A method and device for optimizing sensor deployment based on motor neuron diseases

Abstract

The present invention provides a method and device for optimizing sensor deployment based on motor nerve diseases, wherein the method for optimizing sensor deployment based on motor nerve diseases comprises: deploying sensors at the concerned part corresponding to the target motor nerve disease, collecting motion posture data of the concerned part; training a support vector machine model according to the collected motion posture data, using the trained support vector machine model to obtain the sensitivity of each sensor corresponding to the motion posture data; obtaining a candidate sensor set according to the sensitivity of each sensor; retraining the trained support vector machine model according to the new motion posture data collected by the candidate sensor set; updating the candidate sensor set using a genetic algorithm, obtaining a candidate sensor set when the output result accuracy of the retrained support vector machine model is optimal and the number of candidate sensor sets is the least, and obtaining an optimized sensor set. The efficiency of motion posture detection can be improved.

Description

Technical Field

[0001] The present invention relates to the field of motion detection technology, and in particular to a method and device for optimizing sensor deployment based on motor neuron diseases. Background Art

[0002] As people's demand for healthy life increases, various chronic motor neurological diseases, such as hemiplegia after stroke, have become important factors affecting the quality of life. Therefore, long-term tracking and detection of patients' movement postures and adjustment of rehabilitation programs based on the test results have become an efficient means of rehabilitation for patients with motor neurological diseases.

[0003] The current motion posture detection technology mainly includes two categories, one is optical motion capture technology, and the other is motion posture detection technology based on inertial measurement unit (IMU). The former records the spatial changes of various parts of the patient and has detailed positioning, but the equipment is relatively complex and requires a fixed site, which is not suitable for long-term measurement. In contrast, the latter is a wearable device that can measure motion posture without the need for additional equipment, and is more suitable for long-term tracking and detection of patients. However, the current IMU motion posture detection system requires the placement of sensors for each suspected part, resulting in a large number of sensors, a large amount of data, and a long processing time, which reduces the efficiency of motion posture detection, and requires a long wearing time and a complicated use process. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a method and device for optimizing sensor deployment based on motor neuron diseases, so as to improve the efficiency of motion posture detection.

[0005] In a first aspect, an embodiment of the present invention provides a method for optimizing sensor deployment based on motor neuron diseases, including:

[0006] Deploy sensors at the concerned parts corresponding to the target motor nerve diseases to collect the motion posture data of the concerned parts;

[0007] A support vector machine model is trained based on the collected motion posture data, and the sensitivity of each sensor corresponding to the motion posture data is obtained using the trained support vector machine model;

[0008] According to the sensitivity of each sensor, a candidate sensor set is obtained;

[0009] Retrain the trained support vector machine model based on the new motion posture data collected by the candidate sensor set;

[0010] The genetic algorithm is used to update the candidate sensor set, and the candidate sensor set with the best output accuracy of the retrained support vector machine model and the least number of candidate sensor sets is obtained, thereby obtaining the optimized sensor set.

[0011] In combination with the first aspect, an embodiment of the present invention provides a first possible implementation manner of the first aspect, wherein the training of a support vector machine model based on the collected motion posture data includes:

[0012] The collected motion posture data is filtered using a high-pass filter to obtain angle data with 0 as the reference;

[0013] For each action, according to the angle data corresponding to each sensor of the action, the peak-to-peak value data of each angle data is obtained;

[0014] The peak-to-peak value data was normalized to obtain normalized data in the range of 0-1, and the normalized data was used to train the support vector machine model.

[0015] In combination with the first possible implementation of the first aspect, the embodiment of the present invention provides a second possible implementation of the first aspect, wherein the using the trained support vector machine model to obtain the sensitivity of each sensor corresponding to the motion posture data includes:

[0016] For each motion posture data in the motion posture data group input to the support vector machine model, the value of the motion posture data in the motion posture data group is changed in sequence to obtain motion posture data groups with different changed values;

[0017] For each motion posture data set with a changed value, input the motion posture data set with the changed value into a trained support vector machine model to obtain a label value of the motion posture data set with the changed value;

[0018] For each change value of the motion posture data, the difference between the maximum value and the minimum value in the label value is calculated, and the ratio of the difference to the maximum value is calculated to obtain the sensitivity of the motion posture data.

[0019] In combination with the second possible implementation of the first aspect, the embodiment of the present invention provides a third possible implementation of the first aspect, wherein acquiring the candidate sensor set according to the sensitivity of each sensor includes:

[0020] The obtained sensitivities of each motion posture data are sorted from large to small, and candidate motion posture data are selected from the sorted values;

[0021] The sensors to which the candidate motion posture data belongs are obtained to obtain a candidate sensor set.

[0022] In combination with the first aspect and any possible implementation manner of the first aspect to the third possible implementation manner, the embodiment of the present invention provides a fourth possible implementation manner of the first aspect, which further includes:

[0023] Wearing each optimized sensor in the optimized sensor set on a corresponding part of the monitored object to collect posture data of the moving object;

[0024] Perform high-pass filtering and normalization on the posture data of the moving object;

[0025] The normalized motion object posture data are grouped to obtain grouped data, and the grouped data are input into a retrained support vector machine to obtain label values ​​corresponding to the normalized motion object posture data.

[0026] In combination with the fourth possible implementation manner of the first aspect, the embodiment of the present invention provides a fifth possible implementation manner of the first aspect, wherein, after collecting the motion object posture data and before performing high-pass filtering on the motion object posture data, the method further includes:

[0027] If optimized sensors are arranged on the left and right sides of the same part, the dynamic time warping algorithm is used to align the posture data of the moving object collected by the optimized sensors on both sides.

[0028] In combination with the fourth possible implementation of the first aspect, the embodiment of the present invention provides a sixth possible implementation of the first aspect, wherein grouping the normalized motion object posture data to obtain grouped data includes:

[0029] The optimized sensors are divided into three groups: upper limbs, lower limbs, and torso. The torso includes the upper chest and hips, the upper limbs include the upper arms and forearms, and the lower limbs include the thighs, calves, and insteps.

[0030] For each group of sensors, the normalized motion object posture data corresponding to the sensors in the group are grouped together.

[0031] In a second aspect, an embodiment of the present invention further provides a device for optimizing sensor deployment based on motor nerve diseases, including:

[0032] A data acquisition module, used to place sensors at the concerned parts corresponding to the target motor nerve diseases, and collect the motion posture data of the concerned parts;

[0033] A sensitivity acquisition module is used to train a support vector machine model based on the collected motion posture data, and use the trained support vector machine model to obtain the sensitivity of each sensor corresponding to the motion posture data;

[0034] The initial screening module is used to obtain a set of candidate sensors based on the sensitivity of each sensor;

[0035] A retraining module, used to retrain the trained support vector machine model based on new motion posture data collected by the candidate sensor set;

[0036] The optimization module is used to update the candidate sensor set by using a genetic algorithm, obtain the candidate sensor set when the output result accuracy of the retrained support vector machine model is optimal and the number of candidate sensor sets is the least, and obtain the optimized sensor set.

[0037] In a third aspect, an embodiment of the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0038] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are executed.

[0039] The method and device for optimizing sensor deployment based on motor nerve diseases provided by the embodiment of the present invention deploy sensors at the concerned part corresponding to the target motor nerve disease, collect the motion posture data of the concerned part; train the support vector machine model according to the collected motion posture data, and use the trained support vector machine model to obtain the sensitivity of each sensor corresponding to the motion posture data; obtain the candidate sensor set according to the sensitivity of each sensor; retrain the trained support vector machine model according to the new motion posture data collected by the candidate sensor set; update the candidate sensor set using a genetic algorithm, obtain the candidate sensor set when the output result accuracy of the retrained support vector machine model is optimal and the number of candidate sensor sets is the least, and obtain the optimized sensor set. In this way, the support vector machine model is used to obtain the sensitivity of the sensor, the sensor is preliminarily selected, and then the optimal sensor is determined from the preliminarily selected sensors based on the genetic algorithm and the support vector machine model, which can effectively reduce the number of deployed sensors and reduce the amount of collected motion posture data, thereby improving the efficiency of motion posture detection.

[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0042] Figure 1 A schematic diagram of the method flow for optimizing sensor deployment based on motor nerve diseases provided by an embodiment of the present invention is shown;

[0043] Figure 2 A schematic diagram of the structure of a device for optimizing sensor deployment based on motor nerve diseases provided by an embodiment of the present invention is shown;

[0044] Figure 3 A schematic diagram showing the arrangement of sensors provided in an embodiment of the present invention is shown;

[0045] Figure 4 A schematic diagram of the structure of a sensor provided by an embodiment of the present invention is shown;

[0046] Figure 5 A schematic diagram of sensor deployment based on stroke optimization provided by an embodiment of the present invention is shown;

[0047] Figure 6 A schematic diagram of the structure of a computer device 600 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0049] For each motor neuron disease, the body parts of concern are different. For example, Parkinson's disease focuses on hand shaking and calf movement rhythm, while hemiplegia after stroke focuses on the movement pattern of the affected side. Moreover, not every part of concern needs to be equipped with sensors. Therefore, by optimizing the currently deployed sensors, the number of deployed sensors can be effectively reduced, thereby shortening the time required for motion posture data processing and improving the detection efficiency of motion posture.

[0050] The embodiments of the present invention provide a method and device for optimizing sensor deployment based on motor neuron diseases, which will be described below through embodiments.

[0051] Figure 1 The schematic diagram of the method flow for optimizing sensor deployment based on motor neuron diseases provided by an embodiment of the present invention is shown. Figure 1 As shown, the method includes:

[0052] Step 101, placing sensors at the concerned part corresponding to the target motor nerve disease, and collecting motion posture data of the concerned part;

[0053] In the embodiment of the present invention, taking a target motor nerve disease as an example, the focus areas are located at the back of the upper chest, the back of the hips, the front of the upper arms, the front of the lower arms, the front of the thighs, the front of the lower legs, and the insteps on the uppers. In this way, a sensor is worn at each focus area, for a total of 12 sensors. As an optional embodiment, each sensor is provided with a three-axis gyroscope, a three-axis accelerometer, and a three-axis magnetometer.

[0054] In the embodiment of the present invention, by wearing sensors, multiple motion posture data of the wearer's various actions can be recorded, for example, walking motion posture data, squatting motion posture data, both hands side-raising motion posture data, elbow flexion motion posture data, etc. As an optional embodiment, the motion posture data is three-axis motion posture data, and each sensor collects three-axis motion posture data of the corresponding position.

[0055] In the embodiment of the present invention, the collected motion posture data includes both the three-axis motion posture data of patients with target motor nerve diseases at various parts of interest and the three-axis motion posture data of healthy people at various parts of interest. Part of the three-axis motion posture data of healthy people at various parts of interest is used as test data, and the remaining motion posture data collected is used as training data.

[0056] Step 102, training a support vector machine model based on the collected motion posture data, and using the trained support vector machine model to obtain the sensitivity of each sensor corresponding to the motion posture data;

[0057] In an embodiment of the present invention, as an optional embodiment, training a support vector machine model according to the collected motion posture data includes:

[0058] A11, filtering the collected motion posture data using a high-pass filter to obtain angle data with 0 as the reference;

[0059] In the embodiment of the present invention, the motion posture data is filtered to remove the DC bias. As an optional embodiment, taking the left and right legs as an example, the angle data includes: rotation data along the skeletal axis, lifting angle data and plane swing data.

[0060] In an embodiment of the present invention, the acceleration data collected by the three-axis gyroscope and the acceleration data collected by the three-axis accelerometer are subjected to Kalman complementary filtering to obtain the pitch angle and the roll angle, which are analyzed with the geomagnetic field data collected by the three-axis magnetometer to obtain the three-axis attitude angle.

[0061] A12, for each action, according to the angle data corresponding to each sensor of the action, obtain the peak-to-peak value data of each angle data;

[0062] In the embodiment of the present invention, the peak-to-peak value is the average variation of the angle data. For each action, 12 sensors are arranged, and the motion posture data collected by each sensor is three-axis motion posture data, and each axis motion posture data corresponds to a peak-to-peak value data. Therefore, for this action, a total of 12×3=36 peak-to-peak value data are obtained.

[0063] In the embodiment of the present invention, the motion posture data collected by each sensor are coaxial in time, that is, for each action, each sensor collects motion posture data at the same time, and collects data for a period of time at the same time. Then, for each axis motion posture data, the maximum and minimum value points of the axis motion posture data are extracted, and finally the average value of the peak-to-peak value is taken.

[0064] In the embodiment of the present invention, the peak-to-peak value corresponding to the motion posture data can characterize the range of motion of the corresponding joint of the corresponding part, and can reflect the motion pattern of the tester. At the same time, by extracting the peak-to-peak value data features corresponding to the motion posture data of each axis, every three peak-to-peak value data features characterize one IMU, the number of inputs of the SVM model can be reduced as much as possible, so as to more clearly and conveniently analyze the sensitivity of the features.

[0065] A13, normalizing the peak-to-peak value data to obtain normalized data in the range of 0-1, and using the normalized data to train a support vector machine model.

[0066] In the embodiment of the present invention, the input of the support vector machine (SVM) model is the normalized data corresponding to the 36 peak-to-peak data of each action, and the output is the label value of whether there is a disease. For example, the label of a healthy person is 0, and the label of a patient is 1.

[0067] In an embodiment of the present invention, for each motion posture data collected, the motion posture data corresponding to the action is used as the input of the support vector machine model, and an output label value between zero and one is obtained, and the trained support vector machine model is verified based on the test data.

[0068] In an embodiment of the present invention, as an optional embodiment, the sensitivity of each sensor corresponding to the motion posture data is obtained by using a trained support vector machine model, including:

[0069] A21, for each motion posture data in the motion posture data group input into the support vector machine model, sequentially changing the value of the motion posture data in the motion posture data group to obtain motion posture data groups with different changed values;

[0070] A22, for each motion posture data set with a changed value, inputting the motion posture data set with a changed value into a trained support vector machine model to obtain a label value of the motion posture data set with a changed value;

[0071] In the embodiment of the present invention, the normalized data corresponding to 36 peak-to-peak data is a motion posture data set, each peak-to-peak data corresponds to a sub-feature, and for each sub-feature in the motion posture data set, the value of the sub-feature is changed to obtain a motion posture data set with a changed value. For example, for the sub-feature with a changed value, the value of the sub-feature can be set to 0, 0.25, 0.5, 0.75, 1 respectively, and the other sub-features remain unchanged. In this way, five motion posture data sets corresponding to the sub-feature can be obtained, and the support vector machine model outputs five results (label values) respectively. Wherein, the output result is a value between 0 and 1.

[0072] A23, for each change value of the motion posture data, calculate the difference between the maximum value and the minimum value in the label value, and calculate the ratio of the difference to the maximum value to obtain the sensitivity of the motion posture data.

[0073] In the embodiment of the present invention, a sensor includes three types of motion posture data, and thus the sensitivity of the sensor includes the sensitivities corresponding to the three types of motion posture data respectively.

[0074] Step 103, obtaining a candidate sensor set according to the sensitivity of each sensor;

[0075] In an embodiment of the present invention, as an optional embodiment, obtaining a candidate sensor set according to the sensitivity of each sensor includes:

[0076] A31, sorting the obtained sensitivities of the motion posture data from large to small, and selecting candidate motion posture data from the sorted sensitivities;

[0077] In the embodiment of the present invention, 36 motion posture data correspond to 36 sub-features, and the sensitivities of the 36 sub-features are ranked. As an optional embodiment, sub-features with sensitivities in the top 20% are selected as candidate sub-features.

[0078] A32, obtaining the sensor to which the candidate motion posture data belongs, and obtaining a candidate sensor set.

[0079] In the embodiment of the present invention, sensors corresponding to sub-features with top 20% sensitivity are found, assuming that there are m sensors, and the found m candidate sensors are recorded as an initial sensor group (candidate sensor set) S1.

[0080] In the embodiment of the present invention, a plurality of candidate motion posture data may belong to the same sensor.

[0081] Step 104, retraining the trained support vector machine model based on the new motion posture data collected by the candidate sensor set;

[0082] In the embodiment of the present invention, based on the features (3m) corresponding to the initial sensor group S1, the motion posture data (new motion posture data) collected but not used for support vector machine training is used as input to retrain the SVM model.

[0083] Step 105 , using a genetic algorithm to update the candidate sensor set, obtain the candidate sensor set when the output result accuracy of the retrained support vector machine model is optimal and the number of candidate sensor sets is minimal, and obtain the optimized sensor set.

[0084] In the embodiment of the present invention, as an optional embodiment, a genetic algorithm is used to update the candidate sensor set, and a candidate sensor set is obtained when the output result accuracy of the retrained support vector machine model is optimal and the number of candidate sensor sets is the least, and an optimized sensor set is obtained, including:

[0085] A41, randomly selecting a predetermined number of candidate sensors from the candidate sensor set to obtain a candidate sensor group;

[0086] In the embodiment of the present invention, k candidate sensors s1, s2, ..., sk are randomly selected from the initial sensor group S1 to form a candidate sensor population P(0) = {s1, s2, ..., sk} in the genetic algorithm. Let the number of iterations t = 0, and the maximum number of iterations T = 500.

[0087] In the embodiment of the present invention, as an optional embodiment, the number of candidate sensors in the initial sensor group S1 is greater than 5, and the number of candidate sensors in the candidate sensor group should be no less than 3.

[0088] A42, for each candidate sensor in the candidate sensor group, obtaining a fitness value of the candidate sensor according to the support vector machine classification model, and obtaining the fitness of the candidate sensor based on the fitness value of the candidate sensor and the fitness value of each candidate sensor in the candidate sensor group;

[0089] In the embodiment of the present invention, an iterative operation is performed to calculate the fitness. For each candidate sensor in the candidate sensor group P(t), the SVM classification model is trained respectively, and the corresponding fitness value f(s1), f(s2), ..., f(sk) is calculated. Then the fitness of the candidate sensor si (i = 1, 2, ..., k) is f(si) / (f(s1) + f(s2) + ... + f(sm)).

[0090] A43, retaining the candidate sensors with the fitness threshold number in the candidate sensor group, deleting other candidate sensors, and completing the candidate sensors in the candidate sensor group based on the preset crossover and mutation algorithm, and performing the next round of iteration with the completed candidate sensor group;

[0091] In the embodiment of the present invention, as an optional embodiment, the candidate sensors with fitness in the top 30% are retained, a total of n, and the remaining (kn) are discarded. Then, through crossover and mutation operations, (kn) candidate sensors are supplemented from the candidate sensor set to obtain the iterated candidate sensor group P(t+1).

[0092] In an embodiment of the present invention, as an optional embodiment, based on a preset crossover and mutation algorithm, completing the candidate sensors in the candidate sensor group includes:

[0093] Take fitness as the probability of each candidate sensor being selected. According to the probability, randomly select two candidate sensors each time, and take 50% of the sensors from each to form a new sensor group scross, which is the result after crossover. Then randomly select 2 candidate sensors in scross and replace them with any two candidate sensors outside scross to obtain the mutated result. Perform this operation (kn) times in total to generate (kn) new candidate sensors, which form a candidate sensor group with the original n candidate sensors.

[0094] A44, if the number of iterations reaches a preset iteration number threshold, or the maximum fitness of a candidate sensor in the candidate sensor group is different from the maximum fitness of a candidate sensor in the candidate sensor group of the previous iteration by no more than a preset error threshold, the candidate sensor group of this iteration is obtained as the optimized sensor set.

[0095] In the embodiment of the present invention, if t=T, or the maximum fitness no longer increases, the candidate sensor group with the current maximum fitness is used as the optimized sensor set S2 output and the calculation is terminated; otherwise, t=t+1 is set and step A42 is executed.

[0096] In the embodiment of the present invention, the problem of optimizing the sensor can be expressed as:

[0097]

[0098] stnum(s)≥3,S∈D

[0099] Among them, s is the solution to the problem, that is, the optimized sensor set.

[0100] D is the feasible domain, that is, the candidate sensor set.

[0101] acc SVM is the accuracy of the SVM classification model.

[0102] num is the number of candidate sensors in the optimized sensor set or candidate sensor set.

[0103] In the embodiment of the present invention, the optimization objective f(s) is to maximize the classification accuracy while minimizing the ratio of the number of sensors in the optimization sensor set to the total number of sensors in the candidate sensor set.

[0104] In the embodiment of the present invention, the initial sensor group is combined and optimized, and the optimized initial sensor group is S1. S1 is updated by using a genetic algorithm, and the goal is to make the classification result accuracy of the SVM model the highest and the number of sensors in S1 the smallest. In this way, the optimized sensor set is obtained based on the sensors that make the classification result accuracy of the SVM model the best and the number of candidate sensors the smallest.

[0105] In the embodiment of the present invention, after optimization, an optimal sensor set for the target motor nerve disease is finally obtained.

[0106] In an embodiment of the present invention, as an optional embodiment, the method further includes:

[0107] A51, wearing each optimized sensor in the optimized sensor set on a corresponding part of the monitored object to collect posture data of the moving object;

[0108] A52, high-pass filtering and normalization of the motion object posture data;

[0109] In the embodiment of the present invention, the posture data of the moving object on each axis collected by the optimization sensor is subjected to high-pass filtering to remove the bias.

[0110] In the embodiment of the present invention, as an optional embodiment, after collecting the motion object posture data and before performing high-pass filtering on the motion object posture data, the method further includes:

[0111] If optimized sensors are arranged on both sides of the same part, the dynamic time warping (DTW) algorithm is used to align the posture data of the moving object collected by the optimized sensors on both sides to reduce the complexity of data processing.

[0112] A53, grouping the normalized motion object posture data to obtain grouped data, inputting the grouped data into the retrained support vector machine to obtain label values ​​corresponding to the normalized motion object posture data.

[0113] In an embodiment of the present invention, as an optional embodiment, the normalized motion object posture data is grouped to obtain grouped data, including:

[0114] The optimized sensors are divided into three groups: upper limbs, lower limbs, and torso. The torso includes the upper chest and hips, the upper limbs include the upper arms and forearms, and the lower limbs include the thighs, calves, and insteps.

[0115] For each group of sensors, the normalized motion object posture data corresponding to the sensors in the group are grouped together.

[0116] In the embodiment of the present invention, the data contained in the group is used as the input data unit of the support vector machine.

[0117] In the embodiment of the present invention, when performing motion posture monitoring, the time series data collected by the optimized sensor is used. For example, for m optimized sensors, the sampling time is t, the sampling rate is f, and a total of 3*m data sequences are collected, each of which has a length of t*f. After the collected data sequences are grouped, they are directly input into the support vector machine model, and the information contained in the data sequences can be comprehensively analyzed. For example, the peak-to-peak value information, stability information, symmetry information, and other information contained in the data sequences.

[0118] In an embodiment of the present invention, by grouping data sequences, data with strong correlation are grouped together, so that the data within the group are strongly correlated and the data between groups are weakly correlated. The support vector machine model structure can also be simplified, the amount of calculation can be reduced, and the accuracy of the output results can be improved.

[0119] In the embodiment of the present invention, as an optional embodiment, the grouped data is input into a retrained support vector machine to obtain a label value corresponding to the normalized motion object posture data, including:

[0120] For each group of data, features are extracted through a one-dimensional CNN network, and the features corresponding to each group of data extracted are merged in a fully connected network to obtain the label value corresponding to the normalized motion object posture data.

[0121] In an embodiment of the present invention, in each group of data, the three-axis time series data of all sensors are combined into a group of inputs, and features are extracted through a one-dimensional CNN network respectively. The three groups of extracted features are merged in a fully connected network to finally obtain the disease diagnosis result.

[0122] In the embodiment of the present invention, as an optional embodiment, for each group of data, features are extracted through a one-dimensional CNN network, and the features corresponding to each extracted group of data are merged in a fully connected network to obtain a label value corresponding to the normalized motion object posture data, including:

[0123] The three-axis time series data of all sensors in the upper limb group are combined to obtain an upper limb posture data input group, the upper limb posture data input group is input into the SVM model, after the operation of the first convolution layer, the first convolution feature of the upper limb is obtained, the first convolution feature of the upper limb is input into the second convolution layer, the second convolution feature of the upper limb is obtained, and the second convolution feature of the upper limb is input into the pooling layer to obtain the upper limb pooling feature;

[0124] The three-axis time series data of all sensors in the lower limb group are combined to obtain a lower limb posture data input group, the lower limb posture data input group is input into the SVM model, after the operation of the first convolution layer, the first convolution feature of the lower limb is obtained, the first convolution feature of the lower limb is input into the second convolution layer, the second convolution feature of the lower limb is obtained, and the second convolution feature of the lower limb is input into the pooling layer to obtain the lower limb pooling feature;

[0125] The three-axis time series data of all sensors in the torso group are combined to obtain a torso posture data input group, and the torso posture data input group is input into the SVM model. After the operation of the first convolution layer, the first convolution feature of the torso is obtained, and the first convolution feature of the torso is input into the second convolution layer to obtain the second convolution feature of the torso. The second convolution feature of the torso is input into the pooling layer to obtain the pooled feature of the torso;

[0126] The upper limb pooling features, lower limb pooling features, and torso pooling features are input into the fully connected layer to obtain the label value corresponding to the normalized motion object posture data.

[0127] Figure 2 FIG. 1 is a schematic diagram showing the structure of a device for optimizing sensor deployment based on motor nerve diseases provided by an embodiment of the present invention. Figure 2 As shown, the device comprises:

[0128] The data acquisition module 201 is used to place sensors at the concerned part corresponding to the target motor nerve disease and collect the motion posture data of the concerned part;

[0129] Figure 3 FIG. 1 shows a schematic diagram of the sensor arrangement provided by an embodiment of the present invention. Figure 3 As shown, in the embodiment of the present invention, the focus areas are respectively located at the back of the upper chest, the back of the hips, the front of both upper arms, the front of both forearms, the front of both thighs, the front of both calves and the insteps of both feet on the shoe surface.

[0130] In the embodiment of the present invention, the motion posture data is three-axis motion posture data, including but not limited to: walking motion posture data, squatting motion posture data, both hands side-raising motion posture data, elbow flexion motion posture data, etc.

[0131] The sensitivity acquisition module 202 is used to train a support vector machine model based on the collected motion posture data, and use the trained support vector machine model to acquire the sensitivity of each sensor corresponding to the motion posture data;

[0132] In the embodiment of the present invention, as an optional embodiment, the sensitivity acquisition module 202 includes:

[0133] A filtering unit (not shown in the figure) is used to filter the collected motion posture data using a high-pass filter to obtain angle data with 0 as the reference;

[0134] A peak-to-peak value calculation unit, for obtaining peak-to-peak value data of each angle data according to the angle data corresponding to each sensor of the action for each action;

[0135] A training unit is used to normalize the peak-to-peak value data to obtain normalized data in the range of 0-1, and use the normalized data to train a support vector machine model;

[0136] A data adjustment unit is used for sequentially changing the value of each motion posture data in the motion posture data group input into the support vector machine model to obtain motion posture data groups with different changed values;

[0137] A label value output unit is used for inputting the motion posture data group with the changed value into the trained support vector machine model for each motion posture data group with the changed value, so as to obtain the label value of the motion posture data group with the changed value;

[0138] The sensitivity calculation unit is used to calculate the difference between the maximum value and the minimum value in the label value for each motion posture data of the changed value, and calculate the ratio of the difference to the maximum value to obtain the sensitivity of the motion posture data.

[0139] The initial screening module 203 is used to obtain a set of candidate sensors according to the sensitivity of each sensor;

[0140] In the embodiment of the present invention, as an optional embodiment, the initial screening module 203 includes:

[0141] A sorting unit (not shown in the figure), used to sort the obtained sensitivities of each motion posture data from large to small, and select candidate motion posture data from the sorting;

[0142] The initial screening unit is used to obtain the sensor to which the candidate motion posture data belongs, and obtain a candidate sensor set.

[0143] A retraining module 204, for retraining the trained support vector machine model based on new motion posture data collected by the candidate sensor set;

[0144] The optimization module 205 is used to update the candidate sensor set using a genetic algorithm to obtain a candidate sensor set with the best output result accuracy of the retrained support vector machine model and the least number of candidate sensor sets, thereby obtaining an optimized sensor set.

[0145] In an embodiment of the present invention, as an optional embodiment, the device further includes:

[0146] The posture monitoring module (not shown in the figure) is used to wear each optimized sensor in the optimized sensor set on the corresponding part of the monitored object to collect the posture data of the moving object; perform high-pass filtering and normalization on the posture data of the moving object; group the normalized posture data of the moving object to obtain grouped data, input the grouped data into the retrained support vector machine, and obtain the label value corresponding to the normalized posture data of the moving object.

[0147] In the embodiment of the present invention, as another optional embodiment, the device further includes:

[0148] The data alignment module uses a dynamic time warping algorithm to align the posture data of the moving object collected by the optimization sensors on both sides if the same part is equipped with optimization sensors on the left and right sides.

[0149] In an embodiment of the present invention, as an optional embodiment, the normalized motion object posture data is grouped to obtain grouped data, including:

[0150] The optimized sensors are divided into three groups: upper limbs, lower limbs, and torso. The torso includes the upper chest and hips, the upper limbs include the upper arms and forearms, and the lower limbs include the thighs, calves, and insteps.

[0151] For each group of sensors, the normalized motion object posture data corresponding to the sensors in the group are grouped together.

[0152] Figure 4 FIG. 1 shows a schematic diagram of the structure of a sensor provided by an embodiment of the present invention. Figure 4 As shown, the sensor includes: an IMU module 41, a clip 42, and an elastic strap 43, wherein:

[0153] The elastic band 43 includes: an elastic band (not shown in the figure) woven from polyester yarn and rubber band, and plastic buckles at both ends of the elastic band;

[0154] In the embodiment of the present invention, the length of the elastic strap can be adjusted by a plastic buckle. The elastic strap has a certain elasticity and can firmly place the IMU module 41 on the body to prevent it from sliding.

[0155] The clip 42 is a metal clip, used in combination with the plastic clip of the elastic band 43 to fix the IMU module 41 on the elastic band of the elastic band 43;

[0156] The IMU module 41 is a sensor circuit encapsulated in a plastic shell, and the plastic shell is adhered to the clip 42.

[0157] In the embodiment of the present invention, as an optional embodiment, the IMU module 41 includes: a data storage module, a Bluetooth module, a central processing unit and an inertial sensor module (not shown in the figure).

[0158] The inertial sensor module is used to collect motion posture data, including: a three-axis gyroscope, a three-axis accelerometer, and a three-axis magnetometer.

[0159] In the embodiment of the present invention, after the inertial sensor module collects the motion posture data, the three-axis posture angle is obtained through classic Kalman complementary filtering and posture solution.

[0160] In the embodiment of the present invention, the central processing unit is an STM32 chip, which is used to generate control instructions to control the working process of each module, including a data analysis module for processing data.

[0161] The data analysis module is used to preprocess data, calculate parameters, and obtain label values ​​through a convolutional neural network (CNN).

[0162] The Bluetooth module is used for data transmission between different sensors and wirelessly connects with external terminals to transmit data for analysis.

[0163] The data storage module is used to store data and label values.

[0164] In an embodiment of the present invention, the acceleration data collected by the three-axis gyroscope and the acceleration data collected by the three-axis accelerometer are subjected to Kalman complementary filtering to obtain the pitch angle and the roll angle, which are analyzed with the geomagnetic field data collected by the three-axis magnetometer to obtain the three-axis attitude angle.

[0165] In the embodiment of the present invention, a stroke is used as an example for specific description. The motor symptoms of a stroke patient include: abnormal range of dorsiflexion of the foot when walking, hand hurdle movement, sudden straightening of the knee during the stance phase, and leg stiffness and circular movement. Figure 2 After optimizing the deployed sensors, the sensor deployment diagram is as follows: Figure 5 As shown, Figure 5 The schematic diagram of sensor layout based on stroke optimization provided by an embodiment of the present invention is shown. A total of seven optimal sensor layout locations shown in Figures 1-7 are obtained, which are located at the upper chest, the affected arm, the bilateral thighs, the bilateral calves, and the affected instep. By arranging sensors in these locations, the collected motion posture data has a strong distinction between healthy people and patients, and can track the motion posture well. The small number of sensors arranged can effectively reduce the amount of collected motion posture data, thereby improving the efficiency of motion posture detection. Furthermore, the sensor is wearable, small in size, and portable, does not affect the wearer's daily life, and can be used for a long time. Moreover, the optimization of the number of sensors makes wearing more convenient and simple, with less discomfort, reduced complexity, reduced cost, and increased battery life. In addition, based on the label value output by the neural network, the effect of rehabilitation can be determined by comparing the label values ​​before and after, and a reference can be provided for the adjustment of the rehabilitation plan.

[0166] like Figure 6 As shown, an embodiment of the present application provides a computer device 600 for executing Figure 1 The method for optimizing sensor deployment based on motor nerve diseases is described in the present invention. The device includes a memory 601, a processor 602 connected to the memory 601 via a bus, and a computer program stored in the memory 601 and executable on the processor 602. When the processor 602 executes the computer program, the steps of the method for optimizing sensor deployment based on motor nerve diseases are implemented.

[0167] Specifically, the above-mentioned memory 601 and processor 602 can be general-purpose memories and processors, which are not specifically limited here. When the processor 602 runs the computer program stored in the memory 601, it can execute the above-mentioned method for optimizing sensor deployment based on motor neuron diseases.

[0168] Corresponds to Figure 1 The method for optimizing sensor deployment based on motor nerve diseases, an embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method for optimizing sensor deployment based on motor nerve diseases are executed.

[0169] Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, etc. When the computer program on the storage medium is run, the above-mentioned method for optimizing sensor deployment based on motor neuron diseases can be executed.

[0170] In the embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. The system embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of systems or units, which can be electrical, mechanical or other forms.

[0171] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0172] In addition, each functional unit in the embodiments provided in the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0173] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0174] It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and are not to be understood as indicating or implying relative importance.

[0175] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. They should all be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for optimizing sensor deployment based on motor neuron diseases, characterized in that: include: Deploy sensors at the concerned parts corresponding to the target motor nerve diseases to collect the motion posture data of the concerned parts; A support vector machine model is trained based on the collected motion posture data, and the sensitivity of each sensor corresponding to the motion posture data is obtained using the trained support vector machine model; According to the sensitivity of each sensor, a candidate sensor set is obtained; Retrain the trained support vector machine model based on the new motion posture data collected by the candidate sensor set; The genetic algorithm is used to update the candidate sensor set, and the candidate sensor set with the best output accuracy of the retrained support vector machine model and the least number of candidate sensor sets is obtained, thereby obtaining the optimized sensor set.

2. The method according to claim 1, characterized in that The method of training a support vector machine model based on the collected motion posture data includes: The collected motion posture data is filtered using a high-pass filter to obtain angle data with 0 as the reference; For each action, according to the angle data corresponding to each sensor of the action, the peak-to-peak value data of each angle data is obtained; The peak-to-peak value data was normalized to obtain normalized data in the range of 0-1, and the normalized data was used to train the support vector machine model.

3. The method according to claim 2, characterized in that The method of using the trained support vector machine model to obtain the sensitivity of each sensor corresponding to the motion posture data includes: For each motion posture data in the motion posture data group input to the support vector machine model, the value of the motion posture data in the motion posture data group is changed in sequence to obtain motion posture data groups with different changed values; For each motion posture data set with a changed value, input the motion posture data set with the changed value into a trained support vector machine model to obtain a label value of the motion posture data set with the changed value; For each change value of the motion posture data, the difference between the maximum value and the minimum value in the label value is calculated, and the ratio of the difference to the maximum value is calculated to obtain the sensitivity of the motion posture data.

4. The method according to claim 3, characterized in that: The step of obtaining a candidate sensor set according to the sensitivity of each sensor includes: The obtained sensitivities of each motion posture data are sorted from large to small, and candidate motion posture data are selected from the sorted values; The sensors to which the candidate motion posture data belongs are obtained to obtain a candidate sensor set.

5. The method according to any one of claims 1 to 4, characterized in that: Also includes: Wearing each optimized sensor in the optimized sensor set on a corresponding part of the monitored object to collect posture data of the moving object; Perform high-pass filtering and normalization on the posture data of the moving object; The normalized motion object posture data are grouped to obtain grouped data, and the grouped data are input into a retrained support vector machine to obtain label values ​​corresponding to the normalized motion object posture data.

6. The method according to claim 5, characterized in that After collecting the motion object posture data and before performing high-pass filtering on the motion object posture data, the method further includes: If optimized sensors are arranged on the left and right sides of the same part, the dynamic time warping algorithm is used to align the posture data of the moving object collected by the optimized sensors on both sides.

7. The method according to claim 5, characterized in that The step of grouping the normalized motion object posture data to obtain grouped data includes: The optimized sensors are divided into three groups: upper limbs, lower limbs, and torso. The torso includes the upper chest and hips, the upper limbs include the upper arms and forearms, and the lower limbs include the thighs, calves, and insteps. For each group of sensors, the normalized motion object posture data corresponding to the sensors in the group are grouped together.

8. A device for optimizing sensor deployment based on motor neuron diseases, characterized in that: include: A data acquisition module, used to place sensors at the concerned parts corresponding to the target motor nerve diseases, and collect the motion posture data of the concerned parts; A sensitivity acquisition module is used to train a support vector machine model based on the collected motion posture data, and use the trained support vector machine model to obtain the sensitivity of each sensor corresponding to the motion posture data; The initial screening module is used to obtain a set of candidate sensors based on the sensitivity of each sensor; A retraining module, used to retrain the trained support vector machine model based on new motion posture data collected by the candidate sensor set; The optimization module is used to update the candidate sensor set by using a genetic algorithm, obtain the candidate sensor set when the output result accuracy of the retrained support vector machine model is optimal and the number of candidate sensor sets is the least, and obtain the optimized sensor set.

9. A computer device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the computer device is running, the processor and the memory communicate via the bus. When the machine-readable instructions are executed by the processor, the steps of the method for optimizing sensor deployment based on motor neuron diseases as described in any one of claims 1 to 7 are performed.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, executes the steps of the method for optimizing sensor deployment based on motor neuron diseases as described in any one of claims 1 to 7.

Images