Wearable touch biological gait feedback system and method for Parkinson patient

Abstract

The invention discloses a wearable tactile biological gait feedback system and method for a Parkinson's patient, and belongs to the field of human gait rehabilitation, and the system comprises the following modules: a posture sensing module which obtains the information of an accelerometer, a magnetometer and a gyroscope at the positions of two feet by using an inertial sensor IMU bound to the surfaces of the two feet; the data processing module is used for receiving the data information of the posture sensing module, estimating the posture of the two feet, estimating the gait phase of the two feet by using the posture of the two feet, and finally outputting a signal according to a gait feedback stimulation strategy; and the tactile feedback module is used for receiving the stimulation signals of the data processing module and providing vibration stimulation signals at the back and thigh gastrocnemius muscles. According to the wearable tactile biological gait feedback system and method for the Parkinson's patient, related posture data provided by an IMU can be utilized, double-foot posture estimation can be completed through multi-rate multiple-correlation-entropy Kalman filtering, double-foot phases are estimated through a self-adaptive oscillator algorithm, and the wearable tactile biological gait feedback system and method for the Parkinson's patient are obtained. And implementing vibration stimulation based on a set gait feedback strategy.

Description

Technical Field

[0001] The present invention relates to the field of human gait rehabilitation, and more particularly to a wearable tactile biological gait feedback system and method for Parkinson's patients. Background Art

[0002] Parkinson's Disease (PD) is one of the common neurodegenerative diseases globally, and its incidence gradually increases with age. In China, the number of Parkinson's patients has exceeded 2.5 million, and this figure is expected to continue rising with the advent of an aging society. The typical symptoms of Parkinson's disease include bradykinesia, tremors, rigidity, etc. As the disease progresses, the quality of life of patients gradually declines. Among all motor symptoms, Freezing of Gait (FOG) is one of the most challenging, and approximately 80% of Parkinson's patients will experience freezing gait during the development of the disease. Freezing gait usually manifests as a sudden "stoppage" or short steps during walking, preventing patients from walking normally, greatly reducing their ability to take care of themselves, and even increasing the risk of falls. Currently, although drug therapy and deep brain stimulation surgery (DBS) can relieve the symptoms of Parkinson's patients to a certain extent, their effectiveness in relieving freezing gait is still limited. Therefore, finding a convenient and effective non-drug treatment method has become the focus of research.

[0003] As an emerging adjuvant treatment method, wearable tactile feedback devices have become a research hotspot in recent years due to their convenience and real-time nature. These devices can give prompts or interventions before or during the occurrence of freezing gait by applying tactile stimuli to the patient's body, thereby effectively reducing the occurrence frequency and duration of freezing gait. In terms of gait analysis, gait performance can be quantified using parameters such as stride length, stride frequency, speed, and double support time.

[0004] Most existing wearable tactile devices use multiple types of sensors to obtain gait information. Specifically, most wearable biofeedback devices developed for PD patients use small wearable sensors and actuators attached to the user's waist, head, or ankles, usually using inertial measurement unit (IMU) sensors and pressure sensors as sensors (Bowman T, Gervasoni E, Arienti C, et al. Wearable devices for biofeedback rehabilitation: a systematic review and meta-analysis to design application rules and estimate the effectiveness on balance and gait outcomes in neurological diseases[J]. Sensors, 2021, 21(10):3444.). Some scholars also use encoders installed on walkers as receivers of gait information (Wu H K, Chen H R, Chen W Y, et al. A novel instrumented walker for individualized visual cue setting for gait training in patients with Parkinson’s disease[J]. Assistive Technology, 2020, 32(4):203-213.).

[0005] In existing gait cueing research, external feedback in the forms of visual, auditory, and tactile cues has been used to help patients overcome movement difficulties. Existing cueing methods focus on providing visual cues at a fixed distance, or auditory and tactile cues at a fixed speed, and are calibrated for each patient. (Delval A, Krystkowiak P, Delliaux M, et al. Effect of external cueing on gait in Huntington's disease[J]. Movement disorders: official journal of the Movement Disorder Society, 2008, 23(10): 1446 - 1452. Schaefer R S. Auditory rhythmic cueing in movement rehabilitation: findings and possible mechanisms[J]. Philosophical Transactions of the Royal Society B: Biological Sciences, 2014, 369(1658): 20130402. Mikos V, Heng C H, Tay A, et al. A wearable, patient - adaptive freezing of gait detection system for biofeedback cueing in Parkinson's disease[J]. IEEE transactions on biomedical circuits and systems, 2019, 13(3): 503 - 515.) Vibrotactile information can be provided through electrical stimulation or vibration motors. However, the research devices based on the above - mentioned bio - tactile feedback use diverse but costly sensors. Moreover, existing cueing methods are mostly open - loop methods based on fixed routes or fixed designs, and there are relatively few studies on the closed - loop method (human - in - the - loop) considering the user himself. Summary of the Invention

[0006] The objective of the present invention is to provide a wearable tactile biological gait feedback system and method for Parkinson's patients, which can utilize relevant attitude data such as accelerometers, magnetometers, and gyroscopes provided by an IMU, complete bipedal attitude estimation through multi - rate multi - correlation entropy Kalman filtering, estimate bipedal phase using an adaptive oscillator algorithm, and implement vibration stimulation based on a set gait feedback strategy.

[0007] To achieve the above object, the present invention provides a wearable tactile biological gait feedback system for Parkinson's patients, including the following modules:

[0008] Attitude perception module: The accelerometer, magnetometer, and gyroscope information of the biped position are obtained by using the inertial sensor IMU strapped to the insteps of both feet;

[0009] Data processing module: Receives the data information of the attitude perception module, estimates the biped attitude, estimates the biped gait phase using the biped attitude, and finally outputs a signal according to the gait feedback stimulation strategy;

[0010] Tactile feedback module: Receives the stimulation signal of the data processing module and provides a vibration stimulation signal at the back and the gastrocnemius of the thigh.

[0011] Preferably, the data processing module includes the following sub-modules:

[0012] State estimation module: Estimates the biped attitude using multi-rate multiple correlation entropy extended Kalman filtering based on sensor information;

[0013] Phase estimation module: Estimates the phase of the discrete attitude signal using an adaptive oscillator method based on state information;

[0014] Vibration strategy module: Makes real-time decisions on vibration information based on the vibration strategy of gait feedback by processing phase information.

[0015] The present invention provides a calculation method for a wearable tactile biological gait feedback system for Parkinson's patients. In the state estimation module, first, a discrete linear time-invariant system equation is established according to the sensor signal and the attitude information:

[0016] x k =Φ k-1 x k-1 +w k

[0017]

[0018] where, H mag,k x k +v mag,k represents the update of magnetometer data, and H acc,k x k +v acc,k represents the update of accelerometer data; the state variable of the system is defined as the quaternion attitude x k =[x1,x2,x3,x4] T ∈R 4 ,Φ k-1 is the rotation matrix obtained by solving the gyroscope data at the k-1 moment and serves as the system matrix; x kis the system state at time k, x k-1 is the system state at time k-1, v mag,k is the observation noise of the magnetometer, v acc,k is the observation noise of the accelerometer, z k is the system observation; H mag,k is the magnetometer data observation matrix, H acc,k is the accelerometer data observation matrix; w k is the process noise;

[0019] Define the kernel bandwidth of the multi-correlation entropy, determine its parameters according to the standard deviation of the sensor data; determine the rotation matrix between the sensor coordinate system and the local navigation coordinate system NED at the initial moment according to the data of the accelerometer and magnetometer, and determine the state vector at the initial moment and the state covariance matrix of the posterior estimate according to the conversion relationship.

[0020] Preferably, the algorithm of the state estimation module includes three sensor data update stages, and the specific algorithm includes gyroscope data update, accelerometer data update and magnetometer data update.

[0021] Preferably, the gyroscope data update is used for the state transition step of the Kalman filter, and the formula for calculating the state transition matrix is as follows:

[0022]

[0023] Φ k = exp(Ω k )

[0024] where Φ k is the rotation matrix obtained by resolving the gyroscope data at time k, ω1, ω2, and ω3 are the three-axis data of the gyroscope at the current moment, and Ω k represents the quaternion rotation matrix obtained from the gyroscope data; then calculate the covariance matrix Q of the process noise:

[0025]

[0026] where CSI represents the information matrix for calculating the covariance, x1, x2, x3, x4 represent the four components of the state variables of the system at the current moment, I₃ represents the identity matrix of dimension 3, δt represents the time step of the gyroscope data, and its value is 0.0025 s; σ g is the standard deviation of the gyroscope data, and T represents the transpose; then update the covariance prior matrix at the current moment according to the covariance posterior matrix at the previous moment:

[0027] P prior,k = Φ k * P post,k-1 * Φ kT +Q;

[0028] Among them, P prior,k is the state covariance matrix of the prior estimate at the k-th moment, and P post,k-1 is the state covariance matrix of the posterior estimate at the (k - 1)-th moment.

[0029] Preferably, in the update of accelerometer data, first, it is necessary to calculate the observation matrix H of the accelerometer acc :

[0030]

[0031] Among them, g is the local gravitational acceleration value;

[0032] According to the principle of multi - correlation entropy, fuse the data of the state variable and the observed quantity; define the normalized change rate ∈ in the update process as the condition for terminating the iteration; update the state and the state covariance matrix according to the following formula:

[0033]

[0034] B r,acc = σ acc I3

[0035] When execute

[0036]

[0037]

[0038]

[0039] at the end

[0040]

[0041] Among them, is the system initial iteration variable at the k-th moment, is the system state variable obtained by rotating the gyroscope data, B r,acc is the multi - correlation entropy accelerometer kernel bandwidth matrix, σ acc is the standard deviation of the accelerometer data, is the system state variable calculated at the t-th iteration at the k-th moment, is the system state variable calculated at the (t - 1)-th iteration at the k-th moment, K acc,k is the Kalman gain of the accelerometer data at the k-th moment, P prior,k is the state covariance matrix of the prior estimate at the k-th moment, z acc is the observed variable of the system when updating the accelerometer data, eacc is the difference between the observable quantity of the system and z pre,acc As the information quantity for updating the system state, z pre,acc is the representation vector of the gravity vector in the sensor coordinate system, C z,acc is the information matrix for updating the accelerometer covariance matrix is the vector cross product, R acc is the covariance matrix of the accelerometer measurement noise, P post,k is the state covariance matrix of the posterior estimate at the k-th moment, and I is the identity matrix

[0042] Preferably, in the magnetometer data update, in the magnetometer data update, eliminating the longitudinal axis component includes the following steps

[0043] S1. The current attitude of the sensor is represented by a quaternion, and the quaternion describes the rotation from the Earth coordinate system to the sensor coordinate system. Calculate the conjugate of the quaternion

[0044] S2. Calculate the projection of the magnetic field in the direction of gravity, and remove the projected part from the magnetic field measurement value, only retaining the horizontal component of the magnetic field in the sensor coordinate system, thereby eliminating the interference caused by gravity

[0045] S3. Normalize the horizontal component of the magnetic field to keep it at unit length to ensure the stability of numerical calculation during attitude estimation, and then use the rotation matrix to calculate the theoretically observable magnetic field direction at the current attitude

[0046] S4. Use the processed horizontal component of the magnetic field as the observation data. Similarly, according to the principle of multiple correlation entropy, fuse the data of the state variable and the observable quantity; define the normalized change rate ∈ during the update process as the termination condition of the iteration; update the state and the state covariance matrix according to the following formula

[0047]

[0048] B r,mag =σ mag I3

[0049] When Execute

[0050]

[0051]

[0052]

[0053] At the end

[0054]

[0055] Among them, B r,mag is the multi-correlation entropy magnetometer kernel bandwidth matrix, σ mag is the standard deviation of the magnetometer data, K mag,k is the Kalman gain of the magnetometer data at the k-th moment, H mag is the observation matrix of the magnetometer, z mag is the observation variable when the system updates the magnetometer data, z pre,mag is the representation vector of the magnetic induction vector in the sensor coordinate system, C z,mag is the information matrix for updating the magnetometer covariance matrix, e mag is the difference between the observed quantity of the system and z pre,mag , and R mag is the covariance matrix of the magnetometer measurement noise.

[0056] Preferably, the gait in the phase estimation module is divided into a swing phase and a stance phase, and the key gait events include: when the foot is lifted and leaves the ground, it enters the swing phase; when the foot lands and touches the ground, it enters the stance phase; the input signal is modeled as an oscillatory function, which includes amplitude, phase, and instantaneous frequency; by adjusting the learning parameters of the adaptive oscillator, the state variables of the oscillator can gradually converge to the true phase of the input signal; a group of parallel adaptive oscillators are used, so that each oscillator learns different harmonic components of the input signal respectively.

[0057] Preferably, the vibration strategy in the vibration strategy module is as follows:

[0058] For users with unstable gait or uneven gait rhythm, rhythmic vibration is used to enhance gait consistency; before the foot is lifted, the tactile module provides a short vibration on the dorsal surface of the foot or the ankle to prompt the user to prepare to step and help form a stable gait rhythm; after the foot lands, the tactile module provides a slight vibration to feedback the completion of the gait, enabling the user to more clearly perceive the changes in the steps;

[0059] For abnormal gait, when a frozen gait is detected, if the step length suddenly shrinks and the step frequency increases abnormally but there is no effective walking, the tactile module provides a strong rhythmic vibration on the sole of the foot or the ankle, and through external rhythmic stimulation, guides the user to step and breaks the frozen state;

[0060] In the case of gait asymmetry, if the system detects that the step length or step frequency on one side is significantly lower than the other side, the vibration duration is increased on the slower side of the gait, enabling the user to realize the imbalance of the two-sided gait, and then actively adjusting the steps to achieve a more symmetrical walking pattern;

[0061] For the problem of dragging steps, if it is detected that the user's foot does not lift enough during the swing phase and there is a risk of dragging steps, the tactile module provides a strong short vibration on the dorsal surface of the foot to remind the user to raise the steps and reduce the risk of tripping or falling.

[0062] Therefore, the present invention adopts the above-mentioned wearable tactile biological gait feedback system and method for Parkinson's patients, and the specific beneficial effects are as follows:

[0063] (1) The perception module in the system is dedicated to real-time monitoring of the patient's gait state and integrates an IMU sensor. The system can be flexibly embedded in lightweight wearable devices such as insoles, straps, or waistbands, etc. Wearing it does not affect the patient's normal activities and is comfortable for long-term wearing.

[0064] (2) The data processing module of the system can continuously analyze gait data (such as gait cycle, gait speed, stride, etc.) according to the sensor data, using an efficient and stable state estimation algorithm and gait phase.

[0065] (2) The tactile stimulation strategy module in the system can adjust the stimulation strategy in real time according to the real-time gait data, using a closed-loop control algorithm. On the basis of considering individual differences, the system can intelligently select appropriate stimulation intensity, frequency, and stimulation site, and provide precise tactile feedback when necessary.

[0066] The technical solution of the present invention will be further described in detail below through the drawings and embodiments. Description of the Drawings

[0067] Figure 1 It is a schematic diagram of the closed-loop control of each module in an embodiment of the wearable tactile biological gait feedback system and method for Parkinson's patients of the present invention. Detailed Embodiment

[0068] The technical solution of the present invention will be further described below through the drawings and embodiments.

[0069] Unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meaning understood by those of ordinary skill in the field to which the present invention belongs.

[0070] Embodiment 1

[0071] Considering problems such as the scarcity of wearable biofeedback devices that only use inertial sensors (IMU) and the relatively scarce research on the closed-loop method (human-in-the-loop) of the user himself, the present invention designs a wearable bio-tactile feedback system and method composed of a micro-vibration motor, a micro-computation chip, a high-precision inertial sensor, and a multiple correlation entropy Kalman filtering algorithm.

[0072] Such as Figure 1As shown in the figure, the present invention provides a wearable tactile feedback system and method based on inertial sensors. The design concept of the present invention is as follows: through devices with gait and posture measurement, adaptive gait phase measurement, and tactile feedback functions, considering the closed-loop control framework of the user himself and considering safety and resource constraints during use, optimal gait perception and tactile feedback are achieved. The above parts as a whole constitute a complete set of software and hardware closed-loop wearable tactile feedback system, including three major modules:

[0073] I. Attitude perception module: A high-precision inertial sensor is strapped to the insteps of both feet. According to the sensor characteristics, the accelerometer, magnetometer, and gyroscope data of the positions of both feet are obtained. And through the CAN bus communication protocol, data is transmitted to the data processing module.

[0074] II. Data processing module: Receives information from the attitude perception module, receives the data information of the attitude perception module, estimates the attitudes of both feet using the attitude, and estimates the gait phases of both feet. Finally, signals are output according to the gait feedback stimulation strategy. The data processing module includes three sub-modules, namely the state estimation module, the phase estimation module, and the vibration strategy module.

[0075] 1) State estimation module:

[0076] Considering that the sampling rates of the accelerometer and magnetometer of the inertial sensor are different, which are 400Hz and 100Hz respectively. The multi-rate multiple correlation entropy extended Kalman filter based on sensor information is used to estimate the attitudes of both feet. First, a discrete linear time-invariant system equation is established according to the sensor signals and attitude information:

[0077] x k =Φ k-1 x k-1 +w k

[0078]

[0079] where, H mag,k x k +v mag,k represents the magnetometer data update, and H acc,k x k +v acc,k represents the accelerometer data update; the state variable of the system is defined as the quaternion attitude x k =[x1,x2,x3,x4] T ∈R 4 ,Φ k-1 is the rotation matrix obtained by resolving the gyroscope data at the (k - 1)th moment and serves as the system matrix; x k is the system state at the kth moment, x k-1 is the system state at the (k - 1)th moment, vmag,k is the observation noise of the magnetometer, v acc,k is the observation noise of the accelerometer, z k is the system observable; H mag,k is the magnetometer data observation matrix, H acc,k is the accelerometer data observation matrix; w k is the process noise;

[0080] Next, define the kernel bandwidth σ of the multi-correlation entropy mag , σ acc , and determine its parameters according to the standard deviation of the sensor data. Determine the rotation matrix R0 between the sensor coordinate system and the local navigation coordinate system (NED) at the initial moment according to the data of the accelerometer and the magnetometer, and determine the state vector at the initial moment and the state covariance matrix of the posterior estimate according to the conversion relationship.

[0081] The algorithm operation includes three sensor data update stages, and the specific algorithm is as follows:

[0082] (1) Gyroscope data update:

[0083] The gyroscope data update is used for the state transition step of the Kalman filter, and the formula for calculating the state transition matrix is as follows:

[0084]

[0085] Φ k = exp(Ω k )

[0086] where Φ k is the rotation matrix obtained by resolving the gyroscope data at time k, ω1, ω2, and ω3 are the three-axis data of the gyroscope at the current moment respectively, and Ω k represents the quaternion rotation matrix obtained from the gyroscope data; then calculate the covariance matrix Q of the process noise:

[0087]

[0088] where CSI represents the information matrix for calculating the covariance, x1, x2, x3, and x4 represent the four components of the state variables of the system at the current moment, I3 represents the identity matrix of dimension 3, δt represents the time step of the gyroscope data, and the value is 0.0025s; σ g is the standard deviation of the gyroscope data, and T represents the transpose; then update the covariance prior matrix at the current moment according to the covariance posterior matrix at the previous moment:

[0089] P prior,k = Φ k * P post,k-1 * Φ k T+Q;

[0090] where P prior,k is the state covariance matrix of the prior estimate at the k-th moment, and P post,k-1 is the state covariance matrix of the posterior estimate at the (k - 1)-th moment.

[0091] (2) Accelerometer data update:

[0092] First, the observation matrix H of the accelerometer needs to be calculated acc :

[0093]

[0094] where g is the local gravitational acceleration value;

[0095] Then, according to the principle of multi-correlation entropy, the data of the state variable and the observed quantity are fused; the normalized change rate ∈ in the update process is defined as the condition for terminating the iteration; the state and the state covariance matrix are updated according to the following formula:

[0096] [[ID=I27]]

[0097] B r,acc = σ acc I3

[0098] When Execute

[0099]

[0100]

[0101]

[0102] At the end

[0103]

[0104] where is the system initial iteration variable at the k-th moment, is the system state variable obtained by rotating the gyroscope data, B r,acc is the multi-correlation entropy accelerometer kernel bandwidth matrix, σ acc is the standard deviation of the accelerometer data, is the system state variable calculated at the t-th iteration at the k-th moment, is the system state variable calculated at the (t - 1)-th iteration at the k-th moment, K acc,k is the Kalman gain of the accelerometer data at the k-th moment, P prior,k is the state covariance matrix of the prior estimate at the k-th moment, z acce is the observed variable when the system updates the accelerometer data acc is the difference between the observed quantity of the system and z pre,acc as the information quantity for updating the system state, z pre,acc is the representation vector of the gravity vector in the sensor coordinate system, C z,acc is the information matrix for updating the accelerometer covariance matrix is the vector cross product, R acc is the covariance matrix of the accelerometer measurement noise, P post,k is the state covariance matrix of the posterior estimate at the k-th moment, and I is the identity matrix

[0105] (3) Magnetometer data update:

[0106] Due to the particularity of the earth's magnetic field, the magnetic vector received by the magnetometer does not point due north but has a longitudinal axis component. To eliminate the longitudinal axis component, the following steps are taken:

[0107] S1. The current attitude of the sensor is represented by a quaternion, which describes the rotation from the earth coordinate system to the sensor coordinate system. To restore from the sensor coordinate system to the earth coordinate system, the conjugate of this quaternion needs to be calculated, and this operation is equivalent to performing a reverse rotation, so that the vector in the sensor coordinate system can be converted back to the earth coordinate system

[0108] S2. Calculate the projection of the magnetic field in the direction of gravity and remove this projection part from the magnetic field measurement value, only retaining the horizontal component of the magnetic field in the sensor coordinate system, so as to eliminate the interference caused by gravity. In the earth coordinate system, the direction of gravity is fixed and usually points vertically downward. Through quaternion transformation, this standard gravity direction can be converted to the sensor coordinate system, so as to obtain the gravity direction in the sensor coordinate system. Since the magnetic field measurement is affected by the current attitude of the sensor, it is necessary to eliminate its component in the direction of gravity to ensure the independence of the magnetic field direction

[0109] S3. To ensure the stability of the magnetic field vector, after removing the gravity component, it is necessary to normalize the horizontal component of the magnetic field to keep it at unit length to ensure the stability of numerical calculation in the attitude estimation process. Subsequently, the rotation matrix is used to calculate the theoretically observable magnetic field direction at the current attitude, and this predicted value can be compared with the actually measured magnetic field direction of the sensor for error analysis and attitude correction. Through the above transformations and processing, the magnetic interference in the attitude estimation can be effectively reduced, so that the magnetometer data can accurately reflect the true direction of the sensor and improve the stability and accuracy of attitude recognition

[0110] S4. Then, use the processed horizontal magnetic field component as the observation data. Similarly, according to the principle of multi-correlation entropy, fuse the data of the state variables and the observables. Define the normalized change rate ∈ in the update process as the condition for terminating the iteration. Update the state and the state covariance matrix according to the following formula:

[0111]

[0112] B r,mag = σ mag I3

[0113] When Execute

[0114]

[0115]

[0116]

[0117] At the end

[0118]

[0119] Among them, B r,mag is the multi-correlation entropy magnetometer kernel bandwidth matrix, σ mag is the standard deviation of the magnetometer data, K mag,k [[ID=�9]]is the Kalman gain of the magnetometer data at the k-th moment, H mag is the observation matrix of the magnetometer, z mag is the observation variable of the system when updating the magnetometer data, z pre,mag is the representation vector of the magnetic induction vector in the sensor coordinate system, C z,mag is the information matrix for updating the magnetometer covariance matrix, e mag is the difference between the observable of the system and z pre,mag and R mag is the covariance matrix of the magnetometer measurement noise.

[0120] 2) Phase Estimation Module:

[0121] This module uses an adaptive oscillator method based on state information to estimate the phase of discrete attitude signals. Gait can be divided into the swing phase and the stance phase, and the key gait events include: Toe-Off (TO): The foot leaves the ground and enters the swing phase. Heel-Strike (HS): The foot touches the ground and enters the stance phase. The input signal can be modeled as an oscillatory function, which contains characteristic information such as amplitude, phase, and instantaneous frequency. By adjusting the learning parameters of the adaptive oscillator, the state variables of the oscillator can gradually converge to the true phase of the input signal, thus achieving accurate phase estimation. To improve stability, a group of parallel adaptive oscillators are adopted, and each oscillator learns different harmonic components of the input signal respectively. This method is equivalent to performing Fourier decomposition in a real-time environment to extract the phase information of the main harmonics, enabling the entire system to provide high-precision global phase estimation.

[0122] 3) Vibration strategy module:

[0123] This module processes the phase information and makes real-time decisions on vibration information based on the gait feedback vibration strategy. The vibration strategies are as follows:

[0124] For users with unstable gait or uneven gait rhythm, rhythmic vibration can be adopted to enhance gait consistency. Before the foot lift (TO), the tactile module provides a short vibration (100 ms) on the dorsal surface of the foot or the ankle to prompt the user to prepare to step and help form a stable gait rhythm. After the foot strike (HS), the tactile module provides a slight vibration (50 ms) to feedback the completion of the gait, enabling the user to more clearly perceive the changes in the steps. In addition, the system can compare the current gait frequency of the user with the target gait frequency. If the step frequency is too slow, the intensity of the vibration can be increased before TO to encourage a faster pace, thus helping the user maintain a reasonable walking rhythm.

[0125] For users with abnormal gaits, the tactile module can provide more targeted feedback to improve gait problems. When detecting Freezing of Gait (FOG), if the step length suddenly shrinks and the step frequency abnormally increases but there is no effective walking, the tactile module can provide strong rhythmic vibration (250 ms vibration + 250 ms interval) on the sole of the foot or the ankle. Through external rhythmic stimulation, it guides the user to step and breaks the frozen state.

[0126] In the case of gait asymmetry, if the system detects that the step length or step frequency on one side is significantly lower than the other side, the vibration duration (150 ms) can be increased on the slower side of the gait, enabling the user to be aware of the imbalance of the two-sided gait and then actively adjust the steps to achieve a more symmetric walking pattern.

[0127] For the shuffling problem, if it is detected that the user's foot does not lift enough during the swing phase and there is a risk of shuffling, the tactile module can provide a strong short vibration (200 ms) on the dorsal side of the foot to remind the user to improve their gait and reduce the risk of tripping or falling. This immediate tactile feedback can help the user adjust their gait in a timely manner and improve the stability and safety of walking.

[0128] III. Tactile feedback module: According to the vibration control signal output by the control module, output a PWM control signal to the motor at the stimulation site to control the micro vibration motor to vibrate. The placement positions of the micro vibration motors are the inner sides of the elbows of both arms, both sides of the spine, both sides of the waist, and the gastrocnemius tendon of the calf.

[0129] Therefore, the present invention adopts the above-mentioned tactile biological gait feedback system and method for Parkinson's patients, which can utilize the relevant attitude data such as accelerometers, magnetometers, and gyroscopes provided by the IMU, complete the bipedal attitude estimation through multi-rate multiple correlation entropy Kalman filtering, estimate the bipedal phase using the adaptive oscillator algorithm, and implement vibration stimulation based on the set gait feedback strategy.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A wearable tactile biological gait feedback system for Parkinson's patients, characterized in that, It includes the following modules: Attitude perception module: The accelerometer, magnetometer, and gyroscope information of the biped position are obtained by using the inertial sensor IMU strapped to the insteps of both feet; Data processing module: Receives the data information of the attitude perception module, estimates the biped attitude, estimates the biped gait phase using the biped attitude, and finally outputs a signal according to the gait feedback stimulation strategy; Tactile feedback module: Receives the stimulation signal of the data processing module and provides vibration stimulation signals at the back and gastrocnemius of the thigh.

2. The wearable tactile biological gait feedback system for Parkinson's patients according to claim 1, wherein: The data processing module includes the following sub-modules: State estimation module: Estimates the biped attitude using multi-rate multi-correlation entropy extended Kalman filtering based on sensor information; Phase estimation module: Estimates the phase of the discrete attitude signal using an adaptive oscillator method based on state information; Vibration strategy module: Based on the processed phase information, makes real-time decisions on vibration information based on the gait feedback vibration strategy.

3. The calculation method of a tactile biological gait feedback system wearable by Parkinson's patients according to any one of claims 1-2, characterized in that: In the state estimation module, first establish a discrete linear time-invariant system equation according to the sensor signal and attitude information: x k = Φ k-1 x k-1 + w k Among them, H mag,k x k +v mag,k represents the update of magnetometer data, and H acc,k x k +v acc,k represents the update of accelerometer data; Define the state variables of the system as the quaternion attitude x k = [x1, x2, x3, x4] T ∈R 4 , Φ k-1 is the rotation matrix obtained by resolving the gyroscope data at time k - 1 and serves as the system matrix; x k is the system state at time k, x k-1 is the system state at time k - 1, v mag,k is the observation noise of the magnetometer, v acc,k is the observation noise of the accelerometer, z k is the system observable; H mag,k is the magnetometer data observation matrix, H acc,k is the accelerometer data observation matrix; w k is the process noise; Define the kernel bandwidth of multi-correlation entropy, and determine its parameters according to the standard deviation of the sensor data; determine the rotation matrix between the sensor coordinate system and the local navigation coordinate system NED at the initial moment according to the data of the accelerometer and magnetometer, and determine the state vector at the initial moment and the state covariance matrix of the posterior estimate according to the conversion relationship.

4. The calculation method of a wearable tactile biological gait feedback system for Parkinson's patients according to claim 3, characterized in that: The algorithm of the state estimation module includes three sensor data update stages, and the specific algorithm includes gyroscope data update, accelerometer data update, and magnetometer data update.

5. The calculation method of a tactile biological gait feedback system wearable by Parkinson's patients according to claim 4, characterized in that: The gyroscope data update is used for the state transition step of Kalman filtering, and the formula for calculating the state transition matrix is as follows: Φ k = exp(Ω k ) Among them, Φ k is the rotation matrix obtained by resolving the gyroscope data at time k, ω1, ω2, and ω3 are the three-axis data of the gyroscope at the current moment, and Ω k represents the quaternion rotation matrix obtained from the gyroscope data; then calculate the covariance matrix Q of the process noise: Among them, CSI represents the information matrix for calculating covariance, x1, x2, x3, x4 represent the four components of the system state variables at the current moment, I3 represents the identity matrix of dimension 3, δt represents the gyroscope data time step, with a value of 0.0025 s; σ g is the standard deviation of the gyroscope data, T represents the transpose; and then the prior covariance matrix at the current moment is updated based on the posterior covariance matrix at the previous moment: P prior,k = Φ k * P post,k-1 * Φ k T + Q; where, P prior,k is the state covariance matrix of the prior estimate at the k-th moment, and P post,k-1 is the state covariance matrix of the posterior estimate at the (k-1)-th moment.

6. The calculation method of a wearable tactile biological gait feedback system for Parkinson's patients according to claim 5, characterized in that: In the acceleration data update, it is first necessary to calculate the observation matrix H of the accelerometer acc : where g is the local gravitational acceleration value; According to the multi-correlation entropy principle, fuse the data of the state variable and the observed quantity; define the normalized change rate ∈ in the update process as the condition for terminating the iteration; update the state and the state covariance matrix according to the following formula: B r,acc = σ acc I3 When Execute At the end Among them, is the system initial iteration variable at the k-th moment, is the system state variable obtained by rotating the gyroscope data, B r,acc is the multi-correlation entropy accelerometer kernel bandwidth matrix, σ acc is the standard deviation of the accelerometer data, is the system state variable calculated at the t-th iteration at the k-th moment, is the system state variable calculated at the (t - 1)-th iteration at the k-th moment, K acc,k is the Kalman gain of the accelerometer data at the k-th moment, P prior,k is the state covariance matrix of the prior estimate at the k-th moment, z acc is the observation variable when the system updates the accelerometer data, e acc is the difference between the observed quantity of the system and z pre,acc as the information quantity for updating the system state, z pre,acc is the representation vector of the gravity vector in the sensor coordinate system, C z,acc is the information matrix for updating the accelerometer covariance matrix, is the vector cross product, R acc is the covariance matrix of the accelerometer measurement noise, P post,k is the state covariance matrix of the posterior estimate at the k-th moment, and I is the identity matrix.

7. The calculation method of a wearable tactile biological gait feedback system for Parkinson's patients according to claim 4, characterized in that: In the magnetometer data update, the steps for eliminating the longitudinal axis component include the following: S1. The current attitude of the sensor is represented by a quaternion, and the conjugate of the quaternion is calculated, where the quaternion describes the rotation from the earth coordinate system to the sensor coordinate system; S2. Calculate the projection of the magnetic field in the direction of gravity, and remove the projected part from the magnetic field measurement value, only retaining the horizontal component of the magnetic field in the sensor coordinate system, so as to eliminate the interference caused by gravity; S3. Normalize the horizontal component of the magnetic field to keep it at unit length to ensure the stability of numerical calculation during attitude estimation, and then calculate the theoretically observable magnetic field direction at the current attitude using the rotation matrix; S4. Take the processed horizontal component of the magnetic field as the observed data. Similarly, according to the multi-correlation entropy principle, fuse the data of the state variable and the observed quantity; define the normalized change rate ∈ in the update process as the condition for terminating the iteration; update the state and the state covariance matrix according to the following formula: B r,mag = σ mag I3 When Execute At the end Among them, B r,mag is the multi-correlation entropy magnetometer kernel bandwidth matrix, σ mag is the standard deviation of the magnetometer data, K mag,k is the Kalman gain of the magnetometer data at the k-th moment, H mag is the observation matrix of the magnetometer, z mag is the observation variable when the system updates the magnetometer data, z pre,mag is the representation vector of the magnetic induction vector in the sensor coordinate system, C z,mag is the information matrix for updating the magnetometer covariance matrix, e mag is the difference between the observed quantity of the system and z pre,mag and R mag is the covariance matrix of the magnetometer measurement noise.

8. The calculation method of a wearable tactile biological gait feedback system for Parkinson's patients according to claim 3, characterized in that: The gait in the phase estimation module is divided into a swing phase and a stance phase. The key gait events include: when the foot leaves the ground at foot lift, entering the swing phase; when the foot touches the ground at foot strike, entering the stance phase. The input signal is modeled as an oscillatory function that includes amplitude, phase, and instantaneous frequency. By adjusting the learning parameters of the adaptive oscillator, the state variables of the oscillator can gradually converge to the true phase of the input signal. A set of parallel adaptive oscillators are used, enabling each oscillator to learn different harmonic components of the input signal respectively.

9. The calculation method of a wearable tactile biological gait feedback system for Parkinson's patients according to claim 3, characterized in that: The vibration strategy in the vibration strategy module is as follows: For users with unstable gait or uneven gait rhythm, rhythmic vibration is used to enhance gait consistency. Before foot lift, the tactile module provides a short vibration on the dorsal foot or ankle to prompt the user to prepare to step and help form a stable gait rhythm. After foot strike, the tactile module provides a slight vibration to feedback on gait completion, enabling the user to more clearly perceive the changes in the steps. For abnormal gait, when a freezing gait is detected, if the step length suddenly shrinks and the step frequency abnormally increases but there is no effective walking, the tactile module provides a strong rhythmic vibration on the sole of the foot or ankle. Through external rhythmic stimulation, the user is guided to step and break the freezing state. In the case of gait asymmetry, if the system detects that the step length or step frequency on one side is significantly lower than the other side, the vibration duration is increased on the slower side of the gait, enabling the user to be aware of the imbalance in the gait on both sides and then actively adjust the steps to achieve a more symmetric walking pattern. For the problem of dragging the foot, if it is detected that the user's foot does not lift enough during the swing phase and there is a risk of dragging the foot, the tactile module provides a strong short vibration on the dorsal foot to remind the user to lift the step and reduce the risk of tripping or falling.

Images