Non-implantable wearable double-layer electrode tongue touch force feedback control system and device

Abstract

The non-implantable wearable double-layer electrode tongue touch force feedback control system and device relate to the field of human-computer interaction technology, and solve the problem that the existing tongue touch device obtains little electrode information, including a support body, an information acquisition and transmission module, a tongue touch information perception layer connected to the support body, a tongue touch force feedback layer and a colloid layer, a lower electrode contact 1 is provided on the lower surface of the tongue touch information perception layer, an upper electrode contact 1 is provided on the upper surface, a through hole is provided on the colloid layer corresponding to the upper electrode contact 1, and an electrode contact 2 is provided on the lower surface of the tongue touch force feedback layer corresponding to the through hole, and the tongue touches the tongue touch information perception layer with force, so that the upper electrode contact 1 can contact the corresponding electrode contact 2, and at this time, the upper electrode contact 1 and / or the electrode contact 2 send the electrode signal of the upper electrode contact 1 to the information acquisition and transmission module. Through the present invention, not only the electrode signal of the lower electrode contact 1 can be obtained, but also the electrode signal of the upper electrode contact 1 with a large feedback tongue touch force can be obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of human-computer interaction, and in particular to a non-implantable wearable double-layer electrode tongue touch force feedback control system and device. Background Art

[0002] At present, the auxiliary interactive feedback devices that solve the problem of independent life for patients with hand disability include brain-computer interface, eye-computer interface, surface electromyography interface, etc. These systems have disadvantages such as unstable signals, poor interactive experience, and weak anti-fatigue ability. There are also problems such as complex device structure, high cost, and unsuitable for patients with moderate to severe disabilities with severe physical disabilities, especially difficult to adapt to the application of patients with severe upper limb or hand disability (spinal nerve injury).

[0003] Relevant research has found that the tongue is the "third hand" of human beings and one of the strongest muscles in the human body. The tip of the tongue functions like the fingertips, and the movement scale of the tongue touch is almost the same as that of the hand touch. The tongue can provide complex movement control and manipulation capabilities. If the characteristics of the tongue can be effectively utilized and intelligent devices can be controlled through the tongue-machine interaction device (Tongue Drive System, TDS), the problems of autonomous rehabilitation in the early stages of disability and self-care after disability for patients with moderate to severe disabilities can be solved.

[0004] A patent entitled "Tongue-machine interaction model and device" with publication number CN106648114A discloses a single-layer electrode feedback tongue touch auxiliary device, the technology of which requires the user to touch the palate electrode point with the tongue to realize the control device of human-machine interaction. The device only realizes the corresponding control of human-machine interaction by touching the sensor with the tongue. The tongue touch device has few functions, and the information obtained by tongue touch is little, and the touch force of the tongue touch cannot be reflected, resulting in few functions that can be realized by human-machine interaction, which greatly limits the control function and control accuracy based on tongue touch. Based on this, the present invention provides a non-implantable wearable double-layer electrode tongue touch force feedback control system and device.

[0005] Existing tongue touch devices do not take into account the different characteristics of human oral cavity, nor the differences in tongue movement ability of disabled patients, and do not provide personalized control methods based on individual characteristics. There is an urgent need for a method to divide command control areas according to the user's physical condition. Summary of the invention

[0006] In order to solve the problem that the existing tongue touch device cannot have a detection structure based on the tongue touch force and can obtain little electrode information, and there is an urgent need for a method of dividing the command control area according to the user's situation, the present invention provides a non-implantable, wearable double-layer electrode tongue touch force feedback control system and device, as well as a command control area division method.

[0007] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0008] The non-implantable wearable double-layer electrode tongue touch force feedback control device comprises a support body, a tongue touch information sensing layer and an information collection and transmission module, wherein the tongue touch information sensing layer is connected to the support body, a lower electrode contact 1 is arranged on the lower surface of the tongue touch information sensing layer, and the lower electrode contact 1 is connected to the information collection and transmission module, and when the tongue touches the lower electrode contact 1, the lower electrode contact 1 sends an electrode signal of the lower electrode contact 1 to the information collection and transmission module, and the tongue touch force feedback control device further comprises a tongue touch force feedback layer and a colloid layer, wherein the tongue touch force feedback layer, the colloid layer and the tongue touch information sensing layer are connected to the support body, and a lower electrode contact 1 is arranged on the lower surface of the tongue touch information sensing layer, and the lower electrode contact 1 is connected to the information collection and transmission module, and when the tongue touches the lower electrode contact 1, the lower electrode contact 1 sends an electrode signal of the lower electrode contact 1 to the information collection and transmission module, and the tongue touch force feedback control device further comprises a tongue touch force feedback layer and a colloid layer, and the tongue touch force feedback layer, the colloid layer and the tongue touch information sensing layer are connected to the support body, and a lower electrode contact 1 is arranged on the lower surface of the tongue touch information sensing layer, and the lower electrode contact 1 is connected to the information collection and transmission module, and the lower electrode contact 1 sends an electrode signal of the lower electrode contact 1 ... The layers are arranged sequentially from top to bottom, an upper electrode contact 1 is arranged on the upper surface of the tongue touch information perception layer, a through hole is arranged on the colloid layer corresponding to the upper electrode contact 1, and an electrode contact 2 is arranged on the lower surface of the tongue touch force feedback layer corresponding to the through hole, and the electrode contact 2 is connected to the information collection and transmission module; by touching the tongue touch information perception layer with force, the colloid layer can be deformed so that the upper electrode contact 1 can contact the corresponding electrode contact 2, and when the electrode contact 1 contacts the electrode contact 2, the upper electrode contact 1 and\or the electrode contact 2 send the electrode signal of the upper electrode contact 1 to the information collection and transmission module.

[0009] The double-layer electrode tongue touch force feedback control system including the non-implantable wearable double-layer electrode tongue touch force feedback control device also includes a wireless communication device, and the electrode signal of the lower electrode contact point 1 and the electrode signal of the upper electrode contact point 1 collected by the information collection and transmission module are sent to the wireless communication device.

[0010] The method for dividing the command control area of ​​the double-layer electrode tongue touch force feedback control system comprises the following steps:

[0011] Step 1: After the user wears the tongue touch force feedback control device, the user touches the tongue touch information sensing layer (1) with the tongue for multiple times, and the information collection and transmission module (5) obtains a user tongue touch data set including electrode signals of a plurality of upper electrode contact points 1 and electrode signals of a plurality of lower electrode contact points 1;

[0012] Step 2, the command input area models the task of learning the precise command control area as a first Markov model, and the command input area pre-trains the feature vector of the first Markov model according to the general tongue touch area data set to obtain the trained first Markov model. At the same time, the command input area division module receives the user tongue touch data set and performs correct data pre-processing on the data in the user tongue touch data set to obtain the correct tongue touch data set, and fine-tunes the parameters of the trained first Markov model according to the correct tongue touch data set to obtain a new first Markov model; after inputting the correct tongue touch data set into the new first Markov model, the state data of the tongue touch area after processing is output, and the state data of the tongue touch area after processing is used as the input of the policy gradient algorithm, and the policy gradient algorithm is used to adjust along the gradient direction until the optimal strategy is obtained, and the precise command control area divided according to the optimal strategy is obtained;

[0013] Step 3: Determine whether it is necessary to perform fuzzy processing on the precise instruction control area. If not, the precise instruction control area is used as the instruction control area to be finally used, and the instruction control area division is completed; if necessary, proceed to step 4;

[0014] Step 4. Input the user tongue touch data in the user tongue touch data set corresponding to the precise command control area and the general tongue touch area data set into the convolutional neural network. The convolutional neural network classifies the user tongue touch data in the user tongue touch data set corresponding to the precise command control area by learning the general tongue touch area data set, establishes a second Markov model and inputs the data output by the convolutional neural network into the second Markov model, performs Q-Learning algorithm training on the output of the second Markov model to obtain a fuzzy command control area. The fuzzy command control area is used as the command control area finally used, and the command control area division is completed.

[0015] The beneficial effects of the present invention are:

[0016] The non-implantable wearable double-layer electrode tongue touch force feedback control system and device of the present invention can not only obtain the electrode signal of the lower electrode contact one, but also obtain the electrode signal of the upper electrode contact one by setting the tongue touch force feedback layer, the colloid layer and the upper electrode contact one. The tongue touch force can be reflected by the contact between the electrode contact two and the upper electrode contact one, and the magnitude of the tongue touch force can be fed back by the electrode signal of the upper electrode contact one. The non-implantable wearable double-layer electrode tongue touch force feedback control system and device of the present invention provide a basis for realizing multi-functional and high-precision control in human-computer interaction. Based on the present invention, it is also possible to avoid human-computer interaction operation events caused by the tongue accidentally touching the lower electrode contact one.

[0017] The command control area division method of the double-layer electrode tongue touch force feedback control system of the present invention adopts end-to-end data processing, establishes the MDP model, and then uses the policy gradient algorithm to continuously learn and adjust along the gradient direction until the appropriate area, so as to obtain the matching control instructions for patients with different degrees; all command control areas are trained simultaneously through CNN, CNN classification is adopted, and then the Q-Learning algorithm is used to continuously update to obtain the optimal matching command area, so that the system can achieve the personalized control purpose for patients with limb disabilities; fully consider the different characteristics of human oral cavity, consider the differences in the ability of disabled patients to move their tongues, and realize personalized command control area division according to individual characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a structural diagram of the disassembled state of the non-implantable wearable double-layer electrode tongue touch force feedback control system and device of the present invention.

[0019] Figure 2 This is a schematic diagram of the human-computer interaction application of the non-implantable, wearable double-layer electrode tongue touch force feedback control system and device of the present invention.

[0020] Figure 3 This is a flow chart of step 2 of the instruction control area division method of the present invention.

[0021] Figure 4 This is a flow chart of step 4 of the instruction control area division method of the present invention.

[0022] In the figure: 1. tongue touch information sensing layer, 101. upper electrode contact one, 2. colloid layer, 201. through hole, 3. tongue touch force feedback layer, 301. electrode contact two, 4. support body, 401. bite groove, 5. information collection and transmission module, 6. wireless communication device. DETAILED DESCRIPTION

[0023] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0025] like Figure 1 As shown, the non-implantable wearable double-layer electrode tongue touch force feedback control device includes a support body 4, a tongue touch force feedback layer 3, a colloid layer 2, a tongue touch information perception layer 1, and an information collection and transmission module 5.

[0026] The support body 4 is used to support the tongue touch force feedback layer 3, the colloid layer 2 and the tongue touch information sensing layer 1. The support body 4 is provided with an occlusal groove 401 corresponding to the upper teeth of the user.

[0027] The information collection and transmission module 5 is disposed in the support body 4 and connected to the tongue touch information sensing layer 1 , and is used for receiving the electrode signal sent by the tongue touch information sensing layer 1 and sending out the received electrode signal.

[0028] The tongue touch information sensing layer 1 is connected to the support body 4 and electrically connected to the information collection and transmission module 5. The tongue touch information sensing layer 1 includes a substrate 1 and an electrode contact 1, wherein the electrode contact 1 is arranged on the substrate 1, and the electrode contact 1 includes a lower electrode contact 1 arranged on a lower side of the substrate and an upper electrode contact 101 arranged on an upper side of the substrate 1. The number of the lower electrode contact 1 is usually multiple, and the lower electrode contact 1 is connected to the information collection and transmission module 5 by connecting the lower flexible circuit 1, and the upper electrode contact 101 is connected to the information collection and transmission module 5 by connecting the upper flexible circuit 1. The lower electrode contact 1 is connected to the information collection and transmission module 5 by the lower connecting wire 1, and the upper electrode contact 101 is connected to the information collection and transmission module 5 by the upper connecting wire 1. Specifically, the lower electrode contact 1 and the lower connecting wire 1 are plated on a lower side surface of the substrate, and the lower electrode contact 1 and the lower connecting wire 1 constitute a flexible electrode layer, and the upper electrode contact 101 and the upper connecting wire 1 are plated on an upper side surface of the substrate 1, and the upper electrode contact 101 and the upper connecting wire 1 constitute a flexible electrode layer. The lower side of the tongue touch information sensing layer 1 corresponds to the tongue, and the tongue can touch the lower electrode contact 1. When the tongue touches the lower electrode contact 1, the touched lower electrode contact 1 sends the electrode signal of the lower electrode contact 1 to the information collection and transmission module 5 through the lower connecting line 1.

[0029] The colloid layer 2 is connected to the upper surface of the tongue touch information sensing layer 1 , and is made of a flexible material. A through hole 201 is provided on the colloid layer 2 corresponding to the upper electrode contact 101 . The upper electrode contact 101 is located in the through hole 201 .

[0030] The tongue touch force feedback layer 3 is connected to the colloid layer 2 and is located on the upper side of the colloid layer 2. The tongue touch force feedback layer 3 includes a substrate 2 and an electrode contact 2 301. The electrode contact 2 301 is arranged on the lower surface of the substrate 2. The electrode contact 2 301 is arranged corresponding to the through hole 201 of the colloid layer 2, that is, the electrode contact 2 301 is arranged corresponding to the upper electrode contact 1 101. The electrode contact 2 301 is located in the through hole 201. The electrode contact 2 301 is connected to the information collection and transmission module 5 through the connecting line 2. Specifically, a flexible circuit 2 and an electrode contact 2 301 are plated on the lower surface of the substrate 2. The flexible circuit 2 and the electrode contact 2 301 constitute a flexible electrode layer. When the tongue touches the tongue touch information sensing layer 1 and the force exerted by the tongue on the tongue touch information sensing layer 1 is greater than a certain value, the colloid layer 2 is squeezed and deformed, so that the electrode contact 2 301 can contact the upper electrode contact 1 101. When the force exerted by the tongue on the tongue touch information sensing layer 1 is not greater than the above-mentioned certain value, the colloid layer 2 is reset, and the electrode contact 2 301 and the upper electrode contact 1 101 are disconnected. The number of the upper electrode contact 1 101 and the electrode contact 2 301 is the same, both of which are several, and it is not limited whether they are evenly arranged. The tongue exerts different forces, the colloid layer 2 is squeezed and deformed differently, and the number of contacts between the electrode contact 2 301 and the upper electrode contact 1 101 is different; the contact positions between the tongue and the tongue touch information sensing layer 1 are different, and the positions where the colloid layer 2 is deformed are different, which leads to different contacts between the electrode contact 2 301 and the upper electrode contact 1 101. When electrode contact 2 301 contacts upper electrode contact 101, upper electrode contact 101 contacting electrode contact 2 301 sends the electrode signal of upper electrode contact 1 to information collection and transmission module 5 through upper connecting line 1, or electrode contact 2 301 contacting upper electrode contact 101 sends the electrode signal of upper electrode contact 1 to information collection and transmission module 5 through connecting line 2, or upper electrode contact 101 contacting electrode contact 2 301 and electrode contact 2 301 contacting upper electrode contact 1 101 work together to send the electrode signal of upper electrode contact 1 to information collection and transmission module 5.

[0031] A colloid layer 2 made of a flexible material is placed between the tongue touch force feedback layer 3 and the tongue touch information sensing layer 1, and a through hole 201 is processed on the surface of the colloid layer 2 according to the distribution of the electrode contacts of the tongue touch force feedback layer 3 and the tongue touch information sensing layer 1; firstly, the personalized oral surface 3D data of the disabled patient is collected, and the tongue touch force feedback layer 3 and the support body 4 are designed according to the oral surface 3D data, and the colloid layer 2 and the tongue touch information sensing layer 1 are designed according to the oral surface 3D data or according to the tongue touch force feedback layer 3 and the support body 4, and after the design is completed, the precision mold is processed and manufactured; the support body 4 and the tongue touch information sensing layer 1 are in an integrated structure to facilitate the disabled patient to wear; the tongue touch force feedback layer 3 is bonded to the upper surface of the colloid layer 2, and the lower surface of the colloid layer 2 is bonded to the tongue touch information sensing layer 1, and the three are sequentially bonded into one by an adhesive, and the electrode contact 2 301 does not contact the upper electrode contact 1 101. The upper surface of the tongue touch force feedback layer 3 contacts the upper palate of the user, and the support body 4 contacts the teeth of the user. After the user puts on the non-implantable wearable double-layer electrode tongue touch force feedback control device of the present invention, the tongue can touch the lower electrode contact one of the tongue touch information sensing layer 1 by moving the tongue, and the tongue can also give the tongue touch information sensing layer 1 an upward force, that is, the tongue exerts force on the palate, and the electrode contact two 301 can contact the upper electrode contact one 101. When neither the tongue touch force feedback layer 3 downward nor the tongue touch information sensing layer 1 upward is given, the electrode contact two 301 does not contact the upper electrode contact one 101, that is, when the glial layer 2 is not further squeezed and deformed, the electrode contact two 301 cannot contact the upper electrode contact one 101.

[0032] A non-implantable wearable double-layer electrode tongue touch force feedback control device is used for human-computer interaction. A posture signal can be obtained based on the electrode signal of the lower electrode contact point 1 obtained by the user's tongue touching the non-implantable wearable double-layer electrode tongue touch force feedback control device. A force signal can be obtained based on the electrode signal of the upper electrode contact point 1. The machine is controlled based on the posture signal and the force signal to achieve human-computer interaction, so that the user can control the machine's parameters such as direction, speed, and accuracy during human-computer interaction.

[0033] The non-implantable wearable double-layer electrode tongue touch force feedback control device also includes a power supply device, which is connected to the tongue touch force feedback layer 3, the tongue touch information perception layer 1 and the information collection and transmission module 5, and the power supply device supplies power to the tongue touch force feedback layer 3, the tongue touch information perception layer 1 and the information collection and transmission module 5.

[0034] A non-implantable wearable double-layer electrode tongue contact force feedback control system includes a non-implantable wearable double-layer electrode tongue contact force feedback control device and a wireless communication device 6. The information acquisition and transmission module 5 is connected to the wireless communication device 6, and the electrode signal of the lower electrode contact 1 and the electrode signal of the upper electrode contact 1 collected by the information acquisition and transmission module 5 can be sent to the wireless communication device 6. The wireless communication device 6 carries a data processor, which is connected to the controller. The data processor performs data processing on the electrode signal of the lower electrode contact 1 and the electrode signal of the upper electrode contact 1. The data processor processes the electrode signal of the lower electrode contact 1 to obtain a posture signal, and the data processor processes the electrode signal of the upper electrode contact 1 to obtain a force signal. Preferably, the data processor can also fuse the electrode signal of the upper electrode contact 1 and the electrode signal of the lower electrode contact 1 to obtain a posture signal and a force signal. The data processor sends the posture signal and the force signal to the controller, and the controller receives and performs relevant control on the human-machine interactive machine accordingly (fusion of the posture signal and the force signal).

[0035] The non-implantable wearable double-layer electrode tongue touch force feedback control system and device of the present invention can not only obtain the electrode signal of the lower electrode contact one, but also obtain the electrode signal of the upper electrode contact one. The contact situation between the electrode contact two 301 and the upper electrode contact one 101 can reflect the tongue touch force, and the electrode signal of the upper electrode contact one can feedback the magnitude of the tongue touch force. The non-implantable wearable double-layer electrode tongue touch force feedback control device of the present invention provides a basis for realizing multi-functional and high-precision control in human-computer interaction. Based on the present invention, it is also possible to avoid human-computer interaction operation events caused by the tongue accidentally touching the lower electrode contact one.

[0036] like Figure 2 Through the non-implantable wearable double-layer electrode tongue touch force feedback control system, the tongue touch and tongue processing can be sensed based on its electrode contacts. The obtained electrode signal is processed by the information acquisition and transmission module 5 and the wireless communication device 6 to obtain the posture signal and the force signal. The corresponding control of the (machine) auxiliary device (such as a manipulator, an intelligent wheelchair, a smart home, etc.) is obtained through the personalized fusion control algorithm, and then the (machine) auxiliary device realizes the effect on the person.

[0037] For a non-implantable wearable double-layer electrode tongue touch force feedback control system, a plurality of instruction input areas are provided thereon, and each instruction control area has its corresponding lower electrode contact one and upper electrode contact one 101. According to different users, the specific position of the instruction input area is different, the corresponding electrode contacts are different, and the number of corresponding electrode contacts may also be different. Usually, the tongue touch device is provided with four instruction control areas, namely, the instruction control area for upward instructions, the instruction control area for downward instructions, the instruction control area for left instructions, and the instruction control area for right instructions. The tongue touch device of the present invention also includes an instruction input area division module, and the data processor is loaded with the instruction input area division module. The instruction input area division module is used to divide the instruction control area according to the electrode signal of the lower electrode contact one and the electrode signal of the upper electrode contact one obtained by the user touching the information perception layer. It is preferred to divide the optimal control area for each type of control instruction, and divide the instruction control area according to the ability of the tongue to achieve touch (tongue touch force, tongue touch area, etc.). The instruction input area division module can be set in the wireless communication device 6. The instruction control area division method of the instruction input area division module includes the following steps:

[0038] Step 1: After the user wears the tongue touch device, he touches the tongue touch information perception layer 1 with his tongue for many times. The information acquisition and transmission module 5 obtains a plurality of electrode signals of the upper electrode contact 1 and a plurality of electrode signals of the lower electrode contact 1. The information acquisition and transmission module 5 sends the electrode signals of the upper electrode contact 1 and the electrode signals of the lower electrode contact 1 to the instruction input area division module. The electrode signal is binary data. All the electrical signal results are used as the user's tongue touch data set, and the user's tongue touch data set includes the electrode signals of the upper electrode contact 1 and the electrode signals of the lower electrode contact 1.

[0039] Step 2: The command input area models the task of learning the precise command control area as a Markov (MDP) model, which is called the first Markov model. The command input area pre-trains the feature vector of the first Markov model according to the general tongue touch area data set to obtain the trained first Markov model. At the same time, the command input area division module receives the user tongue touch data set, and correctly pre-processes the data in the user tongue touch data set to obtain the correct tongue touch data set; according to the correct tongue touch data set, the parameters of the trained first Markov model are fine-tuned to obtain a new first Markov model, and the correct tongue touch data set is input into the new first Markov model. The new first Markov model outputs the state data 1 of the tongue touch area after processing, and the state data 1 of the tongue touch area after processing is used as the input of the policy gradient algorithm, and the policy gradient algorithm is used to adjust along the gradient direction until the optimal strategy is obtained, and the precise command control area on the tongue touch information perception layer 1 is obtained according to the optimal strategy.

[0040] Adopting end-to-end optimization strategy, Figure 3 The specific process of this step is described as follows:

[0041] The instruction input area division module is loaded with a tongue touch area data set that is common to the double-layer electrode tongue touch force feedback control system. The instruction input area models the task of learning the precise instruction control area as a one-step Markov (MDP) model, which is referred to as the first Markov model in this article; the instruction input area division module pre-trains the feature vector of the first Markov model according to the common tongue touch area data set to obtain a trained Markov model, and at the same time, the instruction input area division module receives the user's tongue touch data set and performs correct data pre-processing on the data of the user's tongue touch data set to obtain a correct tongue touch data set; then, based on the correct tongue touch data set, the parameters of the trained first Markov model are fine-tuned to obtain a new first Markov model. The instruction input area division module inputs the correct tongue touch data set into the new Markov model and then outputs the state data of the processed tongue touch area (that is, the empty instruction area is eliminated). The state data of the processed tongue touch area includes the force, contact point and number of contacts when the tongue contacts the tongue touch information perception layer 1. The state data of the processed tongue touch area is used as the input of the policy gradient algorithm (corresponding to Figure 3 The output of the policy gradient algorithm is the probability of selecting a certain type of precise instruction control area.

[0042] At this time, the strategy π can be described as a function containing parameters θ, where e∈R d’ As the parameter vector of the strategy, d' represents the dimension of θ, so R d’ Represents d'-dimensional space, that is, the strategy π is:

[0043] π θ (s, a) = P (a | s, θ) ≈ π (a | s) (1)

[0044] Where s represents the state of the tongue touch area, a represents the action of the tongue touch area collected, and s and a in formula (1) are taken from the data in the state data of the tongue touch area after processing. P(α|s, θ) represents the state transition probability, which is the probability of taking any possible action a under the current state s and a certain parameter setting θ.

[0045] After expressing the strategy as a continuous function, the optimization method of continuous function can be used to find the optimal strategy.

[0046] The selected area strategy trajectory τ can be defined as:

[0047] τ={s1,a1,r1,s2,a2,r2,…,s i , a i , r i ,…,sk , a k , r k} (2)

[0048] Among them, s i is the state of the tongue contact area at step i in the strategy trajectory; a i is the tongue contact area action selected in step i; r i To take action a at step i i The reward value obtained after the operation is, i = 1, 2, ..., k, where k represents the total number of steps in the strategy trajectory. If the external device (the machine for human-computer interaction) executes the instruction correctly, the reward is 1, if it is not executed or the execution is wrong, the reward is 0. In order to maximize the expected cumulative reward, a parameterized strategy π is defined θ The expected cumulative rewards are as follows:

[0049]

[0050] in, is the cumulative reward value of the selected regional strategy trajectory, p θ (τ) represents the probability of a trajectory appearing under parameter θ, Denotes τ with respect to the distribution p θ (τ), R(τ) represents the return of a trajectory τ. After maximizing the expectation, the gradient can be estimated, and then the parameter θ is updated according to the policy formula until the policy converges to obtain the optimal policy. The action generated by the optimal policy selects the instruction input area with the largest reward, that is, the optimized instruction input area is divided into areas, and the precise instruction control area is obtained. Since the gradient of the metric function represented by J(θ) is used to learn the policy parameters, the update of the parameters here is approximately the gradient ascent of J, and the formula is as follows:

[0051]

[0052] in, Represents the policy gradient; η represents the learning rate, the step size when the gradient rises. Thus, continuous training and optimization are carried out to achieve the optimal strategy for obtaining the maximum cumulative reward value in the self-learning process, and then to achieve optimal control. According to the optimal strategy, the precise instruction control area is divided, and the division of the precise instruction control area is completed.

[0053] Each time the above step (ie, step 2) is executed, the precise instruction control area division of one type of instruction is completed, and step 2 is repeated until the precise instruction control area division of all types of instructions is completed, and then step 3 is performed.

[0054] Step 3: The user or the user's service personnel determines whether it is necessary to fuzzify the precise instruction control area according to the user's physical condition (the specific situation of the user's tongue touching the tongue touch information perception layer 1, such as the flexibility of the tongue, etc.). If not, the precise instruction control area is used as the final instruction control area, that is, the division of the instruction control area is completed; if necessary, proceed to step 4, and fuzzify the precise instruction control area through step 4 to obtain a fuzzy instruction control area. The precise instruction control area corresponds to one or several specific electrode contacts, and the fuzzy instruction control area corresponds to the specific electrode contact and some electrode contacts located around the specific electrode contact.

[0055] Step 4, in the instruction input area, the user tongue touch data in the general tongue touch area dataset and the user tongue touch data in the precise instruction control area corresponding to the user tongue touch dataset are input into the convolutional neural network. The convolutional neural network classifies the user tongue touch data in the user tongue touch dataset corresponding to the precise instruction control area by learning the general tongue touch area dataset; a Markov model is established (the Markov model here is referred to as the second Markov model in this paper) and the data output by the convolutional neural network is input into the second Markov model. The state data 2 of the processed tongue touch area output by the second Markov model is trained by the Q-Leaming algorithm to obtain a fuzzy instruction control area. The fuzzy instruction control area is used as the final instruction control area to complete the division of the instruction control area.

[0056] Figure 4 This is the block diagram of step 4. Step 4 is specifically as follows:

[0057] For example, in an experiment where the four types of command areas of up, down, left, and right of the user experiment are trained simultaneously, and the user tongue touch data set corresponding to the precise command control area is trained uniformly, a convolutional neural network CNN is used to first classify the data. The input of the convolutional neural network is a general tongue touch area data set and the user tongue touch data in the user tongue touch data set corresponding to the precise command control area. The user tongue touch data in the user tongue touch data set corresponding to the precise command control area is classified by learning the general tongue touch area data set, that is, classified according to the characteristics of the command control area into four categories of up, down, left, and right. During the training process, the areas are understood and classified, and approximate areas are discarded to make the data information more accurate. The second Markov model is established in the instruction input area, and the data output by the convolutional neural network is input into the second Markov model. The data output by the second Markov model is the state data 2 of the tongue touch area after processing. The state data 2 of the tongue touch area after processing is trained by the Q-Learning algorithm, that is, the Q-learning algorithm is used to create a Q table. It is necessary to initialize each Q value of the Q table, take an action, that is, select a certain area and observe the current state and the reward given by the external device. The state represents the sequence of the current tongue touch area and the general control area; the action represents the selection of the appropriate area; the reward refers to the reward obtained by the system. If the external device executes the instruction correctly, the reward is 1, and if it is not executed or the execution is wrong, the reward is 0. Then the Bellman equation is used to update the Q value Q(s, a):

[0058] NewQ(s,a)←Q(s,a)+α[R(s,a)+γQmax(s′,a′)-Q(s,a)] (5)

[0059] Among them, NewQ(s, a) represents the new Q value of the state action; Q(s, a) is the current Q value; α is the step size parameter, also known as the learning rate; R(s, a) is the reward for taking action a in state s; s and a in formula (5) are taken from the state data 2 of the tongue touch area; γ is the discount factor, which defines how much influence the future rewards have on the calculation of the current reward. When the discount factor is 0, the reward only considers the immediate reward. When γ=1, all future rewards will be fully calculated into the current reward; Qmax(s′, a′) is the maximum expected future reward. By updating the Q table, the best control area can be selected all the time, and the optimal control area for each type of control command can be obtained, that is, the division of the fuzzy command control area is completed, and personalized control is finally achieved.

[0060] The command control area division method of the double-layer electrode tongue tactile force feedback control system of the present invention uses end-to-end data processing, establishes the MDP model, and then uses the policy gradient algorithm to continuously learn and adjust along the gradient direction until the appropriate area, so as to obtain the matching control instructions for patients with different degrees. All command control areas are trained simultaneously through CNN, CNN classification is adopted, and then the Q-Learning algorithm is used to continuously update to obtain the optimal matching command area, so that the system can achieve the purpose of further personalized control for patients with limb disabilities. The command control area division method of the present invention fully considers the different characteristics of the human oral cavity, and considers the differences in the tongue movement ability of disabled patients, realizes personalized command control area division according to individual characteristics, and constructs a real-time interactive control system of people (disabled people), machines (auxiliary equipment), and tongue tactile interactive interface devices through tactile feedback and force feedback.

Claims

1. A non-implantable wearable double-layer electrode tongue touch force feedback control device, comprising a support body (4), a tongue touch information sensing layer (1) and an information collection and transmission module (5), wherein the tongue touch information sensing layer (1) is connected to the support body (4), a lower electrode contact 1 is provided on the lower surface of the tongue touch information sensing layer (1), and the lower electrode contact 1 is connected to the information collection and transmission module (5), and when the tongue touches the lower electrode contact 1, the lower electrode contact 1 sends an electrode signal of the lower electrode contact 1 to the information collection and transmission module (5), characterized in that: The tongue touch force feedback control device further comprises a tongue touch force feedback layer (3) and a colloid layer (2), wherein the tongue touch force feedback layer (3), the colloid layer (2) and the tongue touch information sensing layer (1) are arranged in sequence from top to bottom, an upper electrode contact point 1 (101) is arranged on the upper surface of the tongue touch information sensing layer (1), a through hole (201) is arranged on the colloid layer (2) corresponding to the upper electrode contact point 1 (101), and an electrode contact point 2 (301) is arranged on the lower surface of the tongue touch force feedback layer (3) corresponding to the through hole (201). 1), the electrode contact 2 (301) is connected to the information collection and transmission module (5); by using the tongue to forcefully touch the tongue touch information sensing layer (1), the colloid layer (2) can be deformed so that the upper electrode contact 1 (101) can contact the corresponding electrode contact 2 (301), and when the electrode contact 1 contacts the electrode contact 2 (301), the upper electrode contact 1 (101) and / or the electrode contact 2 (301) send the electrode signal of the upper electrode contact 1 to the information collection and transmission module (5).

2. The non-implantable wearable double-layer electrode tongue touch force feedback control device according to claim 1, characterized in that: The tongue touch force feedback control device further comprises a power supply device, wherein the power supply device is connected to the tongue touch force feedback layer (3), the tongue touch information perception layer (1) and the information collection and transmission module (5).

3. A double-layer electrode tongue touch force feedback control system comprising a non-implantable wearable double-layer electrode tongue touch force feedback control device as claimed in any one of claims 1 to 2, characterized in that: The system also includes a wireless communication device (6), and the electrode signal of the lower electrode contact point 1 and the electrode signal of the upper electrode contact point 1 collected by the information collection and transmission module (5) are sent to the wireless communication device (6).

4. The double-layer electrode tongue touch force feedback control system according to claim 3, characterized in that: The wireless communication device (6) is loaded with a data processor, which is connected to a controller. The data processor processes the electrode signal of the lower electrode contact 1 to obtain a posture signal, processes the electrode signal of the upper electrode contact 1 to obtain a force signal, and sends the posture signal and the force signal to the controller of the human-computer interaction machine.

5. The double-layer electrode tongue touch force feedback control system according to claim 3, characterized in that: The wireless communication device (6) is loaded with a command input area division module, and the command input area division module is used to divide the command control area according to the electrode signal of the lower electrode contact point 1 and the electrode signal of the upper electrode contact point 1.

6. The command control area division method of the double-layer electrode tongue touch force feedback control system according to claim 5, characterized in that: The steps include: Step 1: After the user wears the tongue touch force feedback control device, the user touches the tongue touch information sensing layer (1) with the tongue for multiple times, and the information collection and transmission module (5) obtains a user tongue touch data set including electrode signals of a plurality of upper electrode contact points 1 and electrode signals of a plurality of lower electrode contact points 1; Step 2, the command input area models the task of learning the precise command control area as a first Markov model, and the command input area pre-trains the feature vector of the first Markov model according to the general tongue touch area data set to obtain the trained first Markov model. At the same time, the command input area division module receives the user tongue touch data set and performs correct data pre-processing on the data in the user tongue touch data set to obtain the correct tongue touch data set, and fine-tunes the parameters of the trained first Markov model according to the correct tongue touch data set to obtain a new first Markov model; after inputting the correct tongue touch data set into the new first Markov model, the state data of the tongue touch area after processing is output, and the state data of the tongue touch area after processing is used as the input of the policy gradient algorithm, and the policy gradient algorithm is used to adjust along the gradient direction until the optimal strategy is obtained, and the precise command control area divided according to the optimal strategy is obtained; Step 3: Determine whether it is necessary to perform fuzzy processing on the precise instruction control area. If not, the precise instruction control area is used as the instruction control area to be finally used, and the instruction control area division is completed; if necessary, proceed to step 4; Step 4. Input the user tongue touch data in the user tongue touch data set corresponding to the precise command control area and the general tongue touch area data set into the convolutional neural network. The convolutional neural network classifies the user tongue touch data in the user tongue touch data set corresponding to the precise command control area by learning the general tongue touch area data set, establishes a second Markov model and inputs the data output by the convolutional neural network into the second Markov model, performs Q-Learning algorithm training on the output of the second Markov model to obtain a fuzzy command control area. The fuzzy command control area is used as the command control area finally used, and the command control area division is completed.

7. The command control area division method of the double-layer electrode tongue touch force feedback control system according to claim 6, characterized in that: The specific process of Q-Learning algorithm training is as follows: Create a Q table and initialize each Q value in the Q table; Use the Bellman equation to update the Q value Q(s,a): NewQ(s,a)←Q(s,a)+α[R(s,a)+γQmax(s′,a′)-Q(s,a)] Among them, NewQ(s, a) represents the new Q value of the state action; Q(s, a) is the current Q value; α is the step size parameter; R(s, a) is the reward for taking action a in state s; γ is the discount factor; Qmax(s′, a′) is the maximum expected future reward; The optimal control area for each type of control instruction is selected by updating the Q table of Q value, and the optimal control area is the fuzzy instruction control area.

Images