Intelligent gait training evaluation system for vestibular rehabilitation

Abstract

The invention discloses an intelligent gait training evaluation system for vestibular rehabilitation. A control host collects posture signals of left and right legs and a head of a patient at each operation speed; the left and right leg and head posture signals corresponding to each left leg walking one step are independently used as a unit to be processed, and the left and right leg and head posture signals corresponding to each right leg walking one step are independently used as a unit to be processed; judging whether a head posture signal in any to-be-processed unit under any running speed signal is within an amplitude range corresponding to the running speed, if so, retaining the to-be-processed unit, and otherwise, discarding the to-be-processed unit; splicing the left leg posture signal, the right leg posture signal and the head posture signal in each to-be-processed unit reserved under a certain running speed signal, and performing preprocessing; and performing feature extraction to obtain an N-dimensional feature matrix formed under N operation speed signals, and inputting the N-dimensional feature matrix into the target balance capability evaluation model to train and evaluate the balance dysfunction degree of the patient.

Description

Technical Field

[0001] The invention relates to the technical field of vestibular rehabilitation training equipment, and in particular to an intelligent gait training evaluation system for vestibular rehabilitation. Background Art

[0002] Vestibular diseases often lead to decreased balance and unstable gait, which greatly inconveniences patients' daily lives. Existing rehabilitation training equipment often cannot provide personalized dynamic gait training, especially accurate monitoring of head and body posture and timely feedback on patients' rehabilitation status, and cannot provide rehabilitation adjustment suggestions in a timely manner. Summary of the invention

[0003] In view of the problems and shortcomings of the prior art, the present invention provides an intelligent gait training system for vestibular rehabilitation.

[0004] The present invention solves the above technical problems through the following technical solutions:

[0005] The present invention provides an intelligent gait training and evaluation system for vestibular rehabilitation, which is characterized in that it comprises an intelligent gait training device, VR glasses for vestibular rehabilitation patients to wear, a left leg strap, a right leg strap and a control host, the intelligent gait training device comprises a fixed base, a treadmill for gait training of vestibular rehabilitation patients is fixed at the middle position of the surface of the fixed base, a plurality of protective pillars are fixed along the edge of the fixed base at intervals, a protective guardrail is fixed on the top of the plurality of protective pillars, a plurality of elastic protective long strips are fixed at intervals on the edge of the protective guardrail, an elastic protective belt for surrounding the waist of the vestibular rehabilitation patient is fixed at the end of the plurality of elastic protective long strips, and the elastic protective belt is located directly above the treadmill;

[0006] The left leg strap is provided with a left leg posture sensor, the right leg strap is provided with a right leg posture sensor, the VR glasses are provided with a head posture sensor, and the treadmill is provided with a Hall sensor. The treadmill, the VR glasses, the left leg posture sensor, the right leg posture sensor, the head posture sensor and the Hall sensor are all electrically connected to the control host, and the control host stores the amplitude range of the head posture signal corresponding to each running speed;

[0007] The left leg posture sensor is used to detect the left leg posture signal of the vestibular rehabilitation patient, the right leg posture sensor is used to detect the right leg posture signal of the vestibular rehabilitation patient, the head posture sensor is used to detect the head posture signal of the vestibular rehabilitation patient, and the Hall sensor is used to detect the running speed signal of the treadmill;

[0008] The control host is used to control the treadmill to be at N different running speeds, collect the left leg posture signal, right leg posture signal and head posture signal of the vestibular rehabilitation patient under each running speed signal, and divide the left leg posture signal, right leg posture signal and head posture signal under the running speed signal in an alternating order of one step of the left leg and one step of the right leg. The left leg posture signal, right leg posture signal and head posture signal corresponding to each step of the left leg under the running speed signal are all separately used as a unit to be processed, and the left leg posture signal, right leg posture signal and head posture signal corresponding to each step of the right leg under the running speed signal are also separately used as a unit to be processed. The running speed includes walking speed and running speed, and N is a positive integer;

[0009] The control host is also used to determine whether the head posture signal in any unit to be processed under any running speed signal is within the amplitude range of the head posture signal corresponding to the running speed, and if so, retain the unit to be processed, otherwise, discard the unit to be processed, thereby obtaining the units to be processed retained under N running speed signals;

[0010] The control host is also used to splice the left leg posture signal, the right leg posture signal and the head posture signal in each unit to be processed that is retained under a certain running speed signal, to obtain the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal under the running speed signal, thereby obtaining the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal under N running speed signals;

[0011] The control host is also used to pre-process the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal under N running speed signals to obtain the pre-processed left leg posture signal, the pre-processed right leg posture signal and the pre-processed head posture signal under N running speed signals;

[0012] The control host is also used to extract features from the pre-processed left leg posture signal, the pre-processed right leg posture signal and the pre-processed head posture signal under any running speed signal, and obtain the left leg posture time domain features, the left leg posture frequency domain features, the right leg posture time domain features, the right leg posture frequency domain features, the head posture time domain features and the head posture frequency domain features under the running speed signal to form a one-dimensional feature matrix, based on which an N-dimensional feature matrix formed under N running speed signals can be obtained;

[0013] The control host is also used to input the N-dimensional feature matrix into a target balance ability assessment model constructed based on a random forest regression model for learning and training, to assess and display the degree of balance dysfunction in vestibular rehabilitation patients, and the degree of balance dysfunction includes no disorder, mild disorder, moderate disorder and severe disorder.

[0014] The positive and progressive effects of the present invention are:

[0015] The present invention can improve the balance ability and gait stability of vestibular rehabilitation patients through dynamic gait training, and can also monitor the head and body postures of vestibular rehabilitation patients and provide immediate adjustment suggestions.

[0016] Improve balance ability: Improving the patient's balance ability through dynamic gait training can help patients better adapt to various dynamic environments in daily life.

[0017] Improve gait stability: Enhance the patient's gait stability and reduce the risk of falling.

[0018] Enhanced rehabilitation outcomes: Provides immediate feedback and guidance to ensure patients receive the best possible rehabilitation outcomes. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Schematic diagram of the structure of an intelligent gait training and evaluation system for vestibular rehabilitation according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, the present embodiment provides an intelligent gait training and evaluation system for vestibular rehabilitation, which includes an intelligent gait training device 10, VR glasses 20 for vestibular rehabilitation patients to wear, a left leg strap 30, a right leg strap 40 and a control host 50. The intelligent gait training device 10 includes a circular fixed base 11, a treadmill 12 for gait training of vestibular rehabilitation patients is fixed at the middle position of the surface of the circular fixed base 11, four protective pillars 13 are evenly fixed along the edge of the circular fixed base 11, a circular protective guardrail 14 is fixed on the top of the four protective pillars 13, four elastic protective strips 15 are evenly fixed on the edge of the circular protective guardrail 14, and an elastic protective belt 16 for surrounding the waist of the vestibular rehabilitation patient is fixed at the end of the four elastic protective strips 15, and the elastic protective belt 16 is located directly above the treadmill 12.

[0022] A control host 50 is placed on the outside of a certain protective pillar 13, and an L-shaped cable arm 17 is fixed to the top of the circular protective guardrail 14. The end of the lateral arm of the L-shaped cable arm 17 is located directly above the treadmill 12. A VR cable 21 is passed through the L-shaped cable arm 17. One end of the VR cable 21 is connected to the control host 50, and the other end passes through the L-shaped cable arm 17 and is connected to the VR glasses 20.

[0023] A left leg posture sensor is provided on the left leg strap 30, a right leg posture sensor is provided on the right leg strap 40, a head posture sensor and a speaker are provided on the VR glasses 20, and a Hall sensor is provided on the treadmill 12. The treadmill 12, the VR glasses 20, the left leg posture sensor, the right leg posture sensor, the head posture sensor, the Hall sensor and the speaker are all electrically connected to the control host 50, and the control host 50 stores the amplitude range of the head posture signal corresponding to each running speed.

[0024] In this embodiment, the treadmill 12 is used to simulate walking or running at different speeds; the VR glasses 20 are used to cooperate with the treadmill to display a virtual training scene, so that the trainees can experience the training process more realistically.

[0025] The left leg posture sensor is used to detect the left leg posture signal of the vestibular rehabilitation patient, the right leg posture sensor is used to detect the right leg posture signal of the vestibular rehabilitation patient, the head posture sensor is used to detect the head posture signal of the vestibular rehabilitation patient, and the Hall sensor is used to detect the running speed signal of the treadmill.

[0026] The control host 50 is used to control the treadmill 12 to be at N different operating speeds (including walking speed and running speed), collect the left leg posture signal, right leg posture signal and head posture signal of I historical vestibular rehabilitation patients under each operating speed signal, and divide the left leg posture signal, right leg posture signal and head posture signal of I historical vestibular rehabilitation patients under the operating speed signal in an alternating order of one step of the left leg and one step of the right leg. The left leg posture signal, right leg posture signal and head posture signal corresponding to each left leg step of each historical vestibular rehabilitation patient under the operating speed signal are each separately regarded as a historical unit to be processed, and the left leg posture signal, right leg posture signal and head posture signal corresponding to each right leg step of each historical vestibular rehabilitation patient under the operating speed signal are also each separately regarded as a historical unit to be processed, N is a positive integer, and I is a positive integer.

[0027] The control host 50 is also used to determine whether the head posture signal in any to-be-processed unit of any historical vestibular rehabilitation patient under any running speed signal is within the amplitude range of the head posture signal corresponding to the running speed. If so, the historical to-be-processed unit is retained; otherwise, the historical to-be-processed unit is discarded, thereby obtaining the historical to-be-processed units retained by each historical vestibular rehabilitation patient under N running speed signals.

[0028] The control host 50 is also used to splice the left leg posture signal, the right leg posture signal and the head posture signal in each historical to-be-processed unit retained by a historical vestibular rehabilitation patient under a certain running speed signal, to obtain the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal of each historical vestibular rehabilitation patient under the running speed signal, thereby obtaining the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal of each historical vestibular rehabilitation patient under N running speed signals.

[0029] In this scheme, each historical vestibular rehabilitation patient has a spliced ​​left leg posture signal, a spliced ​​right leg posture signal and a spliced ​​head posture signal under each running speed signal.

[0030] The control host 50 is also used to pre-process the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal of each historical vestibular rehabilitation patient under N running speed signals, and obtain the pre-processed left leg posture signal, the pre-processed right leg posture signal and the pre-processed head posture signal of each historical vestibular rehabilitation patient under N running speed signals.

[0031] In this scheme, each historical vestibular rehabilitation patient has a preprocessed left leg posture signal, a preprocessed right leg posture signal and a preprocessed head posture signal under each running speed signal.

[0032] The control host 50 is also used to extract features from the pre-processed left leg posture signal, the pre-processed right leg posture signal and the pre-processed head posture signal of any historical vestibular rehabilitation patient under any running speed signal, and obtain the left leg posture time domain features, left leg posture frequency domain features, right leg posture time domain features, right leg posture frequency domain features, head posture time domain features and head posture frequency domain features of the historical vestibular rehabilitation patient under the running speed signal to form a one-dimensional feature matrix, based on which an N-dimensional feature matrix constructed under each running speed signal of each historical vestibular rehabilitation patient can be obtained.

[0033] In this scheme, each historical vestibular rehabilitation patient under each running speed signal corresponds to a one-dimensional feature matrix composed of left leg posture time domain features, left leg posture frequency domain features, right leg posture time domain features, right leg posture frequency domain features, head posture time domain features and head posture frequency domain features; each historical vestibular rehabilitation patient needs to experience N running speeds, thereby obtaining an N-dimensional feature matrix constructed under N running speed signals corresponding to each historical vestibular rehabilitation patient.

[0034] The control host 50 is also used to construct a feature data set {(x1, y1), (x2, y2) ... (x I ,y I )},x i represents the N-dimensional feature matrix corresponding to the i-th historical vestibular rehabilitation patient, y i It represents the degree of balance dysfunction corresponding to the i-th historical vestibular rehabilitation patient. The i-th historical vestibular rehabilitation patients include patients with no disorder, patients with mild disorder, patients with moderate disorder and patients with severe disorder.

[0035] The control host 50 is also used to construct a feature training sample set and a feature verification sample set, wherein the feature training sample set includes I1 feature data sets, and the feature verification sample set includes I2 feature data sets, where I=I1+I2, I1 and I2 are both positive integers and I1 is much larger than I2. For example, 80% of the feature data sets in the feature data set are used as feature training sample sets, and 20% of the feature data sets are used as feature verification sample sets.

[0036] The control host 50 is also used to build a balance ability assessment model based on the random forest regression model, input the feature training sample set into the balance ability assessment model for training, input the feature verification sample set into the trained balance ability assessment model for verification, and build a target balance ability assessment model.

[0037] In this embodiment, a balance ability assessment model is constructed based on a random forest regression model, a particle swarm improvement algorithm is used to optimize the hyperparameters of the balance ability assessment model, and the balance ability assessment model is trained using a feature training sample set. The optimal hyperparameters are obtained through continuous iteration of the particle swarm improvement algorithm, and the target balance ability assessment model is constructed using the optimal hyperparameters.

[0038] Specifically, the control host 50 is used to build a balance ability evaluation model based on the random forest regression model, initialize the hyperparameters of the balance ability evaluation model, initialize the speed and position of the particle swarm with a particle number M, set the particle search space dimension W and the maximum number of iterations T, and the particle dimension W of each particle corresponds to the hyperparameter one by one, and each particle represents a set of hyperparameters.

[0039] Each particle is used as a hyperparameter to construct an initial balance ability evaluation model. The feature training sample set is input into each initial balance ability evaluation model for training. The deviation mean square error of the model is used as the initial fitness of each particle, the initial fitness of each particle is used as the individual optimal fitness, the maximum value of the initial fitness of each particle is used as the initialization population optimal fitness, the current position of each particle is used as the initialization individual optimal position of each particle, and the current position of the particle corresponding to the optimal fitness of the initialization population is used as the initialization population optimal position.

[0040] Update the velocity and position of each particle:

[0041]

[0042] Among them, m represents the current mth particle, w represents the current wth dimension search space, and t represents the current tth iteration number. and Respectively represent the speed of the current particle before and after update, and Respectively represent the position of the current particle before and after update, Q t represents the inertia weight, Q max represents the maximum inertia weight, Q min represents the minimum inertia weight, C1 and C2 represent the learning factors, represents the optimal position of the historical individual, Represents the optimal position of the historical population.

[0043] Update the historical individual optimal position and historical population optimal position of the particle swarm: calculate the updated fitness of each particle, compare the updated fitness of each particle with the historical individual optimal fitness of the particle, if the updated fitness of a particle is better than the historical individual optimal fitness, then update the historical individual optimal fitness of the particle = the updated fitness of the particle, and update the historical individual optimal position of the particle = the updated position of the particle, otherwise do not update; then take the maximum value of the historical individual optimal fitness of each particle as the current optimal fitness, if the current optimal fitness is better than the historical population optimal fitness, then update the historical population optimal fitness = the current optimal fitness, and update the historical population optimal position = the current position of the particle corresponding to the current optimal fitness, otherwise do not update.

[0044] When the optimal fitness of the historical population reaches the set fitness or the number of iterations reaches the maximum number of iterations T, the optimal hyperparameters are obtained based on the particles at the optimal position of the historical population, and the target balance ability evaluation model is constructed using the optimal hyperparameters; the particles are arranged from best to worst according to the updated fitness, the sorted M1 particles are eliminated, and M1 particles are randomly generated based on the sorted M2 particles to replace the eliminated particles. The current positions of the randomly generated M1 particles are used as the positions of the corresponding particles after t updates, and the current speed of each particle in the randomly generated M1 particles is the average speed of the particles on both sides of the particle, M=M1+M2, t=t+1, and the operation of updating the speed and position of each particle is performed again.

[0045] Based on the above-mentioned particle swarm improved algorithm, the optimal hyperparameters can be obtained. The target balance ability evaluation model can be constructed using the optimal hyperparameters, and the feature verification sample set can be input into the target balance ability evaluation model for verification.

[0046] The target balance ability assessment model can be used to predict the degree of balance dysfunction in a vestibular rehabilitation patient. Based on the prediction and assessment results, doctors or rehabilitation therapists can promptly give vestibular rehabilitation exercise adjustment suggestions to vestibular rehabilitation patients.

[0047] The control host 50 is used to control the treadmill 12 to be at N different running speeds, collect the left leg posture signal, right leg posture signal and head posture signal of the vestibular rehabilitation patient under each running speed signal, and divide the left leg posture signal, right leg posture signal and head posture signal under the running speed signal in an alternating order of one step of the left leg and one step of the right leg. The left leg posture signal, right leg posture signal and head posture signal corresponding to each step of the left leg under the running speed signal are each treated as a unit to be processed separately, and the left leg posture signal, right leg posture signal and head posture signal corresponding to each step of the right leg under the running speed signal are also treated as a unit to be processed separately. The running speed includes walking speed and running speed.

[0048] The control host 50 is also used to determine whether the head posture signal in any unit to be processed under any running speed signal is within the amplitude range of the head posture signal corresponding to the running speed. If so, the unit to be processed is retained; otherwise, the unit to be processed is discarded, thereby obtaining the units to be processed retained under N running speed signals.

[0049] Specifically, determine whether the head posture signal in any unit to be processed under the nth running speed signal is within the amplitude range of the head posture signal corresponding to the running speed. If so, confirm to retain the unit to be processed and use it as a retained processing unit; otherwise, confirm to discard the unit to be processed and use it as a discarded processing unit. Count the retained number of reserved processing units under the nth running speed signal, and discard all discarded processing units when the retained number reaches a set number. When the retained number does not reach the set number, determine the first number of discarded processing units where both the front and rear units to be processed are reserved processing units. When the first number reaches the difference between the set number and the reserved number, select the discarded processing units where the front and rear units to be processed corresponding to the difference are both reserved processing units. For any selected processing unit to be discarded, replace the head posture signal in the selected processing unit to be discarded based on the average value of the amplitudes of the head posture signals in the front and rear reserved processing units. amplitude, and the replaced processing units constitute the reserved processing units; when the first number does not reach the difference between the set number and the reserved number, first, for any of the to-be-discarded processing units whose front and rear to-be-discarded units are all reserved processing units, the amplitude of the head posture signal in the selected to-be-discarded processing unit is replaced based on the average value of the amplitude of the head posture signal in the front and rear reserved processing units, and the replaced processing units constitute the reserved processing units, and then only one to-be-discarded processing unit corresponding to the second number of continuous segments of to-be-discarded processing units is retained, and for the only to-be-discarded processing units corresponding to the second number, the amplitude of the head posture signal in the only to-be-discarded processing unit is replaced based on the average value of the amplitude of the head posture signal in the front and rear reserved processing units, and the replaced processing units constitute the reserved processing units, so as to obtain the reserved processing units under the nth running speed signal, and similarly, the reserved processing units under N running speed signals are obtained, and the first number + the second number = the difference between the set number and the reserved number.

[0050] The control host 50 is also used to splice the left leg posture signal, the right leg posture signal and the head posture signal in each reserved processing unit reserved under a certain running speed signal, to obtain the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal under the running speed signal, thereby obtaining the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal under N running speed signals.

[0051] The control host 50 is also used to pre-process the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal under N running speed signals to obtain the pre-processed left leg posture signal, the pre-processed right leg posture signal and the pre-processed head posture signal under N running speed signals.

[0052] The control host 50 is also used to extract features from the preprocessed left leg posture signal, the preprocessed right leg posture signal and the preprocessed head posture signal under any running speed signal, and obtain the time domain features of the left leg posture, the frequency domain features of the left leg posture, the time domain features of the right leg posture, the frequency domain features of the right leg posture, the time domain features of the head posture and the frequency domain features of the head posture under the running speed signal to form a one-dimensional feature matrix, based on which an N-dimensional feature matrix constructed under N running speed signals can be obtained.

[0053] The control host 50 is also used to input the N-dimensional feature matrix into a target balance ability assessment model constructed based on a random forest regression model for learning and training, to assess and display the degree of balance dysfunction in vestibular rehabilitation patients, and the degree of balance dysfunction includes no degree of dysfunction, mild degree of dysfunction, moderate degree of dysfunction and severe degree of dysfunction.

[0054] The control host 50 is used to use a speaker to play the vestibular rehabilitation exercise adjustment suggestions given by the doctor or rehabilitation therapist based on the assessed degree of balance dysfunction of the vestibular rehabilitation patient.

[0055] In this embodiment, based on the left leg posture signal, right leg posture signal and head posture signal of the vestibular rehabilitation patient, the degree of balance dysfunction of the vestibular rehabilitation patient can be obtained, and instant adjustment suggestions can be provided to the vestibular rehabilitation patient, and voice commands can be issued through the built-in speaker of the VR glasses.

[0056] In this embodiment, advanced signal processing technology and machine learning algorithms are used to quickly and accurately analyze the patient's posture changes and provide immediate adjustment suggestions to help the patient correct gait and balance problems.

[0057] In this embodiment, the control host can be equipped with a wireless communication module, allowing doctors or rehabilitation therapists to remotely monitor the training progress and posture changes of vestibular rehabilitation patients and adjust the treatment plan in time.

[0058] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that these are only examples, and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. An intelligent gait training and evaluation system for vestibular rehabilitation, characterized in that: It includes an intelligent gait training device, VR glasses for vestibular rehabilitation patients to wear, a left leg strap, a right leg strap and a control host. The intelligent gait training device includes a fixed base, a treadmill for gait training of vestibular rehabilitation patients is fixed at the middle position of the surface of the fixed base, a plurality of protective pillars are fixed along the edge of the fixed base, a plurality of protective pillars are fixed on the top of the plurality of protective pillars, a plurality of elastic protective long strips are fixed at the edge of the protective guardrail, an elastic protective belt for surrounding the waist of the vestibular rehabilitation patient is fixed at the end of the plurality of elastic protective long strips, and the elastic protective belt is located directly above the treadmill; The left leg strap is provided with a left leg posture sensor, the right leg strap is provided with a right leg posture sensor, the VR glasses are provided with a head posture sensor, and the treadmill is provided with a Hall sensor. The treadmill, the VR glasses, the left leg posture sensor, the right leg posture sensor, the head posture sensor and the Hall sensor are all electrically connected to the control host, and the control host stores the amplitude range of the head posture signal corresponding to each running speed; The left leg posture sensor is used to detect the left leg posture signal of the vestibular rehabilitation patient, the right leg posture sensor is used to detect the right leg posture signal of the vestibular rehabilitation patient, the head posture sensor is used to detect the head posture signal of the vestibular rehabilitation patient, and the Hall sensor is used to detect the running speed signal of the treadmill; The control host is used to control the treadmill to be at N different running speeds, collect the left leg posture signal, right leg posture signal and head posture signal of the vestibular rehabilitation patient under each running speed signal, and divide the left leg posture signal, right leg posture signal and head posture signal under the running speed signal in an alternating order of one step of the left leg and one step of the right leg. The left leg posture signal, right leg posture signal and head posture signal corresponding to each step of the left leg under the running speed signal are all separately used as a unit to be processed, and the left leg posture signal, right leg posture signal and head posture signal corresponding to each step of the right leg under the running speed signal are also separately used as a unit to be processed. The running speed includes walking speed and running speed, and N is a positive integer; The control host is also used to determine whether the head posture signal in any unit to be processed under any running speed signal is within the amplitude range of the head posture signal corresponding to the running speed, and if so, retain the unit to be processed, otherwise, discard the unit to be processed, thereby obtaining the units to be processed retained under N running speed signals; The control host is also used to splice the left leg posture signal, the right leg posture signal and the head posture signal in each unit to be processed that is retained under a certain running speed signal, to obtain the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal under the running speed signal, thereby obtaining the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal under N running speed signals; The control host is also used to pre-process the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal under N running speed signals to obtain the pre-processed left leg posture signal, the pre-processed right leg posture signal and the pre-processed head posture signal under N running speed signals; The control host is also used to extract features from the pre-processed left leg posture signal, the pre-processed right leg posture signal and the pre-processed head posture signal under any running speed signal, and obtain the left leg posture time domain features, the left leg posture frequency domain features, the right leg posture time domain features, the right leg posture frequency domain features, the head posture time domain features and the head posture frequency domain features under the running speed signal to form a one-dimensional feature matrix, based on which an N-dimensional feature matrix formed under N running speed signals can be obtained; The control host is also used to input the N-dimensional feature matrix into a target balance ability assessment model constructed based on a random forest regression model for learning and training, to assess and display the degree of balance dysfunction in vestibular rehabilitation patients, and the degree of balance dysfunction includes no disorder, mild disorder, moderate disorder and severe disorder.

2. The intelligent gait training and evaluation system for vestibular rehabilitation according to claim 1, characterized in that: The control host is also used to determine whether the head posture signal in any unit to be processed under the nth running speed signal is within the amplitude range of the head posture signal corresponding to the running speed, and if so, confirm to retain the unit to be processed and use it as a reserved processing unit, otherwise confirm to discard the unit to be processed and use it as a discarded processing unit, count the reserved number of reserved processing units under the nth running speed signal, and discard all the discarded processing units when the reserved number reaches a set number, and when the reserved number does not reach the set number, determine the first number of discarded processing units where both the front and rear units to be processed are reserved processing units, and when the first number reaches the difference between the set number and the reserved number, select the discarded processing units where the front and rear units to be processed corresponding to the difference are both reserved processing units, and for any selected processing unit to be discarded, replace the selected processing unit to be discarded based on the average value of the amplitude of the head posture signal in the front and rear reserved processing units. The amplitude of the head posture signal in the element is replaced to form a reserved processing unit. When the first number does not reach the difference between the set number and the reserved number, first, for any of the front and rear to-be-processed units are reserved processing units, the amplitude of the head posture signal in the selected to-be-discarded processing unit is replaced based on the average value of the amplitude of the head posture signal in the front and rear reserved processing units, and the reserved processing unit is replaced. Then, only one to-be-discarded processing unit corresponding to the second number of continuous segments of to-be-discarded processing units is retained. For the only to-be-discarded processing unit corresponding to the second number, the amplitude of the head posture signal in the only to-be-discarded processing unit is replaced based on the average value of the amplitude of the head posture signal in the front and rear reserved processing units, and the reserved processing unit is replaced. Thus, the reserved processing unit under the nth running speed signal is obtained, and similarly, the reserved processing units under N running speed signals are obtained. The first number + the second number = the difference.

3. The intelligent gait training and evaluation system for vestibular rehabilitation according to claim 1, characterized in that: The control host is used to control the treadmill to be at N different running speeds, collect the left leg posture signal, right leg posture signal and head posture signal of I historical vestibular rehabilitation patients under each running speed signal, and divide the left leg posture signal, right leg posture signal and head posture signal of I historical vestibular rehabilitation patients under the running speed signal in an alternating order of left leg taking a step and right leg taking a step, and the left leg posture signal, right leg posture signal and head posture signal corresponding to each left leg step of each historical vestibular rehabilitation patient under the running speed signal are all individually used as a historical unit to be processed, and the left leg posture signal, right leg posture signal and head posture signal corresponding to each right leg step of each historical vestibular rehabilitation patient under the running speed signal are also individually used as a historical unit to be processed, and I is a positive integer; The control host is also used to determine whether the head posture signal in any unit to be processed of any historical vestibular rehabilitation patient under any running speed signal is within the amplitude range of the head posture signal corresponding to the running speed, and if so, retain the historical unit to be processed, otherwise, discard the historical unit to be processed, thereby obtaining the historical units to be processed retained by each historical vestibular rehabilitation patient under N running speed signals; The control host is also used to splice the left leg posture signal, the right leg posture signal and the head posture signal in each historical to-be-processed unit retained by a historical vestibular rehabilitation patient under a certain running speed signal, and obtain the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal of each historical vestibular rehabilitation patient under the running speed signal, thereby obtaining the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal of each historical vestibular rehabilitation patient under N running speed signals; The control host is also used to pre-process the spliced ​​left leg posture signal, the spliced ​​right leg posture signal and the spliced ​​head posture signal of each historical vestibular rehabilitation patient under N running speed signals, and obtain the pre-processed left leg posture signal, the pre-processed right leg posture signal and the pre-processed head posture signal of each historical vestibular rehabilitation patient under N running speed signals; The control host is also used to extract features from the pre-processed left leg posture signal, the pre-processed right leg posture signal and the pre-processed head posture signal of any historical vestibular rehabilitation patient under any running speed signal, and obtain the left leg posture time domain features, left leg posture frequency domain features, right leg posture time domain features, right leg posture frequency domain features, head posture time domain features and head posture frequency domain features of the historical vestibular rehabilitation patient under the running speed signal to form a one-dimensional feature matrix, based on which an N-dimensional feature matrix formed under each running speed signal of each historical vestibular rehabilitation patient can be obtained; The control host is also used to construct a feature data set {(x1, y1), (x2, y2) ... (x I ,y I )},x i represents the N-dimensional feature matrix corresponding to the i-th historical vestibular rehabilitation patient, y i represents the degree of balance dysfunction corresponding to the i-th historical vestibular rehabilitation patient, and the i-th historical vestibular rehabilitation patient includes patients with no disability, patients with mild disability, patients with moderate disability, and patients with severe disability; The control host is also used to construct a feature training sample set and a feature verification sample set, the feature training sample set includes I1 feature data sets, and the feature verification sample set includes I2 feature data sets, I=I1+I2, I1 and I2 are both positive integers and I1 is much larger than I2; The control host is also used to build a balance ability assessment model based on the random forest regression model, input the feature training sample set into the balance ability assessment model for training, input the feature verification sample set into the trained balance ability assessment model for verification, and build a target balance ability assessment model.

4. The intelligent gait training and evaluation system for vestibular rehabilitation according to claim 3, characterized in that: A target balance ability evaluation model is constructed based on the random forest regression model, specifically: The control host is used to build a balance ability evaluation model based on a random forest regression model, initialize the hyperparameters of the balance ability evaluation model, initialize the speed and position of a particle swarm with a particle number M, set the particle search space dimension W and the maximum number of iterations T, the particle dimension W of each particle corresponds to the hyperparameter one by one, and each particle represents a set of hyperparameters; Each particle is used as a hyperparameter to construct an initial balance ability evaluation model. The feature training sample set is input into each initial balance ability evaluation model for training. The deviation mean square error of the model is used as the initial fitness of each particle. The initial fitness of each particle is used as the individual optimal fitness. The maximum value of the initial fitness of each particle is used as the initialization population optimal fitness. The current position of each particle is used as the initialization individual optimal position of each particle. The current position of the particle corresponding to the initialization population optimal fitness is used as the initialization population optimal position. Update the speed and position of each particle; Among them, m represents the current mth particle, w represents the current wth dimension search space, and t represents the current tth iteration number. and Respectively represent the speed of the current particle before and after update, and Respectively represent the position of the current particle before and after update, Q t represents the inertia weight, Q max represents the maximum inertia weight, Q min represents the minimum inertia weight, C1 and C2 represent the learning factors, represents the optimal position of the historical individual, represents the optimal position of the historical population; Update the historical individual optimal position and historical population optimal position of the particle swarm: calculate the updated fitness of each particle, compare the updated fitness of each particle with the historical individual optimal fitness of the particle, if the updated fitness of a particle is better than the historical individual optimal fitness, then update the historical individual optimal fitness of the particle = the updated fitness of the particle, and update the historical individual optimal position of the particle = the updated position of the particle, otherwise do not update; then take the maximum value of the historical individual optimal fitness of each particle as the current optimal fitness, if the current optimal fitness is better than the historical population optimal fitness, then update the historical population optimal fitness = the current optimal fitness, and update the historical population optimal position = the current position of the particle corresponding to the current optimal fitness, otherwise do not update; When the optimal fitness of the historical population reaches the set fitness or the number of iterations reaches the maximum number of iterations T, the optimal hyperparameters are obtained based on the particles at the optimal position of the historical population, and the target balance ability evaluation model is constructed using the optimal hyperparameters; the particles are arranged from best to worst according to the updated fitness, the sorted M1 particles are eliminated, and M1 particles are randomly generated based on the sorted M2 particles to replace the eliminated particles. The current positions of the randomly generated M1 particles are used as the positions of the corresponding particles after t updates, and the current speed of each particle in the randomly generated M1 particles is the average speed of the particles on both sides of the particle, M=M1+M2, t=t+1, and the operation of updating the speed and position of each particle is performed again.

5. The intelligent gait training and evaluation system for vestibular rehabilitation according to claim 1, characterized in that: A control host is placed on the outside of one of the protective pillars, an L-shaped cable arm is fixed to the top of the protective guardrail, the end of the lateral arm of the L-shaped cable arm is located directly above the treadmill, a VR cable is passed through the L-shaped cable arm, one end of the VR cable is connected to the control host, and the other end is connected to the VR glasses after passing through the L-shaped cable arm.

6. The intelligent gait training and evaluation system for vestibular rehabilitation according to claim 1, characterized in that: The fixed base is a circular fixed base, and four protective pillars are fixed on the circular fixed base at even intervals along the edge. A circular protective guardrail is fixed on the top of the four protective pillars, and four elastic protective strips are fixed on the edge of the circular protective guardrail at even intervals. Elastic protective belts are fixed at the ends of the four elastic protective strips.

7. The intelligent gait training and evaluation system for vestibular rehabilitation according to claim 1, characterized in that: The VR glasses are provided with a built-in speaker, and the control host is used to use the speaker to play the vestibular rehabilitation exercise adjustment suggestions given by the doctor or rehabilitation therapist based on the assessed degree of balance dysfunction of the vestibular rehabilitation patient.

Images