrix N, and the corresponding disease types are used to construct a vector M. The matrix N and the vector M are then substituted into a logistic regression probability model so that the output symptom prediction probability is used to characterize the severity of the corresponding symptom. The higher the prediction probability of the corresponding symptom, the higher the severity of the corresponding symptom.

4. The intelligent portable gait detection device according to claim 1, characterized in that, The data processing module is configured to: establish a planar coordinate system in the sole plane and determine the coordinates of each pressure sensor. The location of the plantar pressure center is obtained by weighted averaging the pressure values ​​at corresponding coordinates, where the coordinates are... In this context, 'i' represents the pressure sensor serial number; 'i' represents the i-th pressure sensor. This represents the lateral coordinate of the i-th pressure sensor; Represents the longitudinal coordinate of the i-th pressure sensor; This indicates the pressure magnitude of the i-th pressure sensor; The data processing module is also configured to: continuously solve the plantar pressure center during the process of single foot contact with the ground, from heel impact to toe lift-off, and connect them sequentially to obtain the plantar pressure center trajectory line, and determine the pressure distribution pattern through the pressure center trajectory line; The data processing module is also configured to perform curve similarity analysis by comparing the pressure center trajectory lines of normal people with the DTWS algorithm. The DTWS algorithm includes the DTW algorithm and the Similarity algorithm.

5. The intelligent portable gait detection device according to claim 4, characterized in that, When the data processing module is configured to execute the DTW algorithm, it stores the coordinates of the pressure center trajectory line into a sequence Q, compares and analyzes the test subject's sequence Q with the pre-calibrated normal gait sequence C, and uses interpolation to adjust the length of the Q sequence to be the same as the C sequence. The length of sequence C is set to n, then C = , , , ..., , ..., Q= , , , ..., , ..., ;In sequence The coordinates of the element storage point are ( , ), The coordinates of the element storage point are ( , ); The data processing module is configured to construct an n*n matrix from the elements of sequence Q and sequence C. The minimum cumulative total distance ED(Q,C) from (1,1) to (n,n) in the matrix is ​​calculated, where ED(Q,C)=γ(n,n). The smaller the minimum cumulative total distance ED(Q,C), the higher the similarity of the pressure center trajectory lines of the corresponding sequence Q and sequence C, and the more normal the plantar pressure distribution is.

6. The intelligent portable gait detection device according to claim 5, characterized in that, When the data processing module is configured to execute the Similarity algorithm, it stores the minimum cumulative total distance ED(Q,C) and the corresponding similarity in the Q sequence and the C sequence, respectively. The similarity calculation formula is as follows: satisfy: Where a and b represent coefficients; ED(Q,C) represents the total distance calculated by DTW; similarity The result of the operation is located in the range of [0, 1]. The larger the similarity value, the more the plantar pressure distribution of the test subject is close to the normal gait.

7. The intelligent portable gait detection device according to claim 1, characterized in that, The raw dataset established by the data processing module includes plantar pressure values, dorsiflexion and angle values, and ankle acceleration and angle values.

8. The intelligent portable gait detection device according to claim 1, characterized in that, The data processing module is configured to use an outlier factor function. Calculate the ratio of the average local reachability density of all points in the k-th neighborhood of each data point p to the local reachability density of the data point itself. If the ratio is greater than 1, it means that the density of data point p is less than the density of its surrounding points, and the data point p is marked as an outlier and removed. If the ratio is less than or equal to 1, it means that the density of data point p is greater than the density of its surrounding points, and the data point p currently participating in the calculation is retained.

9. The intelligent portable gait detection device according to claim 1, characterized in that, The data processing module is used to determine whether the test subject's gait is abnormal, and the determination method includes: Analyze the angle change curve to determine whether the inversion and eversion angles of the ankle joint are normal, that is, whether the toes are pointing excessively inward or outward, and detect problems such as inversion or eversion of the foot. If the ankle joint should have no inversion and eversion angle values ​​of 0, then it is determined that the foot has no inversion or eversion. By observing the range of changes in the ankle dorsiflexion angle, it can be determined whether there is a functional impairment such as insufficient strength in the ankle joint during walking. If the dorsiflexion angle is 0~20 degrees and the toe flexion angle is 0~30 degrees, and the extreme values ​​deviate too much from this normal range, it can be determined that there is a problem with insufficient strength in the ankle joint. Acceleration analysis can reflect the magnitude of force on the foot to some extent. Observing whether there are large fluctuations in the acceleration curve can supplement the measurement results of the pressure sensor. If the acceleration curve changes significantly, it can be judged that there is a problem with the smoothness of force application in the foot. By comparing the data curves of both feet, it can be determined whether the gait symmetry is normal; if there is a large difference in the shape of the curves, it can be determined that there is a gait asymmetry in both feet.

10. The intelligent portable gait detection device according to claim 1, characterized in that, The data processing module is used to determine the logistic regression probability model function using the following formula: ； The logistic regression probability model has a range of (0, 1), which is suitable for representing predicted probabilities. During the data fitting process, a logistic regression probability model is used. A multiple linear regression equation was constructed. Let x = y in the formula f(x), then we can obtain an independent variable with x1, x2, ..., xy. n multivariate functions ; The data processing module is used to treat each parameter as independent variables x1, x2, ..., xn. n Input multivariate function Output the predicted probability value of the corresponding symptom.

Images