An assisted expression system for a tracheostomy patient and a wearable device

Abstract

The application relates to the field of intelligent auxiliary expression, and discloses an auxiliary expression system for a tracheostomy patient and a wearable device, wherein the system is provided with an operation collection device, a feature collection element, a training module, a data processing module and an alarm module; the operation collection device obtains an operation signal of a patient operation action in an operable state; the feature collection element collects individual feature data to obtain a feature data sequence in combination; the training module optimizes parameters of a feature correlation model; the data processing module calculates an individual feature data interval and a feature change parameter interval corresponding to each operation signal; the operation collection device is automatically connected to the alarm module in the operable state, and the alarm module is started to send an alarm signal when the operation collection device collects the operation signal. In the application, the operation action and the individual feature data are not matched in a mistaken touch state, the operation collection device cannot obtain the operation signal, the occurrence of the mistaken touch condition is avoided, and the accuracy of the auxiliary expression is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent auxiliary expression technology, specifically to an auxiliary expression system and wearable device for patients after tracheotomy. Background Technology

[0002] Tracheotomy is a surgical procedure used to treat severe airway obstruction. Patients often face the problem of being unable to communicate effectively through sound after tracheotomy. In response, patients can express their needs by writing on a whiteboard, using communication cards, or by using hand gestures. However, these methods of expression cannot meet the patient's needs in emergency situations.

[0003] To address these issues, some electronic auxiliary communication devices have emerged on the market, such as electronic speech synthesizers and gesture recognition devices. Speech synthesizers can emit corresponding voice information by pressing specific buttons to express the patient's needs. Medical staff will then take appropriate actions based on the received needs. For example, if the machine emits the voice of "pain," it will prompt medical staff to examine the patient's injury.

[0004] During the use of a speech synthesizer, in an emergency, the patient may accidentally press multiple buttons or other buttons, or press buttons when there is no need for the device. This can lead to incorrect expression of needs and improper operation by medical staff. Summary of the Invention

[0005] Therefore, the present invention provides an auxiliary expression system and wearable device for patients after tracheotomy, which effectively solves the technical problem in the prior art where accidental button touches lead to incorrect expression of needs, resulting in improper operation by medical staff.

[0006] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution: an auxiliary expression system for patients after tracheotomy, comprising: The operation acquisition device is switched to a locked state or an operable state. In the operable state, the operation acquisition device acquires the operation signal of the patient's operation action. Feature acquisition element, used to collect individual feature data of patients, in order to combine them to obtain a feature data sequence; The training module is communicatively connected to the operation acquisition device and the feature acquisition element. The training module constructs a feature association model and receives a dataset obtained by processing multiple sets of historical operation signals and feature data sequences. The dataset is then input into the neural network for training to optimize the parameters of the feature association model. The data processing module calculates the individual feature data range and feature change parameter range corresponding to each operation signal based on the feature association model. The data processing module acquires real-time individual feature data and determines whether it is within the individual feature data range. It combines the real-time individual feature data into a real-time data subsequence and calculates the real-time feature change parameters within the effective time period t based on the real-time data subsequence. It then determines whether the real-time feature change parameters are within the feature change parameter range. An alarm module is connected to the operation acquisition device. The operation acquisition device automatically connects to the alarm module when it is in an operable state, and activates the alarm module to send an alarm signal when the operation acquisition device acquires an operation signal. Specifically, when the real-time individual feature data is within the individual feature data range and the real-time feature change parameters are within the feature change parameter range, the operation acquisition device switches from the locked state to the operable state.

[0007] Furthermore, after acquiring multiple sets of historical operation signals and feature data sequences, the feature data sequences are cut into data segments with a time length of T to obtain several reference data subsequences; Each reference data subsequence consists of individual feature data of several consecutive time nodes. The time node corresponding to one of the individual feature data in each reference data subsequence is the same as the time node of one of the operation signals, and the overlapping time node is marked as the standard node t0. The reference data subsequences correspond one-to-one with the operation signals to form a dataset.

[0008] Furthermore, the starting time node of the reference data subsequence is t1, and the ending time node is t2; Where t2-t1=T, t0-t1>t2-t0.

[0009] Furthermore, the time period T0 within T is selected as the measurement period, and the characteristic change parameters of the reference data subsequence within the measurement period are calculated; The feature change parameter is the rate of change of individual feature data. The starting time node of time period T0 is t3, and the ending time node is t4. The individual feature data corresponding to the starting time node is a, and the individual feature data corresponding to the ending time node is b. The feature change parameter is S. ; The feature change parameters and their corresponding measurement periods are used as input data and substituted into the neural network.

[0010] Furthermore, the dataset is divided into a training set and a test set. The data in the training set is used as a single-channel grayscale image and input into the neural network for training. The parameters in the neural network are adjusted a second time using the test set to obtain the final feature association model.

[0011] Furthermore, different operation signals are used as inputs and substituted into the feature association model. After calculation, the corresponding individual feature data range, feature change parameter range, and measurement time period are output. The data processing module acquires real-time individual feature data, compares the real-time individual feature data with each individual feature data interval, and determines whether it is within one of the individual feature data intervals. Real-time individual characteristic data and all individual characteristic data within time period T are combined to obtain a real-time data subsequence. Based on the real-time data subsequence, the real-time characteristic change parameters within the effective time period t are calculated. The real-time characteristic change parameters are compared with each characteristic change parameter interval to determine whether the real-time characteristic change parameters are within one of the characteristic change parameter intervals. Among them, the effective time period t is less than or equal to the calculated time period T0; When it is determined that the real-time feature change parameter is not within one of the feature change parameter intervals, the real-time feature change parameter within the effective time period t is calculated using t-tx as the new effective time period t. The real-time feature change parameter is compared with each feature change parameter interval to determine whether the real-time feature change parameter is within one of the feature change parameter intervals. The process stops when the real-time feature change parameter is within one of the feature change parameter intervals. Set a threshold value x, and stop and output the final judgment result when the effective time period t reaches the value of x.

[0012] Furthermore, if the operation signal of the patient's operation action is acquired more than once at a single time point by the operation acquisition device, the operation acquisition device is switched to a locked state; An operation threshold is set. When the frequency of multiple operation signals of the same operation action acquired by the operation acquisition device exceeds the operation threshold, the operation acquisition device is switched to an operable state.

[0013] Furthermore, the individual characteristic data is the patient's heart rate.

[0014] To address the aforementioned technical problems, the present invention further provides the following technical solution: an assistive wearable device, comprising: Wearing main body; Operate the data acquisition equipment, feature acquisition elements, and alarm module; The wearable device is a wristband, on which a watch face is mounted; The operation acquisition device includes several function buttons on the dial and a circuit board inside the dial, with a switch installed on the circuit board; The feature acquisition element is a heart rate sensor located inside the dial, and the alarm module is a speaker located on the dial; The switch is connected to the speaker. When the operation acquisition device is in an operable state, after the function button is pressed, the switch and the circuit board transmits the operation signal to the speaker. The speaker then receives the signal and triggers an alarm.

[0015] To address the aforementioned technical problems, the present invention further provides the following technical solution: an assistive wearable device, comprising: Wearing main body; Operate the data acquisition equipment, feature acquisition elements, and alarm module; The main body to be worn is a finger sleeve; The operation acquisition device includes several function buttons on the finger sleeve and a circuit board inside the finger sleeve, with a switch installed on the circuit board; The feature acquisition element is a heart rate sensor installed inside the finger sleeve, and the alarm module is a speaker installed on the finger sleeve; The switch is connected to the speaker. When the operation acquisition device is in an operable state, after the function button is pressed, the switch and the circuit board transmits the operation signal to the speaker. The speaker then receives the signal and triggers an alarm.

[0016] Compared with the prior art, the present invention has the following advantages: In this invention, an operation acquisition device is used to acquire the operation signals of the patient's operation actions, and a feature acquisition element is used to collect the patient's individual feature data. The feature association model is optimized through a training module, and the individual feature data range and feature change parameter range corresponding to each operation signal are calculated based on the feature association model. In actual operation, when the real-time individual feature data is within the individual feature data range and the real-time feature change parameter is within the feature change parameter range, the operation acquisition device switches from a locked state to an operable state. In the case of accidental touch, the operation action does not match the individual feature data, and the operation acquisition device cannot acquire the operation signal, thus avoiding accidental touch and effectively improving the accuracy of auxiliary expression. Attached Figure Description

[0017] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0018] Figure 1 A structural block diagram of an auxiliary expression system for patients after tracheotomy provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the pruning of the reference data subsequence in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the selection of the measurement time period in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the selection of the effective time period in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, the present invention provides an auxiliary expression system for patients after tracheotomy, which includes an operation acquisition device, a feature acquisition element, a training module, a data processing module, and an alarm module.

[0021] The data acquisition device can be switched to a locked state or an operable state. In the operable state, the data acquisition device acquires the operation signals of the patient's operation actions.

[0022] The feature acquisition element is used to collect individual feature data of patients to combine them into a feature data sequence.

[0023] The training module communicates with the operation acquisition device and feature acquisition element. The training module builds a feature association model and receives a dataset obtained by processing multiple sets of historical operation signals and feature data sequences. The dataset is then input into the neural network for training to optimize the parameters of the feature association model.

[0024] Among them, the feature association model mainly aims to construct the association between different operational signals and the patient's individual feature data.

[0025] The data processing module calculates the individual feature data range and feature change parameter range corresponding to each operation signal based on the feature association model.

[0026] The data processing module acquires real-time individual characteristic data and determines whether it is within the individual characteristic data range. It combines the real-time individual characteristic data into real-time data subsequences and calculates the real-time characteristic change parameters within the effective time period t based on the real-time data subsequences. It then determines whether the real-time characteristic change parameters are within the characteristic change parameter range.

[0027] The alarm module is connected to the operation acquisition device. When the operation acquisition device is in an operable state, it automatically connects to the alarm module and activates the alarm module to issue an alarm signal when the operation acquisition device acquires an operation signal.

[0028] Specifically, when the real-time individual characteristic data is within the individual characteristic data range and the real-time characteristic change parameters are within the characteristic change parameter range, the operation acquisition device switches from the locked state to the operable state. When the operation acquisition device is in the operable state, the operation signal of the patient can be linked to the alarm module to issue an alarm signal.

[0029] In this invention, an operation acquisition device is used to acquire the operation signals of the patient's operation actions, and a feature acquisition element is used to collect the patient's individual feature data. The feature association model is optimized through a training module, and the individual feature data range and feature change parameter range corresponding to each operation signal are calculated based on the feature association model. In actual operation, when the real-time individual feature data is within the individual feature data range and the real-time feature change parameter is within the feature change parameter range, the operation acquisition device switches from a locked state to an operable state. In the case of accidental touch, the operation action does not match the individual feature data, and the operation acquisition device cannot acquire the operation signal, thus avoiding accidental touch and effectively improving the accuracy of auxiliary expression.

[0030] The entire process of system application includes the data acquisition stage, the training stage, and the operation stage. The data acquisition stage and the training stage are both preliminary data training stages, which are used to build the matching relationship between the operation signals of different operation actions and individual feature data.

[0031] During the data acquisition phase, the operation acquisition device is in an operable state by default. The feature acquisition element will collect the patient's individual feature data throughout the entire time period and combine them to obtain the feature data sequence. The operation acquisition device will acquire the operation signal of the patient's operation action when the patient performs the operation action. During the training phase, the feature data sequence and operation signal are processed and trained. During the operation phase, the data acquisition device is in a locked state by default, and the state will only be switched under specific conditions.

[0032] Before the training phase, after acquiring multiple sets of historical operation signals and feature data sequences, the feature data sequences need to be processed. Specifically, the feature data sequences are cut into data segments with a time length of T to obtain several reference data subsequences. Each reference data subsequence consists of individual feature data of several consecutive time nodes. The time node corresponding to one of the individual feature data in each reference data subsequence is the same as the time node of one of the operation signals, and the overlapping time node is marked as the standard node t0. The reference data subsequences correspond one-to-one with the operation signals to form a dataset.

[0033] In the above embodiments, the timelines of the feature data sequence and the operation signal are parallel, and there is an overlapping time node t0 in the two timelines. At this standard node t0, the operation acquisition device acquires the operation signal of the patient's operation action, such as... Figure 2 As shown, with the standard node t0 as the reference point, data segments of length T are diffused outward. In other words, the data segments of length T obtained from the diffusion are used as reference data subsequences. In actual data acquisition, several operation signals are acquired. For each operation signal, the standard node t0 can be cut into data segments of length T to obtain several reference data subsequences.

[0034] To ensure the effectiveness of the data analysis process, we assume that the starting time node of the reference data subsequence is t1 and the ending time node is t2, satisfying the following conditions: t2-t1=T, t0-t1>t2-t0. Since the patient's individual characteristic data generally changes significantly before the operation, the individual characteristic data before the operation has great reference value. The proportion of individual characteristic data collected between t1 and t0 is relatively large, which can ensure the effectiveness of the data analysis results.

[0035] In addition, the values ​​of (t2-t1) / (t0-t1) are consistent in each data segment of length T, ensuring that the proportion of data collected before and after the operation is consistent in each reference data subsequence.

[0036] Each operation signal after data processing corresponds one-to-one with each segment of reference data. The above processed operation signals and reference data segments are used as input data and substituted into the neural network for training.

[0037] Since the individual characteristic parameters under different operation signals may not differ much, and the range of individual characteristic parameters alone is too simplistic to judge whether there is an operation action, and the judgment accuracy is low, this invention further processes the reference data subsequence. Specifically, the T0 time period within T is selected as the calculation time period, and the characteristic change parameters of the reference data subsequence within the calculation time period are calculated. The feature change parameter is the rate of change of individual feature data, such as Figure 3 As shown, the starting time node of time period T0 is t3, and the ending time node is t4. The individual feature data corresponding to the starting time node is a, and the individual feature data corresponding to the ending time node is b. The feature change parameter is S. ; The feature change parameters and their corresponding measurement periods are used as input data and substituted into the neural network.

[0038] Each set of data fed into the neural network includes the operating signal, the reference data subsequence, and the feature change parameters. Multiple sets of data are combined to form a dataset. The neural network is used to construct the matching relationship between each operating signal and the individual feature data and feature change parameters in the reference data subsequence.

[0039] The dataset is divided into a training set and a test set. The data in the training set is used as a single-channel grayscale image to be input into the neural network for training. The parameters of the neural network are adjusted a second time using the test set to obtain the final feature association model.

[0040] Specifically, a commonly used recurrent neural network can be used for training. The specific steps are as follows: set the training functions, connection functions, and output functions for the hidden layers, connection layers, and output layers; set the expected minimum error, maximum number of iterations, and learning rate of the network; input the training set into the recurrent neural network model and train it; adjust the weights of each layer in the recurrent neural network model according to the error; determine whether the recurrent neural network model has converged, and training is complete after convergence; the algorithm converges when the error is less than the expected minimum error; and the algorithm ends when the maximum number of iterations is reached.

[0041] The optimized feature association model is obtained through neural network training.

[0042] Since in practical applications, patients' actions are limited to only a few fixed types, their corresponding individual characteristic data and characteristic change parameters are generally only within a certain range. Therefore, the range of individual characteristic data and characteristic change parameters corresponding to different operation signals can be directly obtained through the feature association model. Specifically, different operation signals are used as inputs and substituted into the feature association model. After calculation, the corresponding individual characteristic data range, characteristic change parameter range, and measurement time period are output.

[0043] The calculated individual characteristic data range is generally the range of individual characteristic data at time t0. This is mainly because, within the time period T, the initial individual characteristic data may not differ much from normal human data. Using individual characteristic data near time t1 as a reference to calculate the individual characteristic data range may not be of much reference value. However, individual characteristic data at time t0 often differs from normal human data. Therefore, the individual characteristic data at time t0 is directly used as a reference to obtain the individual characteristic data range after processing multiple datasets.

[0044] Each operation signal corresponds to an individual characteristic data range and a characteristic change parameter range.

[0045] The calculation period output by the neural network is the most informative period for analyzing overall characteristic changes. In practice, when a patient experiences an emergency, individual characteristic data increases (for example, heart rate), and then tends to stabilize after reaching a certain value. Before this, individual characteristic data undergoes significant changes at an unusual rate. This unusual change in individual characteristic data is the calculation period that needs to be calculated. Within this calculation period, individual characteristic data changes significantly, which is of great significance for judging whether the patient has a need to express themselves.

[0046] In actual patient use, the data processing module acquires real-time individual characteristic data, compares the real-time individual characteristic data with each individual characteristic data interval, and determines whether it falls within one of the individual characteristic data intervals. The real-time individual characteristic data and all individual characteristic data within time period T are combined to obtain a real-time data subsequence. Based on the real-time data subsequence, the real-time characteristic change parameters within the effective time period t are calculated. The real-time characteristic change parameters are compared with each characteristic change parameter interval to determine whether the real-time characteristic change parameters are within one of the characteristic change parameter intervals. The effective time period t is less than or equal to the measurement time period T0.

[0047] like Figure 4 As shown, assuming the current time point is q, q is the end time point of the effective time period t, the beginning time point of the effective time period t is p, the individual characteristic data at the end time point is B, and the individual characteristic data at the beginning time point is A, then the real-time characteristic change parameter corresponding to the effective time period t is (BA) / (qp). Determine whether the calculated real-time change parameter is within the characteristic change parameter range.

[0048] When it is determined that the real-time feature change parameter is not within one of the feature change parameter intervals, the real-time feature change parameter within the effective time period t is calculated using t-tx as the new effective time period t. The real-time feature change parameter is compared with each feature change parameter interval to determine whether the real-time feature change parameter is within one of the feature change parameter intervals. The process stops when the real-time feature change parameter is within one of the feature change parameter intervals. Set a threshold value x, and stop and output the final judgment result when the effective time period t reaches the value of x.

[0049] In the above embodiments, an initial t value is set, which is slightly less than the measurement period T0. Real-time feature change parameters are calculated based on this t value. If the real-time feature change parameters are not within one of the feature change parameter intervals, the t value is reduced and the real-time feature change parameters are calculated again. This process is repeated until it is determined that the real-time feature change parameters are within one of the feature change parameter intervals, or until the t value reaches the x value.

[0050] In addition to the above judgment methods, a smaller t value can be set directly to calculate the real-time characteristic change parameter of the t value. If the real-time characteristic change parameter is not within one of the characteristic change parameter intervals, the calculation is stopped and the calculation of the next real-time time point is performed. If the real-time characteristic change parameter is within one of the characteristic change parameter intervals, the calculation is stopped.

[0051] The above process is merely a judgment on whether an operation is possible at the current point in time. The above process needs to be repeated at the next point in time.

[0052] Based on the above process, and after the above analysis, it is determined that if the real-time individual feature data is in one of the individual feature data ranges and the real-time feature change parameters are in one of the feature change parameter ranges, the operation acquisition device should be directly switched to an operable state.

[0053] If more than one operation signal of the patient's operation is acquired during a single time point of the operation acquisition device, the operation acquisition device will be switched to the locked state; An operation threshold is set. When the frequency of multiple operation signals of the same operation action of the patient obtained by the operation acquisition device exceeds the operation threshold, the operation acquisition device is switched to an operable state.

[0054] In other words, when the data acquisition device receives more than one operation signal while it is in an operable state, it immediately switches back to the locked state and cannot trigger an alarm for the multiple operation signals.

[0055] If the frequency of multiple operation signals exceeds the operation threshold, it indicates that the patient is in an emergency, and the operation acquisition device is switched to an operable state.

[0056] In emergency situations, a patient's heart rate usually increases; therefore, the patient's individual heart rate is collected as an individual characteristic data.

[0057] Therefore, the above process uses real-time heart rate and heart rate change rate to determine whether the patient is likely to perform an operation. If an operation is likely to occur, the operation acquisition device is switched to an operable state; otherwise, it remains locked to prevent the patient from accidentally performing an operation.

[0058] The present invention also provides an assistive wearable device, comprising: Wearing main body; Such as the operation acquisition equipment and feature acquisition elements, alarm modules mentioned above; The main wearable component is a wristband, on which a watch face is mounted. The operation and data acquisition device includes several function buttons on the dial and a circuit board inside the dial, with a switch installed on the circuit board; The feature acquisition element is a heart rate sensor located inside the dial, and the alarm module is a speaker located on the dial. The switch is connected to the speaker. When the data acquisition device is in an operable state, the function button is pressed and the circuit board transmits the operation signal to the speaker. The speaker then receives the signal and triggers an alarm.

[0059] In this wearable device, multiple function buttons can be labeled with "holding breath", "pain", "sucking phlegm", and "holding urine". When in an operable state, the operation and data acquisition device can collect the corresponding operation signals and the speaker will emit the corresponding alarm signals.

[0060] The switch is a relay switch. In the operable state, the switch closes when the function button is pressed, and the circuit board can transmit the operation signal to the speaker. The speaker will then sound an alarm upon receiving the signal. In the locked state, the switch cannot be closed, and the corresponding operation signal cannot be transmitted to the speaker.

[0061] A volume control button can be located on the side of the dial to adjust the speaker volume.

[0062] The bracelet features a streamlined, ergonomic design to ensure comfort when worn. It is made of lightweight and soft silicone or TPU (thermoplastic polyurethane) to reduce skin irritation and is suitable for prolonged wear.

[0063] The present invention also provides an assistive wearable device, comprising: Wearing main body; Such as the operation acquisition equipment and feature acquisition elements, alarm modules mentioned above; The main part worn is a finger sleeve; The operation and acquisition device includes several function buttons on the finger sleeve and a circuit board inside the finger sleeve, on which a switch is installed; The feature acquisition element is a heart rate sensor located inside the finger sleeve, and the alarm module is a speaker located on the finger sleeve; The switch is connected to the speaker. When the data acquisition device is in an operable state, the function button is pressed and the circuit board transmits the operation signal to the speaker. The speaker then receives the signal and triggers an alarm.

[0064] In this wearable device, multiple function buttons can be labeled with "holding breath", "pain", "sucking phlegm", and "holding urine". When in an operable state, the operation and data acquisition device can collect the corresponding operation signals and the speaker will emit the corresponding alarm signals.

[0065] The finger sleeves feature a streamlined design with a simple and beautiful appearance. They are ergonomically designed to ensure comfort when worn. In addition, the finger sleeves are made of lightweight and soft medical-grade silicone to reduce skin irritation and make them suitable for long-term wear.

[0066] This wearable device is easier to operate than a wristband. The finger sleeve is worn on the middle finger, and the function button is placed on the palm side. The thumb of the hand can press the button. If the condition permits, the opposite hand can also be used to operate it.

[0067] A volume control button can be located on the side of the finger sleeve to adjust the speaker volume.

[0068] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. An auxiliary expression system for patients after tracheotomy, characterized in that, have: The operation acquisition device is switched to a locked state or an operable state. In the operable state, the operation acquisition device acquires the operation signal of the patient's operation action. Feature acquisition element, used to collect individual feature data of patients, in order to combine them to obtain a feature data sequence; The training module is communicatively connected to the operation acquisition device and the feature acquisition element. The training module constructs a feature association model and receives a dataset obtained by processing multiple sets of historical operation signals and feature data sequences. The dataset is then input into the neural network for training to optimize the parameters of the feature association model. The data processing module calculates the individual feature data range and feature change parameter range corresponding to each operation signal based on the feature association model. The data processing module acquires real-time individual feature data and determines whether it is within the individual feature data range. It combines the real-time individual feature data into a real-time data subsequence and calculates the real-time feature change parameters within the effective time period t based on the real-time data subsequence. It then determines whether the real-time feature change parameters are within the feature change parameter range. An alarm module is connected to the operation acquisition device. The operation acquisition device automatically connects to the alarm module when it is in an operable state, and activates the alarm module to send an alarm signal when the operation acquisition device acquires an operation signal. Specifically, when the real-time individual feature data is within the individual feature data range and the real-time feature change parameters are within the feature change parameter range, the operation acquisition device switches from the locked state to the operable state.

2. The auxiliary expression system for patients after tracheotomy according to claim 1, characterized in that, After acquiring multiple sets of historical operation signals and feature data sequences, the feature data sequences are cut into data segments with a time length of T to obtain several reference data subsequences. Each reference data subsequence consists of individual feature data of several consecutive time nodes. The time node corresponding to one of the individual feature data in each reference data subsequence is the same as the time node of one of the operation signals, and the overlapping time node is marked as the standard node t0. The reference data subsequences correspond one-to-one with the operation signals to form a dataset.

3. The auxiliary expression system for patients after tracheotomy according to claim 2, characterized in that, The starting time node of the reference data subsequence is t1, and the ending time node is t2; Where t2-t1=T, t0-t1>t2-t0.

4. The auxiliary expression system for patients after tracheotomy according to claim 2, characterized in that, Select the time period T0 within T as the measurement period, and calculate the characteristic change parameters of the reference data subsequence within the measurement period; The feature change parameter is the rate of change of individual feature data. The starting time node of time period T0 is t3, and the ending time node is t4. The individual feature data corresponding to the starting time node is a, and the individual feature data corresponding to the ending time node is b. The feature change parameter is S. ； The feature change parameters and their corresponding measurement periods are used as input data and substituted into the neural network.

5. The auxiliary expression system for patients after tracheotomy according to claim 4, characterized in that, The dataset is divided into a training set and a test set. The data in the training set is used as a single-channel grayscale image and input into the neural network for training. The parameters in the neural network are adjusted a second time using the test set to obtain the final feature association model.

6. The auxiliary expression system for patients after tracheotomy according to claim 4, characterized in that, Different operational signals are used as inputs and substituted into the feature association model. After calculation, the corresponding individual feature data range, feature change parameter range, and measurement time period are output. The data processing module acquires real-time individual feature data, compares the real-time individual feature data with each individual feature data interval, and determines whether it is within one of the individual feature data intervals. Real-time individual characteristic data and all individual characteristic data within time period T are combined to obtain a real-time data subsequence. Based on the real-time data subsequence, the real-time characteristic change parameters within the effective time period t are calculated. The real-time characteristic change parameters are compared with each characteristic change parameter interval to determine whether the real-time characteristic change parameters are within one of the characteristic change parameter intervals. Among them, the effective time period t is less than or equal to the calculated time period T0; When it is determined that the real-time feature change parameter is not within one of the feature change parameter intervals, the real-time feature change parameter within the effective time period t is calculated using t-tx as the new effective time period t. The real-time feature change parameter is compared with each feature change parameter interval to determine whether the real-time feature change parameter is within one of the feature change parameter intervals. The process stops when the real-time feature change parameter is within one of the feature change parameter intervals. Set a threshold value x, and stop and output the final judgment result when the effective time period t reaches the value of x.

7. The auxiliary expression system for patients after tracheotomy according to claim 6, characterized in that, If the operation signal of the patient's operation action is acquired more than once at a single time point in the operation acquisition device, the operation acquisition device will be switched to a locked state. An operation threshold is set. When the frequency of multiple operation signals of the same operation action acquired by the operation acquisition device exceeds the operation threshold, the operation acquisition device is switched to an operable state.

8. The auxiliary expression system for patients after tracheotomy according to claim 1, characterized in that, The individual characteristic data refers to the patient's heart rate.

9. A wearable device for assisting expression, characterized in that, have: Wearing main body; The operation acquisition device and feature acquisition element, and alarm module as described in any one of claims 1 to 8; The wearable device is a wristband, on which a watch face is mounted; The operation acquisition device includes several function buttons on the dial and a circuit board inside the dial, with a switch installed on the circuit board; The feature acquisition element is a heart rate sensor located inside the dial, and the alarm module is a speaker located on the dial; The switch is connected to the speaker. When the operation acquisition device is in an operable state, after the function button is pressed, the switch and the circuit board transmits the operation signal to the speaker. The speaker then receives the signal and triggers an alarm.

10. A wearable device for assisting expression, characterized in that, have: Wearing main body; The operation acquisition device and feature acquisition element, and alarm module as described in any one of claims 1 to 8; The main body to be worn is a finger sleeve; The operation acquisition device includes several function buttons on the finger sleeve and a circuit board inside the finger sleeve, with a switch installed on the circuit board; The feature acquisition element is a heart rate sensor installed inside the finger sleeve, and the alarm module is a speaker installed on the finger sleeve; The switch is connected to the speaker. When the operation acquisition device is in an operable state, after the function button is pressed, the switch and the circuit board transmits the operation signal to the speaker. The speaker then receives the signal and triggers an alarm.

Images