Pressing type insulin needle assisting device based on sound detection and working method

Abstract

The invention discloses a pressing type insulin needle assisting device based on sound detection and a working method.The outer shell, the inner shell and the middle shell of the needle assisting device are all vertically arranged, the inner shell and the middle shell are connected through a first buckle, a pull ring is arranged on the upper portion of the inner shell, a key connected with the inner shell is arranged in the middle of the pull ring, and the key is sleeved with a first spring; the button moves downwards to push the bottom of the inner shell to move downwards, the bottom of the inner shell is provided with a second buckle used for being connected with an infusion pipeline, the middle shell is sleeved with a second spring, the top of the inner wall of the outer shell is provided with an extending supporting and blocking part extending inwards, the bottom of the middle shell is provided with a step part, the step part is provided with at least two sliding locks extending outwards, and the sliding locks are transversely arranged. The sliding lock is sleeved with a third spring, an inclined face part is arranged on the outer side of the sliding lock, and a transverse guide rib extending inwards is arranged in the middle of the inner wall of the shell. The needle assisting device is simple in structure, convenient to operate and capable of being repeatedly used, and the technical problems that an existing needle assisting device cannot be repeatedly used, so that resources are wasted, and the use cost is high are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical devices, and in particular, relates to a push-type insulin injection aid based on sound detection and a working method thereof. Background Art

[0002] Most of the existing insulin injection aids use side button operation, which is not convenient in use. Since the side button design requires the user to operate at a specific angle and position, it is easy to cause instability during operation, which in turn affects the working effect. Some insulin injection aids use button-type injection aids, but these products are usually disposable and cannot be reused. This not only causes a waste of resources, but also significantly increases the user's cost of use. Therefore, there is a need for a reusable and effective push-type insulin injection aid based on sound detection and a working method. Summary of the invention

[0003] The purpose of the present invention is to provide a sound detection-based push-type insulin injection aid and a working method which is reusable and has good injection aid effect.

[0004] To achieve the above object, the present invention adopts the following technical solution:

[0005] A push-type insulin needle aid based on sound detection comprises an outer shell, a middle shell, an inner shell, a pull ring, a button, a first spring, a second spring, a sliding lock, a third spring, a sound detection module, a sphere, and a cylindrical tube; the outer shell, the inner shell, and the middle shell are all arranged vertically, the inner shell and the middle shell are connected to each other through a first buckle, a pull ring is arranged above the inner shell, a button connected to the inner shell is arranged in the middle of the pull ring, a first spring is sleeved on the button, and the downward movement of the button pushes the bottom of the inner shell to move downward, a second buckle for connecting an infusion line is arranged at the bottom of the inner shell, a second spring is sleeved on the middle shell, an inwardly extending extending support portion is arranged at the top of the inner wall of the outer shell, a step portion is arranged at the bottom of the middle shell, the top end of the second spring is connected to the lower surface of the extending support portion, the bottom end of the second spring is connected to the upper surface of the step portion, and the step portion is provided at the bottom of the middle shell. At least two sliding locks are extended outward from the upper part, the sliding locks are arranged horizontally, a third spring is sleeved on the sliding locks, an inclined portion is arranged on the outer side of the sliding locks, and a transverse guide rib extending inward is arranged in the middle part of the inner wall of the outer shell. When the middle shell moves upward, the transverse guide rib moves downward relative to the middle shell along the inclined portion, and the third spring is compressed until the lower surface of the sliding lock moves above the upper surface of the transverse guide rib, and the third spring returns to its original state, so that the transverse guide rib supports the sliding lock. At this time, the second spring is in a compressed state. When the inside of the outer shell is pressed hard to move the outer side of the sliding lock to the inner surface of the transverse guide rib, the second spring returns to its original state, so that the middle shell moves downward to its original state; the cylindrical tube is arranged on the inner shell, the sphere is installed in the cylindrical tube, and the sound detection module is arranged on the inner shell and is located on one side of the cylindrical tube.

[0006] The sound detection module includes a microphone, a chip, a button battery, and a speaker;

[0007] The sound detection module and the cylindrical tube are both bonded to the inner shell by colloid;

[0008] A plurality of slide grooves are arranged on the inner wall of the outer shell, and a plurality of slide bars sliding in the slide grooves are arranged on the outer wall of the middle shell, and the slide grooves correspond to the slide bars one by one.

[0009] A pressure switch is arranged on the button, a ring-shaped indicator light is arranged on the circumference of the pull ring, and the microphone, chip, ring-shaped indicator light and speaker are all connected to the button battery through wires.

[0010] The working method of the push-type insulin needle aid based on sound detection is implemented based on the above-mentioned push-type insulin needle aid based on sound detection, and includes the following steps:

[0011] (1) Install the infusion line on the second buckle and remove the auxiliary sticker on the infusion line;

[0012] (2) Pull the pull ring upward until the pull ring drives the inner shell and then drives the middle shell to move upward so that the slide lock is placed above the transverse guide rib and supported by the transverse guide rib. At this time, the second spring is in a compressed state;

[0013] (3) Align the bottom of the outer shell with the use part, apply pressure to the outer shell, and move the slide lock inward to disengage from the support of the transverse guide rib. At this time, the second spring returns to its original state. Under the action of the second spring, the inner shell drives the infusion pipeline to complete the injection. Press the button downward to compress the third spring, so that the infusion pipeline is disengaged from the clamping of the second buckle, and the injection is completed.

[0014] (4) When the button is pressed down, the ball moves in the cylinder and makes a sound. The sound detection module detects and determines whether the injection is successful.

[0015] The specific judgment process of step (4) is as follows:

[0016] ①Use wavelet entropy as a feature extraction tool to analyze the timbre, sound intensity and sound frequency of the sound signal;

[0017] Sound intensity formula:

[0018] I=P / A

[0019] Where I represents the sound intensity, P represents the power of the sound passing through, and A represents the area where the sound propagates;

[0020] Sound frequency formula:

[0021] f=1 / T

[0022] Where f represents frequency and T represents time;

[0023] ② Decompose the original signal through wavelet transform and decompose it into different frequency levels. Wavelet transform can perform multi-scale analysis of the signal in the time-frequency domain and capture the changing characteristics of each scale component in the signal;

[0024] ③ At each decomposition level, the successful injection has a thick timbre, a sound frequency range of 250-750Hz, and a sound intensity of about 40 decibels. On the contrary, the timbre is dull and the sound frequency is lower than 250Hz, which means the injection is unsuccessful. The corresponding entropy value is calculated. In this way, wavelet entropy can accurately identify subtle changes and local features in the signal;

[0025] ④ Calculate the energy or probability distribution of the signal components at each scale;

[0026] ⑤Use the information entropy formula to calculate the wavelet entropy. The information entropy formula is usually:

[0027] W=-∑PjlogPj

[0028] Among them, Pj is the probability of the signal component accounting for the total energy at each scale;

[0029] ⑥ Using the least squares method, the extracted energy or probability distribution data is fitted into a geometric equation through experiments, where the timbre is the independent variable x 1 , sound intensity is the independent variable x 2 , the sound frequency is the independent variable x 3 , the dependent variable is the output result y, and the fitting equations are:

[0030] y 1 =a 1 +b 1 x 1 ;

[0031] y 2 =a 2 +b 2 x 2 ;

[0032] y 3 =a 3 +b 3 x 3 ;

[0033] ⑦ Model the relationship between different sound signals according to the collective equation in step ⑥, and further analyze the characteristic performance of the signals during the injection process;

[0034] ⑧According to the fitting results of the geometric equation, the output classification results are expressed in binary form, using 0 and 1 to identify different classification situations.

[0035] The specific classification of step ⑧ is as follows: if the output result is between 0 and 0.5, the final result is considered to be 0 and the output is 0; if the output result is between 0.5 and 1, the final result is considered to be 1 and the output is 1; wherein, 0 indicates that the injection is unsuccessful, the horn sounds once, and the annular indicator light displays red; 1 indicates that the injection is successful, the horn sounds twice, and the annular indicator light displays green.

[0036] The specific judgment process of step (4) is as follows:

[0037] ①Pre-process the sound signal generated by the ball colliding with the cylinder;

[0038] ② Extract key features from the preprocessed sound signal;

[0039] ③Build the model structure of recurrent neural network RNN;

[0040] ④ Use a large amount of labeled impact sound data to train the RNN model;

[0041] ⑤ Use the trained RNN model to detect and classify impact sounds in real time;

[0042] ⑥Build a convolutional neural network (CNN) model and perform feature extraction and classification on the spectrum graph, while preprocessing the impact sound signal to remove environmental noise;

[0043] ⑦Convert the preprocessed sound signal into a spectrum using short-time Fourier transform (STFT);

[0044] ⑧ Use the labeled impact sound spectrogram data to train the CNN model;

[0045] ⑨Convert the real-time impact sound signal into a spectrogram and input it into the trained CNN model.

[0046] The preprocessing in step ① includes sampling, quantization, and segmentation, so as to convert it into a digital signal suitable for model processing, and eliminate background noise through filtering technology, making the impact sound signal clearer;

[0047] The features in step ② include timbre, sound intensity and sound frequency;

[0048] The specific training method of step ④ is: continuously adjusting the model parameters through the back propagation algorithm and optimizer, so that the RNN model can accurately identify different types of impact sounds;

[0049] The specific detection and classification process in step ⑤ is as follows: the RNN model will output the corresponding classification result based on the input sound signal;

[0050] The specific processing method of step ⑥ is: by using a filter to process the sound signal, an effective impact sound signal is extracted.

[0051] The specific judgment process of step (4) is as follows:

[0052] ① The timbre, sound intensity and sound frequency of the sound signal generated by the ball hitting the cylinder are used to establish three least squares models: 1 =a 1 +b 1 x 1 ,y 2 =a 2 +b 2 x 2 ,y 3 =a 3 +b 3 x 3 , and fit the relationship between input data and output results;

[0053] ②Use the sound signal generated by the ball hitting the cylinder to establish two deep learning models, the recurrent neural network RNN ​​output result y4 and the convolutional neural network CNN output result y 5 ;

[0054] ③ The input sound data is passed into three least squares models and two deep learning models respectively, the output of each model is obtained, and the results of the least squares model and the deep learning model are binary-encoded, using "1" and "0" to indicate whether the result meets the conditions;

[0055] ④Combining 5 models 1 ,y 2 ,y 3 ,y 4 ,y 5 The binary output result is as follows:

[0056] If 3 or more of the 5 model outputs are "0", the final result is judged as "0";

[0057] If 3 or more of the 5 model outputs are "1", the final result is judged as "1";

[0058] ⑤The final result is "0" and "1". 0 means that the injection is unsuccessful, the horn sounds once, and the ring indicator light turns red; 1 means that the injection is successful, the horn sounds twice, and the ring indicator light turns green.

[0059] The needle aid of the present invention is provided with a sound detection module, which can detect the sound changes during the injection process, and then determine whether the injection is successful through a corresponding algorithm, thereby achieving accurate injection and instant completion prompts, effectively improving the user experience.

[0060] The needle-assisting device of the present application hits the infusion line through a button to push the infusion line out, thereby completing the injection, which greatly improves the success rate of the injection and facilitates operation.

[0061] To sum up, the present application has a simple structure, is easy to operate and can be reused. It can not only achieve accurate injection and instant completion prompts through the feedback mechanism of the sound detection module to improve the user experience, but also reduce resource waste and reduce usage costs through reuse, effectively solving the technical problems of existing needle aids that cannot be reused, resulting in resource waste and high usage costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a structural schematic diagram of the push-type insulin injection aid of the present invention.

[0063] Figure 2 It is a schematic diagram of the pull ring of the needle assisting device of the present invention pulling the infusion pipeline upward.

[0064] Figure 3 It is a schematic diagram of the needle assisting device pull ring of the present invention pulling the infusion pipeline upward to complete the assembly.

[0065] Figure 4 It is a schematic diagram of the structure of the cylindrical barrel and the sphere of the present invention.

[0066] Figure 5 It is a structural schematic diagram of the sound detection module of the present invention.

[0067] Figure 6 It is a structural schematic diagram of the pressure switch and the ring indicator light of the present invention. DETAILED DESCRIPTION

[0068] The embodiments of the present invention are further described below with reference to the accompanying drawings.

[0069] Embodiment 1

[0070] like Figure 1-6As shown, a push-type insulin needle aid based on sound detection comprises an outer shell 1, a middle shell 2, an inner shell 3, a pull ring 4, a button 5, a first spring 6, a second spring 7, a slide lock 8, a third spring 9, a sound detection module 10, a ball 11, and a cylindrical barrel 12; the outer shell 1, the inner shell 3 and the middle shell 2 are all arranged vertically, the inner shell 3 and the middle shell 2 are connected by a first buckle, a pull ring 4 is arranged above the inner shell 3, a button 5 connected to the inner shell 3 is arranged in the middle of the pull ring 4, a first spring 6 is sleeved on the button 5, and the button 5 moves downward to push the bottom of the inner shell 3 to move downward, a second buckle for connecting the infusion pipeline 23 is arranged at the bottom of the inner shell 3, a second spring 7 is sleeved on the middle shell 2, an inwardly extending extension support portion 13 is arranged on the top of the inner wall of the outer shell 1, a step portion 14 is arranged at the bottom of the middle shell 2, the top end of the second spring 7 is connected to the lower surface of the extension support portion 13, and the bottom end of the second spring 7 is connected to the upper surface of the step portion 14, At least two slide locks 8 are extended outward on the step portion 14, and the slide locks 8 are arranged horizontally. A third spring 9 is sleeved on the slide locks 8. An inclined portion 15 is arranged on the outer side of the slide locks 8. A transverse guide rib 22 extending inward is arranged in the middle part of the inner wall of the outer shell 1. When the middle shell 2 moves upward, the transverse guide rib 22 moves downward relative to the middle shell 2 along the inclined portion 15, and the third spring 9 is compressed until the lower surface of the slide lock 8 moves above the upper surface of the transverse guide rib 22. The third spring 9 returns to its original state, so that the transverse guide rib 22 supports the slide lock 8. At this time, the second spring 7 is in a compressed state. When the inside of the outer shell 1 is pressed hard to move the outer side of the slide lock 8 to the inner surface of the transverse guide rib 22, the second spring 7 returns to its original state, so that the middle shell 2 moves downward to its original state; the cylindrical tube 12 is arranged on the inner shell 3, and the ball 11 is installed in the cylindrical tube 12. The sound detection module 10 is arranged on the inner shell 3 and is located on one side of the cylindrical tube 12.

[0071] The sound detection module 10 includes a microphone 16, a chip 17, a button battery 18, and a speaker 19;

[0072] The sound detection module 10 and the cylindrical tube 12 are both bonded to the inner shell 3 by colloid;

[0073] A plurality of slide grooves are arranged on the inner wall of the outer shell 1, and a plurality of slide bars sliding in the slide grooves are arranged on the outer wall of the middle shell 2, and the slide grooves correspond to the slide bars one by one.

[0074] The button 5 is provided with a pressure switch 20, the circumference of the pull ring 4 is provided with a ring indicator light 21, and the microphone 16, the chip 17, the ring indicator light 21, and the speaker 19 are all connected to the button battery 18 through wires. The needle assisting device of the present application is provided with two horizontal bars extending from the top of the inner shell 3 to push out the infusion line 23 by pressing the button 5, thereby completing the injection, greatly improving the success rate of the injection and facilitating the operation.

[0075] Embodiment 2

[0076] The working method of the push-type insulin injection aid based on sound detection is implemented based on the push-type insulin injection aid based on sound detection described in the first embodiment, and includes the following steps:

[0077] (1) Install the infusion line 23 on the second buckle and remove the auxiliary sticker on the infusion line 23;

[0078] (2) Pull the pull ring 4 upward until the pull ring 4 drives the inner shell 3 and then drives the middle shell 2 to move upward so that the slide lock 8 is placed above the transverse guide rib 22 and is supported by the transverse guide rib 22. At this time, the second spring 7 is in a compressed state;

[0079] (3) Align the bottom of the outer shell 1 with the use part, apply pressure to the outer shell 1, and move the slide lock 8 inward to disengage from the support of the transverse guide rib 22. At this time, the second spring 7 returns to its original state. Under the action of the second spring 7, the inner shell 3 drives the infusion pipeline 23 to complete the injection. Press the button 5 downward to compress the third spring 9, so that the infusion pipeline 23 is disengaged from the clamping of the second buckle, and the injection is completed.

[0080] (4) After the button 5 is quickly pressed twice, the pressure switch 20 is turned on. The pressure switch 20 is set in the middle of the button 5, and the pressure switch 20 is connected to the button battery 18 through a wire; the button battery 18 starts to supply power to the chip 17. If there is no pressure for more than one minute, the pressure switch 20 is turned off, and the button battery 18 stops supplying power to the chip 17. During operation, the vibration generated by the inner shell 3 hitting the user object is transmitted to the cylindrical tube 12 at the top of the inner shell 3, and the cylindrical tube 12 and the inner shell 3 are colloidally connected; the ball 11 is inside the cylindrical tube 12, and the vibration causes the ball 11 to colloidal ...

[0081] The specific judgment process of step (4) is as follows:

[0082] ①Use wavelet entropy as a feature extraction tool to analyze the timbre, sound intensity and sound frequency of the sound signal;

[0083] Sound intensity formula:

[0084] I=P / A

[0085] Where I represents the sound intensity, P represents the power of the sound passing through, and A represents the area where the sound propagates;

[0086] Sound frequency formula:

[0087] f=1 / T

[0088] Where f represents frequency and T represents time;

[0089] ② Decompose the original signal through wavelet transform and decompose it into different frequency levels. Wavelet transform can perform multi-scale analysis of the signal in the time-frequency domain and capture the changing characteristics of each scale component in the signal;

[0090] ③ At each decomposition level, the successful injection has a thick timbre, a sound frequency range of 250-750Hz, and a sound intensity of about 40 decibels. On the contrary, the timbre is dull and the sound frequency is lower than 250Hz, which means the injection is unsuccessful. The corresponding entropy value is calculated. In this way, wavelet entropy can accurately identify subtle changes and local features in the signal;

[0091] ④ Calculate the energy or probability distribution of the signal components at each scale;

[0092] ⑤Use the information entropy formula to calculate the wavelet entropy. The information entropy formula is usually:

[0093] W=-∑PjlogPj

[0094] Among them, Pj is the probability of the signal component accounting for the total energy at each scale;

[0095] ⑥Use the least squares method to fit the extracted energy or probability distribution data into a geometric equation through experiments. The timbre is the independent variable x 1 , sound intensity is the independent variable x 2 , the sound frequency is the independent variable x 3 , the dependent variable is the output result y, and the fitting equations are:

[0096] y 1 =a 1 +b 1 x 1 ;

[0097] y 2 =a 2 +b 2 x 2 ;

[0098] y 3 =a 3 +b 3 x 3 ;

[0099] ⑦ Model the relationship between different sound signals according to the collective equation in step ⑥, and further analyze the characteristic performance of the signals during the injection process;

[0100] ⑧According to the fitting results of the geometric equation, the output classification results are expressed in binary form, using 0 and 1 to identify different classification situations.

[0101] The specific classification of step ⑧ is as follows: if the output result is between 0 and 0.5, the final result is considered to be 0, and the output is 0; if the output result is between 0.5 and 1, the final result is considered to be 1, and the output is 1; wherein, 0 indicates that the injection is unsuccessful, the horn 19 sounds once, and the annular indicator light 21 displays red; 1 indicates that the injection is successful, the horn 19 sounds twice, and the annular indicator light 21 displays green. A sound detection module 10 is provided in the needle aid of the present invention, and the sound changes during the injection process can be detected by the sound detection module 10, and then the corresponding algorithm is used to determine whether the injection is successful, so as to achieve accurate injection and immediate completion prompts, and effectively improve the user experience. The present application has a simple structure, is easy to operate, and can be reused. Not only can accurate injection and immediate completion prompts be achieved through the feedback mechanism of the sound detection module 10, thereby improving the user experience, but also the waste of resources and the cost of use can be reduced through reuse, effectively solving the technical problem that the existing needle aid cannot be reused, resulting in waste of resources and high cost of use.

[0102] Embodiment 3

[0103] The difference from the second embodiment is that the specific judgment process of step (4) is as follows: ① pre-processing the sound signal generated by the ball colliding with the cylinder,

[0104] ② Extract key features from the pre-processed sound signal. These features include timbre, sound intensity, and sound frequency;

[0105] ③Build the model structure of recurrent neural network (RNN). Since RNN has the ability to process sequence data, it can capture the dynamic characteristics of sound signals changing over time. The input of the model is the sound signal characteristics (timbre, sound intensity and sound frequency), and the output is the recognition result of the impact sound;

[0106] ④ Use a large amount of labeled impact sound data to train the RNN model. Use the back propagation algorithm and optimizer to continuously adjust the model parameters so that it can accurately identify different types of impact sounds;

[0107] ⑤ The trained RNN model is used in actual applications to detect and classify impact sounds in real time. The model will output the corresponding classification results based on the input sound signal, such as "0" and "1". 0 means that the injection is unsuccessful, the horn 19 sounds once, and the ring indicator 21 displays red; 1 means that the injection is successful, the horn 19 sounds twice, and the ring indicator 21 displays green;

[0108] ⑥Build a convolutional neural network (CNN) model to extract features and classify the spectrum. Preprocess the impact sound signal, remove environmental noise and perform standardization. Use filters to process the sound signal and extract the effective impact sound signal;

[0109] ⑦Convert the preprocessed sound signal into a spectrogram using short-time Fourier transform (STFT). The spectrogram converts the sound signal from the time domain to the frequency domain, making it easier for CNN to extract features;

[0110] ⑧ Use the labeled impact sound spectrogram data to train the CNN model. During the training process, the optimizer adjusts the model parameters so that it can accurately classify and identify different types of impact sound features (timbre, sound intensity and sound frequency);

[0111] ⑨Convert the real-time acquired impact sound signal into a spectrum graph and input it into the trained CNN model. The model outputs the classification results of the impact sound as "0" and "1". 0 means that the injection is unsuccessful, the horn 19 sounds once, and the ring indicator 21 displays red; 1 means that the injection is successful, the horn 19 sounds twice, and the ring indicator 21 displays green.

[0112] Embodiment 4

[0113] The difference from the second and third embodiments is that the specific judgment process of step (4) is:

[0114] ① The timbre, sound intensity and sound frequency of the sound signal generated by the ball 11 hitting the cylinder 12 are used to establish three least squares

[0115] Legal model: y 1 =a 1 +b 1 x 1 ,y 2 =a 2 +b 2 x 2 ,y 3 =a 3 +b 3 x 3 , and fit the relationship between input data and output results;

[0116] ②Use the sound signal generated by the ball hitting the cylinder to establish two deep learning models, the recurrent neural network RNN ​​output result y4 and the convolutional neural network CNN output result y 5 ;

[0117] ③ The input sound data is passed into three least squares models and two deep learning models respectively, the output of each model is obtained, and the results of the least squares model and the deep learning model are binary-encoded, using "1" and "0" to indicate whether the result meets the conditions;

[0118] ④Combining 5 models 1 ,y 2 ,y 3 ,y 4 ,y 5 The binary output result is as follows:

[0119] If 3 or more of the 5 model outputs are "0", the final result is judged as "0";

[0120] If 3 or more of the 5 model outputs are "1", the final result is judged as "1";

[0121] ⑤ The final result is "0" and "1", 0 means that the injection is unsuccessful, the horn 19 sounds once, and the ring indicator light 21 displays red; 1 means that the injection is successful, the horn 19 sounds twice, and the ring indicator light 21 displays green.

[0122] The above embodiments are only used to illustrate rather than limit the technical solutions of the present invention. Although the present invention is described in detail with reference to the above embodiments, those skilled in the art should understand that the present invention can still be modified or replaced by equivalents. Any modification or partial replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A push-type insulin injection device based on sound detection, characterized in that: It includes an outer shell, a middle shell, an inner shell, a pull ring, a button, a first spring, a second spring, a slide lock, a third spring, a sound detection module, a sphere, and a cylinder; the outer shell, the inner shell and the middle shell are all arranged vertically, the inner shell and the middle shell are connected by a first buckle, a pull ring is arranged above the inner shell, a button connected to the inner shell is arranged in the middle of the pull ring, a first spring is sleeved on the button, the button moves downward to push the bottom of the inner shell to move downward, a second buckle for connecting an infusion line is arranged at the bottom of the inner shell, a second spring is sleeved on the middle shell, an extending support portion extending inward is arranged on the top of the inner wall of the outer shell, a step portion is arranged at the bottom of the middle shell, the top end of the second spring is connected to the lower surface of the extending support portion, the bottom end of the second spring is connected to the upper surface of the step portion, and a spring is arranged on the step portion to extend outward There are at least two sliding locks, which are arranged horizontally. A third spring is sleeved on the sliding lock. An inclined portion is arranged on the outer side of the sliding lock. An inwardly extending transverse guide rib is arranged in the middle of the inner wall of the outer shell. When the middle shell moves upward, the transverse guide rib moves downward relative to the middle shell along the inclined portion, and the third spring is compressed until the lower surface of the sliding lock moves above the upper surface of the transverse guide rib. The third spring returns to its original state, so that the transverse guide rib supports the sliding lock. At this time, the second spring is in a compressed state. When the inner side of the outer shell is pressed hard to move the outer side of the sliding lock to the inner surface of the transverse guide rib, the second spring returns to its original state, so that the middle shell moves downward to its original state. The cylindrical tube is arranged on the inner shell, and the sphere is installed in the cylindrical tube. The sound detection module is arranged on the inner shell and is located on one side of the cylindrical tube.

2. The sound detection-based push-type insulin injection device according to claim 1, characterized in that: The sound detection module includes a microphone, a chip, a button battery, and a speaker; The sound detection module and the cylindrical tube are both bonded to the inner shell by colloid; A plurality of slide grooves are arranged on the inner wall of the outer shell, and a plurality of slide bars sliding in the slide grooves are arranged on the outer wall of the middle shell, and the slide grooves correspond to the slide bars one by one.

3. The sound detection-based push-type insulin injection device according to claim 2, characterized in that: A pressure switch is arranged on the button, a ring-shaped indicator light is arranged on the circumference of the pull ring, and the microphone, chip, ring-shaped indicator light and speaker are all connected to the button battery through wires.

4. A method for operating a push-type insulin injection device based on sound detection, which is implemented based on the push-type insulin injection device based on sound detection as claimed in any one of claims 1 to 3, characterized in that: The steps include: (1) Install the infusion line on the second buckle and remove the auxiliary sticker on the infusion line; (2) Pull the pull ring upward until the pull ring drives the inner shell and then drives the middle shell to move upward so that the slide lock is placed above the transverse guide rib and supported by the transverse guide rib. At this time, the second spring is in a compressed state; (3) Align the bottom of the outer shell with the use part, apply pressure to the outer shell, and move the slide lock inward to disengage from the support of the transverse guide rib. At this time, the second spring returns to its original state. Under the action of the second spring, the inner shell drives the infusion pipeline to complete the injection. Press the button downward to compress the third spring, so that the infusion pipeline is disengaged from the clamping of the second buckle, and the injection is completed. (4) When the button is pressed down, the ball moves in the cylinder and makes a sound. The sound detection module detects and determines whether the injection is successful.

5. The working method of the push-type insulin injection aid based on sound detection according to claim 4 is characterized in that: The specific judgment process of step (4) is as follows: ①Use wavelet entropy as a feature extraction tool to analyze the timbre, sound intensity and sound frequency of the sound signal; Sound intensity formula: I=P / A Where I represents the sound intensity, P represents the power of the sound passing through, and A represents the area where the sound propagates; Sound frequency formula: f=1 / T Where f represents frequency and T represents time; ② Decompose the original signal through wavelet transform and decompose it into different frequency levels. Wavelet transform can perform multi-scale analysis of the signal in the time-frequency domain and capture the changing characteristics of each scale component in the signal; ③ At each decomposition level, the successful injection has a thick timbre, a sound frequency range of 250-750Hz, and a sound intensity of about 40 decibels. On the contrary, the timbre is dull and the sound frequency is lower than 250Hz, which means the injection is unsuccessful. The corresponding entropy value is calculated. In this way, wavelet entropy can accurately identify subtle changes and local features in the signal; ④ Calculate the energy or probability distribution of the signal components at each scale; ⑤Use the information entropy formula to calculate the wavelet entropy. The information entropy formula is usually: W=-∑PjlogPj Among them, Pj is the probability of the signal component accounting for the total energy at each scale; ⑥ Using the least squares method, the extracted energy or probability distribution data is fitted into a geometric equation through experiments, where the timbre is the independent variable x1, the sound intensity is the independent variable x2, the sound frequency is the independent variable x3, and the dependent variable is the output result y. The fitting equations are: y1=a1+b1x1; y2=a2+b2x2; y3=a3+b3x3; ⑦ Model the relationship between different sound signals according to the collective equation in step ⑥, and further analyze the characteristic performance of the signals during the injection process; ⑧According to the fitting results of the geometric equation, the output classification results are expressed in binary form, using 0 and 1 to identify different classification situations.

6. The working method of the push-type insulin injection aid based on sound detection according to claim 5 is characterized in that: The specific classification of step ⑧ is as follows: if the output result is between 0 and 0.5, the final result is considered to be 0 and the output is 0; if the output result is between 0.5 and 1, the final result is considered to be 1 and the output is 1; wherein, 0 indicates that the injection is unsuccessful, the horn sounds once, and the annular indicator light displays red; 1 indicates that the injection is successful, the horn sounds twice, and the annular indicator light displays green.

7. The working method of the push-type insulin injection aid based on sound detection according to claim 4 is characterized in that: The specific judgment process of step (4) is as follows: ①Pre-process the sound signal generated by the ball colliding with the cylinder; ② Extract key features from the preprocessed sound signal; ③Build the model structure of recurrent neural network RNN; ④ Use a large amount of labeled impact sound data to train the RNN model; ⑤ Use the trained RNN model to detect and classify impact sounds in real time; ⑥Build a convolutional neural network (CNN) model and perform feature extraction and classification on the spectrum graph, while preprocessing the impact sound signal to remove environmental noise; ⑦Convert the preprocessed sound signal into a spectrum using short-time Fourier transform (STFT); ⑧ Use the labeled impact sound spectrogram data to train the CNN model; ⑨Convert the real-time impact sound signal into a spectrogram and input it into the trained CNN model.

8. The working method of the push-type insulin injection aid based on sound detection according to claim 7 is characterized in that: The preprocessing in step ① includes sampling, quantization, and segmentation, so as to convert it into a digital signal suitable for model processing, and eliminate background noise through filtering technology, making the impact sound signal clearer; The features in step ② include timbre, sound intensity and sound frequency; The specific training method of step ④ is: continuously adjusting the model parameters through the back propagation algorithm and optimizer, so that the RNN model can accurately identify different types of impact sounds; The specific detection and classification process in step ⑤ is as follows: the RNN model will output the corresponding classification result based on the input sound signal; The specific processing method of step ⑥ is: by using a filter to process the sound signal, an effective impact sound signal is extracted.

9. The working method of the push-type insulin injection aid based on sound detection according to claim 4, characterized in that: The specific judgment process of step (4) is as follows: ① The timbre, sound intensity and sound frequency of the sound signal generated by the ball colliding with the cylinder are used to establish three least squares models: y1=a1+b1x1, y2=a2+b2x2, y3=a3+b3x3, and fit the relationship between the input data and the output results; ②Use the sound signal generated by the ball hitting the cylinder to establish two deep learning models, the recurrent neural network RNN ​​output result y4 and the convolutional neural network CNN output result y5; ③ The input sound data is passed into three least squares models and two deep learning models respectively, the output of each model is obtained, and the results of the least squares model and the deep learning model are binary-encoded, using "1" and "0" to indicate whether the result meets the conditions; ④ Combining the binary output results of the five models y1, y2, y3, y4, and y5, the specific judgment is as follows: If 3 or more of the 5 model outputs are "0", the final result is judged as "0"; If 3 or more of the 5 model outputs are "1", the final result is judged as "1"; ⑤The final result is "0" and "1". 0 means that the injection is unsuccessful, the horn sounds once, and the ring indicator light turns red; 1 means that the injection is successful, the horn sounds twice, and the ring indicator light turns green.

Images