A snore stopping method based on an airbag pillow

Abstract

The application is a snoring stopping method based on air bag pillow, first, the neck vibration signal and the neck air bag internal pressure of the subject during sleep are collected, and the neck vibration signal segment and the neck air bag internal pressure segment during snoring are extracted; then, the neck vibration signal segment is converted into time-frequency graph, and the neck air bag internal pressure segment is converted into pressure cloud chart; the time-frequency graph and the pressure cloud chart are fused in the channel dimension to obtain the snoring image, and the model is trained; finally, the trained model is transplanted into the control system for snoring identification during sleep; according to the air bag internal pressure data and the neck vibration signal distribution form, the current head posture of the user is judged, and according to the snoring identification result and the current head posture of the user, the corresponding air bag is controlled to inflate / deflate, the head posture is adjusted, and the purpose of stopping snoring is achieved. The application identifies the snoring event based on the neck vibration signal and the neck air bag internal pressure, is not affected by environmental noise and other snorers' snoring sound, has stronger anti-interference, and avoids false detection.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent pillows, and particularly relates to a snoring stopping method based on an air bag pillow. BACKGROUND

[0002] Snoring is an unhealthy sleep state, which not only affects the sleep quality of the patient himself and others, but also affects the health of the patient himself. A large number of studies have shown that long-term snoring can induce various diseases, such as cardiovascular diseases, respiratory diseases, obstructive sleep apnea, etc. The direct cause of snoring is that the sleeping posture and head position are not appropriate, which causes the throat muscles to block the respiratory tract, making the respiratory tract not smooth. When the airflow passes through the narrow part, vortex is generated and vibration is caused, thereby producing snoring sound. Therefore, snoring monitoring is carried out, the head position during sleep is changed, the respiratory tract of the snorer is made smooth, snoring can be directly and effectively relieved, the sleep quality can be improved, and various diseases induced by snoring can be prevented.

[0003] The prior art has carried out a large amount of research on snoring monitoring. An application No. 202110861783.2 discloses an adaptive adjustment method, system and computer program of an air bag pillow. The presence or absence of snoring sound of a user in a sleep cycle is identified, and the amount of snoring sound is measured. After the snoring sound is detected, the height of the pillow is slowly adjusted through the inflation / deflation of the air bag, so as to achieve the purpose of relieving the symptoms of snoring. An invention patent No. 201911104786.0 discloses a device and method for preventing snoring based on DSP sound and image recognition technology. The environmental sound in the room is collected, and the snoring sound is quickly identified on the DSP. If there is snoring, the head position information of the user is collected by the image collection system. The head position is quickly positioned on the DSP by using the image recognition algorithm, so as to improve the efficiency of head position and sleeping posture adjustment.

[0004] The above research shows that most of the existing technologies detect snoring sound through a sound sensor. This snoring sound detection method is easily affected by environmental noise or snoring sound of other snorers, is prone to false detection, and reduces the accuracy of snoring identification.

[0005] Human respiration can cause the chest to perform regular lifting and falling movements, which can simultaneously pull the head to perform regular floating movements. Therefore, human respiration during sleep can cause the neck to vibrate. During normal sleep, respiration can cause the neck to vibrate regularly. Snoring causes the amplitude of the chest lifting and falling movement to increase, the amplitude of the neck vibration to increase, and the waveform of the neck vibration signal to present irregular and non-smooth characteristics, and the vibration frequency to obviously increase. Therefore, snoring can be identified through the neck vibration signal, and is not affected by environmental factors. SUMMARY

[0006] In view of the deficiencies of the prior art, the technical problem to be solved by the present application is to provide a snoring stopping method based on an air bag pillow.

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

[0008] A snoring stopping method based on an air bag pillow, the air bag pillow comprising a head air bag and a neck air bag, the head air bag comprising a left inner air bag, a left outer air bag, a right inner air bag and a right outer air bag, and the neck air bag being provided with an array of vibration sensors for collecting neck vibration signals; the method comprising the following contents:

[0009] I. Design a turning-over and pillow-removing test, and collect the neck vibration signals and the internal pressure of the neck air bag during the turning-over and pillow-removing;

[0010] II. Select a plurality of healthy adults prone to snoring as subjects, and require the subjects to use the air bag pillow for normal sleep, collect the neck vibration signals and the internal pressure of the neck air bag of the subjects during sleep, and extract the neck vibration signal segments and the internal pressure segments of the neck air bag during snoring from the collected neck vibration signals and the internal pressure of the neck air bag;

[0011] III. Divide the neck vibration signal segments and the internal pressure segments of the neck air bag during snoring into segments with the same data length, convert the neck vibration signal segments into time-frequency diagrams, and convert the internal pressure segments of the neck air bag into pressure cloud diagrams; fuse the time-frequency diagrams and the pressure cloud diagrams in the channel dimension to obtain a snoring image; similarly, obtain a turning-over image and a pillow-removing image; and train a convolutional neural network model using the turning-over image, the pillow-removing image and the snoring image;

[0012] IV. Transplant the trained convolutional neural network model into the control system of the air bag pillow for snoring identification during sleep; determine the current head position of the user according to the internal pressure data of the air bag and the distribution form of the neck vibration signals, if the neck vibration signals are concentratedly distributed on the left side of the array of vibration sensors, and the internal pressure of the left inner air bag and the left outer air bag increases significantly, it indicates that the head is in a left lateral position; if the neck vibration signals are concentratedly distributed on the right side of the array of vibration sensors, and the internal pressure of the right inner air bag and the right outer air bag increases significantly, it indicates that the head is in a right lateral position; if the neck vibration signals are concentratedly distributed in the middle of the array of vibration sensors, and the internal pressure of the air bags other than the neck air bag hardly changes, it indicates that the head is in a supine position;

[0013] The control system controls the corresponding air bag to inflate / deflate according to the snoring identification result and the current head position of the user, and adjusts the head position; if the identification result is snoring and the head is in a left lateral position, the left outer air bag is inflated and the left inner air bag is deflated, so that the left side of the air bag pillow is raised and the head of the user is deflected towards the middle of the air bag pillow; after a period of time, the inflation of the left outer air bag is stopped and the left inner air bag is inflated, so that the internal pressure difference between the left inner air bag and the right inner air bag is less than the internal pressure threshold; if the identification result is snoring and the head is in a right lateral position, the right outer air bag is inflated and the right inner air bag is deflated, so that the right side of the air bag pillow is raised and the head of the user is deflected towards the middle of the air bag pillow; after a period of time, the inflation of the right outer air bag is stopped and the right inner air bag is inflated, so that the internal pressure difference between the right inner air bag and the left inner air bag is less than the internal pressure threshold; if the identification result is snoring and the head is in a supine position, the left inner air bag and the left outer air bag are inflated, and the right inner air bag and the right outer air bag are deflated, so that the head is deflected to the left side; or the right inner air bag and the right outer air bag are inflated, and the left inner air bag and the left outer air bag are deflated, so that the head is deflected to the right side, and the inflation / deflation is stopped after a period of time.

[0014] Compared with the prior art, the present application has the following advantages:

[0015] 1. The present application sets up a vibration sensor array on the neck air bag, collects the neck vibration signal of the user during sleep through the vibration sensor array, and identifies the snoring event based on the neck vibration signal and the internal pressure of the neck air bag. Compared with the snoring sound identification based on the sound sensor, the present application method is not affected by environmental noise and the snoring sound of other snorers, has stronger anti-interference performance, better identification accuracy, and avoids false detection.

[0016] 2. In the data processing process, in order to accurately extract the neck vibration signal segment during snoring from the original neck vibration signal, the one-dimensional vibration signal is converted into a two-dimensional time-frequency graph, the two-dimensional time-frequency graph reflects more information, and the accuracy of determining the start and end time of snoring is improved; therefore, the neck vibration signal is fused with the internal pressure of the neck air bag, the snoring information reflected by the multi-source data is more comprehensive, and the accuracy of snoring identification is further improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 It is a structural schematic diagram of the air bag pillow;

[0018] Figure 2 It is a schematic diagram of the vibration sensor array;

[0019] Figure 3 It is a whole flowchart of the present application;

[0020] In the figure, 1 - neck air bag; 2 - right inner air bag; 3 - left inner air bag; 4 - right outer air bag; 5 - left outer air bag; 6 - vibration sensor; 7 - pillow body. DETAILED DESCRIPTION

[0021] The specific embodiments are described below with reference to the accompanying drawings, which are only used to specifically illustrate the technical solutions of the present application, and do not limit the protection scope of the present application.

[0022] As shown in Figure 1 , 2 , the air bag pillow of the present application comprises a pillow body 7 and head air bags and a neck air bag 1 inlaid on the pillow body 7, the head air bags are used to support the head of the user, comprising left inner air bags 3, right inner air bags 2, left outer air bags 5 and right outer air bags 4, the left inner air bags 3 and the right inner air bags 2 are arranged side by side, and the left outer air bags 5 and the right outer air bags 4 are symmetrically arranged outside the two inner air bags 3; the neck air bag 1 is used to support the neck of the user, and the neck air bag 1 is provided with an array composed of a plurality of vibration sensors 6, since the human body breathing will cause the chest to perform regular lifting movement, and the head will also be simultaneously pulled to perform regular floating during sleep, therefore, the vibration signal of the neck generated due to the human body breathing during sleep is collected through the vibration sensor array; each air bag is connected with the air pump through the air pipe and the electromagnetic valve, and the electromagnetic valve is connected with the control system of the air bag pillow through the relay; each air bag is externally connected with a pressure sensor for measuring the internal pressure, the five air bags are independent of each other and do not interfere with each other, and the head posture of the user is adjusted by cooperatively controlling the inflation and deflation of the five air bags, so as to achieve the purpose of snoring stopping.

[0023] An air bag pillow-based snoring stopping method, comprising the following contents:

[0024] I. Design a turning over and leaving pillow test, and collect the neck vibration signal and the internal pressure of the neck air bag during the turning over and leaving pillow;

[0025] Snoring will cause irregular vibration of the human neck, and will also cause the frequency of the neck vibration to increase, so that the waveform of the vibration signal presents irregular and non-smooth characteristics, and will also cause the frequency of the vibration signal to become large, and the turning over and leaving pillow action will also cause the neck movement, resulting in the change of the internal pressure of the neck air bag and the irregularity and frequency increase of the neck vibration signal, therefore, in order to improve the accuracy of snoring identification, it is necessary to identify the turning over and leaving pillow from the collected neck vibration signal;

[0026] A plurality of healthy adult subjects with different ages, weights, genders and body types are selected as the subjects, the subjects are required to lie flat on the air bag pillow, the pressure sensor is used to collect the internal pressure of the neck air bag, and the vibration sensor array is used to collect the neck vibration signal, if the internal pressure of the neck air bag gradually increases until the internal pressure is balanced, and the vibration sensor array collects regular vibration signals at the same time, it indicates that the head of the subject lies on the air bag pillow at this time; When each subject is tested, the initial inflation amount of each air bag needs to be adjusted according to the subjective comfort feeling by inflating / deflating the air bag, so as to ensure the use comfort; The subjects complete the turning over or leaving pillow action according to the voice prompt, and the neck vibration signal and the internal pressure of the neck air bag during turning over and leaving pillow are collected, for example, the neck vibration signal and the internal pressure of the neck air bag during the process that the head of the subject lies from flat to lateral position are collected; Each subject repeats a plurality of tests, each test includes left turning over, right turning over and leaving pillow, the neck vibration signal and the internal pressure of the neck air bag during the process that the subject completes each action are collected, and then a large number of neck vibration signals and internal pressures of neck air bags during turning over and leaving pillow are obtained through the test, which are used for training and testing of the convolutional neural network model;

[0027] II. A plurality of healthy adult subjects prone to snoring are selected as the subjects, the subjects are required to use the air bag pillow for normal sleep, and the neck vibration signal and the internal pressure of the neck air bag of the subjects during sleep are collected; The original neck vibration signal is processed to determine the start and end time of snoring, and the neck vibration signal segment and the internal pressure segment of the neck air bag during snoring are extracted from the original neck vibration signal and the internal pressure of the neck air bag, respectively, according to the start and end time;

[0028] The original neck vibration signal is filtered, and the filtered neck vibration signal is normalized; the filtering adopts a combination of median filtering and mean filtering; due to the long length of the signal, the collected neck vibration signal is segmented into neck vibration signal segments of the same length at a certain time interval; although snoring makes the neck vibration signal waveform irregular, the information reflected in the time domain is limited, and the sensitivity is low, therefore, discrete wavelet transform (DWT) is adopted to convert each neck vibration segment into a two-dimensional time-frequency graph, the two-dimensional time-frequency graph contains time domain and frequency domain information, the discrete wavelet transform expands the one-dimensional vibration signal into an energy distribution graph on the two-dimensional time-frequency plane through time-frequency analysis of the vibration signal, snoring will cause a large change in the time-frequency distribution of the vibration signal, and a higher frequency component will appear in a short time, which is obviously different from the relatively stable frequency component in the normal sleep state, and has an easily identifiable feature, therefore, the interval in which the vibration frequency increases significantly is determined in the time-frequency graph, and the start and end time of the interval is the start and end time of snoring; according to the start and end time of snoring, the neck vibration signal segment and the neck air bag internal pressure segment during snoring are extracted from the original neck vibration signal and the neck air bag internal pressure;

[0029] Due to the randomness of snoring, the subjects do not snore during the test, therefore, in order to obtain sufficient data for model training, the number of subjects needs to be large, and long-term data accumulation is also needed for collecting snoring data;

[0030] III. The convolutional neural network model is trained to make the model have the functions of identifying turning over, leaving the pillow and snoring;

[0031] The collected neck vibration signal and neck air bag internal pressure during turning over and leaving the pillow, and the extracted neck vibration signal segment and neck air bag internal pressure segment during snoring are processed to generate turning over images, leaving the pillow images and snoring images respectively;

[0032] Taking the generation of snoring images as an example, first, the neck vibration signal segment and the neck air bag internal pressure segment are filtered and normalized, and each neck vibration signal segment and neck air bag internal pressure segment is segmented into small segments of the same length; for a segment, the remaining data after segmentation may not be sufficient, if the data amount is less than or equal to one fifth of the data amount of the segmented segment, the remaining data is discarded; if the data amount is greater than one fifth and less than five fifths of the data amount of the segmented segment, the data is supplemented by copying; if the data amount is greater than five fifths of the data amount of the segmented segment, the data is supplemented by adding zeros;

[0033] Then, the segmented neck vibration signal segment is subjected to discrete wavelet transform to obtain a two-dimensional time-frequency graph of the neck vibration signal segment; the segmented neck air bag internal pressure is processed as a k*k pressure cloud chart, k 2 is the data length of the segmented neck air bag internal pressure;

[0034] Finally, the two-dimensional time-frequency graph of the neck vibration signal segment and the pressure cloud chart of the neck air bag are fused in the channel dimension to generate a two-channel grayscale chart, i.e., a snoring image;

[0035] Due to the small amount of snoring data, there is an imbalance problem in the training sample categories, which will affect the training of the model, so the mix-up method is used for data enhancement, i.e., two images of different categories, such as a snoring image and a normal sleep image, are selected to generate a mixed image through linear interpolation, specifically:

[0036]

[0037]

[0038] wherein x i , y i represent the snoring image i and its corresponding label, respectively, x k , y k represent the other category image k and its corresponding label, respectively, represent the mixed image and its corresponding label, respectively, and a is a hyperparameter satisfying Beta distribution.

[0039] The turning-over image, the off-pillow image, and the snoring image obtained above are input into a convolutional neural network model, the model is trained and optimized through back propagation, and a trained convolutional neural network model is obtained; the convolutional neural network model can be a BP neural network, a YOLO series model, etc.

[0040] Four, the trained convolutional neural network model is transplanted into the control system of the air bag pillow, before falling asleep, the user's head is placed flat on the air bag pillow, and the initial internal pressure of each air bag is collected; during sleep, the neck vibration signal and the internal pressure of the five air bags are collected in real time; the control system segments the original neck vibration signal and the neck air bag internal pressure data into segments of the same time length, processes each segment, obtains a two-dimensional time-frequency graph of the neck vibration signal segment and a pressure cloud chart of the neck air bag, and then fuses the two-dimensional time-frequency graph and the pressure cloud chart to obtain a two-channel grayscale chart, i.e., an image during sleep; the image during sleep is input into the trained convolutional neural network model for turning-over, off-pillow, and snoring recognition.

[0041] Meanwhile, the current head position of the user is determined according to the air bag internal pressure data and the distribution form of the neck vibration signal, if the neck vibration signal is concentratedly distributed on the left side of the vibration sensor array, and the internal pressure of the left inner air bag 3 and the left outer air bag 5 is obviously increased, it indicates that the head is in the left lateral position; if the neck vibration signal is concentratedly distributed on the right side of the vibration sensor array, and the internal pressure of the right inner air bag 2 and the right outer air bag 4 is obviously increased, it indicates that the head is in the right lateral position; if the neck vibration signal is concentratedly distributed in the middle of the vibration sensor array, and the internal pressure of the air bags other than the neck air bag is basically unchanged, it indicates that the head is in the supine position.

[0042] The control system controls the corresponding air bag to inflate and deflate according to the current head position of the user and the snoring recognition result of the model, so as to adjust the head position and achieve the purpose of snoring stop; if the recognition result is snoring and the head is in the left lateral position, the left outer air bag 5 is inflated, and the left inner air bag 3 is deflated, so that the left side of the air bag pillow gradually rises, and the head of the user is deflected by a certain angle towards the middle of the air bag pillow, so as to adjust the head position and achieve the purpose of snoring stop; after a period of time, the inflation of the left outer air bag 5 is stopped, and the left inner air bag 3 is slowly inflated, so that the internal pressure difference between the left inner air bag 3 and the right inner air bag 2 is less than the internal pressure threshold; if the recognition result is snoring and the head is in the right lateral position, the right outer air bag 4 is inflated, and the right inner air bag 2 is deflated, so that the right side of the air bag pillow gradually rises, and the head of the user is deflected by a certain angle towards the middle of the air bag pillow; after a period of time, the inflation of the right outer air bag 4 is stopped, and the right inner air bag 2 is slowly inflated, so that the internal pressure difference between the right inner air bag 2 and the left inner air bag 3 is less than the internal pressure threshold; if the recognition result is snoring and the head is in the supine position, the left inner air bag 2 and the left outer air bag 4 are inflated, and the right inner air bag 3 and the right outer air bag 5 are deflated, so that the head is deflected by a certain angle to the left side, and the inflation and deflation are stopped after a period of time; or the right inner air bag 3 and the right outer air bag 5 are inflated, and the left inner air bag 2 and the left outer air bag 4 are deflated, so that the head is deflected by a certain angle to the right side, and the inflation and deflation are stopped after a period of time, so as to adjust the head position and achieve the purpose of snoring stop.

[0043] The inflation and deflation rates of the above are all 0.4L / min, the inflation and deflation time is 25s, and the internal pressure threshold is 0.2kpa.

[0044] The unmentioned parts of the present application are applicable to the prior art.

Claims

1. A snore stopping method based on an airbag pillow, the airbag pillow comprising a head airbag and a neck airbag, the head airbag comprising a left inner airbag, a left outer airbag, a right inner airbag and a right outer airbag, the neck airbag being provided with an array of vibration sensors for collecting a neck vibration signal; characterized in that, The method comprises the following contents: I. Designing a turning-over and pillow-removing test, collecting neck vibration signals and neck airbag internal pressure during the turning-over and pillow-removing; II. Selecting a plurality of snoring-prone healthy adults as subjects, requiring the subjects to sleep normally using the airbag pillow, collecting the neck vibration signals and the neck airbag internal pressure of the subjects during the sleep, and extracting snoring-period neck vibration signal segments and neck airbag internal pressure segments from the collected neck vibration signals and the neck airbag internal pressure; III. Dividing the snoring-period neck vibration signal segments and the neck airbag internal pressure segments into segments with the same data length, converting the neck vibration signal segments into time-frequency diagrams, and converting the neck airbag internal pressure segments into pressure cloud diagrams; fusing the time-frequency diagrams and the pressure cloud diagrams in the channel dimension to obtain a snoring image; similarly, a turning-over image and a pillow-removing image are obtained; the turning-over image, the pillow-removing image and the snoring image are used to train a convolutional neural network model; IV. Transplanting the trained convolutional neural network model into a control system of the airbag pillow for snoring identification during the sleep; judging the current head posture of the user according to the neck airbag internal pressure data and the neck vibration signal distribution form; if the neck vibration signals are concentratedly distributed on the left side of the vibration sensor array, and the internal pressures of the left inner airbag and the left outer airbag increase obviously, it indicates that the head is in a left lateral position; if the neck vibration signals are concentratedly distributed on the right side of the vibration sensor array, and the internal pressures of the right inner airbag and the right outer airbag increase obviously, it indicates that the head is in a right lateral position; if the neck vibration signals are concentratedly distributed in the middle of the vibration sensor array, and the internal pressures of the airbags except the neck airbag hardly change, it indicates that the head is in a supine position; The control system controls the corresponding airbags to inflate / deflate according to the snoring identification result and the current head posture of the user, and adjusts the head posture; if the identification result is snoring and the head is in a left lateral position, the left outer airbag is inflated and the left inner airbag is deflated, so that the left side of the airbag pillow is raised and the head of the user is deflected towards the middle of the airbag pillow; after a period of time, the inflation of the left outer airbag is stopped and the left inner airbag is inflated, so that the internal pressure difference between the left inner airbag and the right inner airbag is less than an internal pressure threshold; if the identification result is snoring and the head is in a right lateral position, the right outer airbag is inflated and the right inner airbag is deflated, so that the right side of the airbag pillow is raised and the head of the user is deflected towards the middle of the airbag pillow; after a period of time, the inflation of the right outer airbag is stopped and the right inner airbag is inflated, so that the internal pressure difference between the right inner airbag and the left inner airbag is less than the internal pressure threshold; if the identification result is snoring and the head is in a supine position, the left inner airbag and the left outer airbag are inflated and the right inner airbag and the right outer airbag are deflated, so that the head is deflected to the left side; or the right inner airbag and the right outer airbag are inflated and the left inner airbag and the left outer airbag are deflated, so that the head is deflected to the right side, and the inflation / deflation is stopped after a period of time.

2. The airbag pillow based snore stopping method of claim 1, wherein, The process of extracting the snoring-period neck vibration signal segments and the neck airbag internal pressure segments is as follows: The collected original neck vibration signal is filtered and normalized, and is segmented into neck vibration signal segments with the same length; the neck vibration signal segments are converted into time-frequency diagrams through discrete wavelet transform, and an interval with obviously increased vibration frequency is determined in the time-frequency diagram, and the start and end time of the interval are the start and end time of snoring; according to the start and end time of snoring, the neck vibration signal segment and the neck airbag internal pressure segment during snoring are extracted from the original neck vibration signal and the neck airbag internal pressure.

3. The airbag pillow based snore stopping method of claim 1, wherein, During the adjustment of the head posture, the inflation and deflation rates of the airbag are both 0.4 L / min, the inflation and deflation times are both 25 s, and the internal pressure threshold is 0.2 kpa.

Images