Air bag pillow integrating snore monitoring, head posture recognition and snore stopping functions

Abstract

The multifunctional air bag pillow integrating snore monitoring and head posture recognition comprises a pillow shell, a neck air bag, a left head air bag, a right head air bag, a left auxiliary air bag, a right auxiliary air bag and a control box, and is characterized in that four microphones and eight flexible pressure sensors are fixed outside the pillow; a control circuit, an air pipe, an inflation pump, an outer pressure sensing module, an inner pressure sensing module, a sound sensing module and eight electromagnetic valves are arranged in the control box. The snore stopping method comprises the following steps: a) collecting signals; b) sound source positioning and snore recognition; c) head posture recognition; and d) performing snore stopping intervention and awakening. According to the multifunctional air bag pillow, by locating snore, discharging non-snore, judging the head posture of a user and starting a strategy of stopping snore by lying flat or lying on the side, the air passage blockage of the air bag pillow user is improved to the maximum extent, smooth breathing is promoted, and it is guaranteed that the user has high sleep quality to the maximum extent.

Description

Technical Field

[0001] This invention relates to an airbag pillow, and more specifically, to a multifunctional airbag pillow that integrates snoring monitoring and head posture recognition. Background Technology

[0002] With increasing health awareness, sleep quality has gradually become a focus of public attention. Snoring, as a common sleep disorder, not only affects the sleep quality of the individual but may also lead to hypoxia, frequent awakenings, and even be associated with obstructive sleep apnea syndrome (OSAS), increasing the risk of cardiovascular disease. Furthermore, improper sleeping posture can also cause neck discomfort, further affecting sleep.

[0003] At present, snoring recognition technology is mainly based on audio signal processing, which has the following problems: (1) It is easily affected by environmental noise or bed partners, resulting in a high misjudgment rate; (2) Data sets are scarce, and the model has poor generalization ability; (3) There is a lack of personalized intervention strategies for different sleeping positions.

[0004] In terms of head posture recognition, existing technologies mostly employ cameras or flexible pressure sensors, which suffer from problems such as privacy leaks, low recognition accuracy, and coarse posture classification. Meanwhile, commercially available smart pillows have simple structures and fragmented functions, making it difficult to achieve integrated control of snoring monitoring and posture recognition, and lacking an integrated, low-noise, and highly comfortable solution. This invention proposes a multifunctional airbag pillow that integrates snoring monitoring and head posture recognition to address the problems of large recognition errors, limited intervention methods, and poor user experience in existing technologies. Summary of the Invention

[0005] In order to overcome the shortcomings of the above-mentioned technical problems, the present invention provides a multifunctional airbag pillow that integrates snoring monitoring and head posture recognition.

[0006] The present invention relates to a multifunctional airbag pillow integrating snoring monitoring and head posture recognition, comprising a pillow shell and a neck airbag, a left head airbag, a right head airbag, a left auxiliary airbag, a right auxiliary airbag, and a control box disposed within the pillow shell. The neck airbag supports the user's neck, while the left and right head airbags support the head. The left and right auxiliary airbags are located on the outer sides of the left and right head airbags, respectively. The control box is separately disposed from the pillow shell. The invention is characterized by having microphones for collecting sound signals fixed at each of the four corners of the pillow shell, and eight flexible pressure sensors evenly disposed along the length of the pillow shell to measure the pressure signals on the user's head.

[0007] The control box contains a control circuit, air tube, air pump, external pressure sensor module, internal pressure sensor module, sound sensor module, and eight solenoid valves. The air pump, solenoid valves, and air tube constitute four inflation / deflation circuits. The neck airbag, left head airbag, and right head airbag each use one inflation / deflation circuit, while the left and right auxiliary airbags share one inflation / deflation circuit. Pressure sensors are installed on the air inlets and outlets of the neck airbag, left and right head airbags, and left and right auxiliary airbags. The microphone, flexible pressure sensor, and air pressure sensor are connected to the sound sensor module, external pressure sensor module, and internal pressure sensor module, respectively.

[0008] The control circuit measures the sound signal, head pressure signal, and airbag pressure signal through the sound sensing module, external pressure sensing module, and internal pressure sensing module, respectively. Based on the collected sound signal, the control circuit determines whether the sleeper using the airbag pillow is snoring. Based on the collected head pressure signal and airbag pressure signal, the control circuit determines whether the patient's current sleeping position is supine or lateral. Based on the patient's current sleeping position, the control circuit uses an air pump to intermittently inflate each airbag to activate supine or lateral anti-snoring.

[0009] The present invention relates to a multifunctional airbag pillow integrating snoring monitoring and head posture recognition. The neck airbag, left head airbag, right head airbag, left auxiliary airbag, and right auxiliary airbag are in the same plane and form an airbag group layer. The pillow shell above the airbag group layer consists of a soft sponge layer, an array pressure sensor layer, and a hard sponge isolation layer from the outside to the inside. The array pressure sensor layer is formed by a flexible pad and eight flexible pressure sensors fixed on the flexible pad. The pillow shell below the airbag group layer consists of a bottom sponge layer and an acrylic support plate layer from the outside to the inside.

[0010] The multifunctional airbag pillow of the present invention integrates snoring monitoring and head posture recognition. The air outlet of the air pump is connected to the air inlet and outlet of the neck airbag, left head airbag, right head airbag, left auxiliary airbag and right auxiliary airbag through the air tube and four solenoid valves respectively. The air inlet and outlet of the neck airbag, left head airbag, right head airbag, left auxiliary airbag and right auxiliary airbag are connected to the outside through the four solenoid valves.

[0011] The anti-snoring method of the multifunctional airbag pillow integrating snoring monitoring and head posture recognition of the present invention is characterized by the following steps:

[0012] a) Signal acquisition; The control circuit acquires external sound source signals when the user is sleeping through four microphones set on the pillow shell, acquires head pressure signals applied to the flexible pressure sensor by the external pressure sensor module, and acquires internal pressure signals of each airbag by the internal pressure sensor module.

[0013] b) Sound source localization and snoring recognition; based on the geometric spatial relationship of the four microphones, the distance of the sound source from the set origin, as well as the pitch and azimuth angles of the sound source, are calculated to achieve spatial localization of the sound source, so as to distinguish whether the sound source is emitted by the user or the bed partner.

[0014] After determining that the sound source is the user by locating the sound source, the original audio is first subjected to wavelet threshold noise reduction, and the Mel frequency spectrum is extracted as the model input. Then, the MobileNet V2 model based on transfer learning is selected for snoring recognition to achieve binary classification of whether the current sound source is snoring or not.

[0015] c) Head posture recognition; First, head pressure signal data collected by a flexible pressure sensor and airbag pressure signal data collected by a pneumatic pressure sensor are fused to construct a 13-dimensional feature vector containing center of gravity shift; then, a one-dimensional convolutional neural network combined with an SE attention mechanism is used to achieve real-time recognition of six head postures, including no one, lying flat, left side lying, right side lying, left head turning, and right head turning; at the same time, a side-lying rolling judgment mechanism is designed to avoid misjudgment caused by initial value refresh during posture switching.

[0016] d) Anti-snoring intervention and wake-up: Based on the detected head posture, the system automatically selects either a supine or side-lying anti-snoring strategy. By controlling the inflation and deflation of different airbags, it guides the head to turn slightly, improving airway patency. If snoring continues to exceed the set threshold, the system will activate the active wake-up function, using the alternating inflation and deflation of the left and right airbags to push the head to move, achieving a gentle wake-up. A 5-minute intervention cooling-off period is set to avoid frequent interventions affecting sleep quality.

[0017] The anti-snoring method of the multifunctional airbag pillow integrating snoring monitoring and head posture recognition of the present invention, wherein step b) of sound source localization and snoring recognition is specifically achieved through the following steps:

[0018] b-1). Establish a rectangular coordinate system; denote the four microphone positions on the pillow shell as M1, M2, M3 and M4. The four microphones on the pillow shell are located in the same plane and at the four vertices of the same rectangle. Establish an xyz coordinate system with the center of the rectangle where the four microphones are located as the origin, the line connecting the origin and M4 as the y-axis, the line connecting the origin and M3 as the x-axis, and the z-axis as the direction. Let d be the distance of M1, M2, M3 and M4 from the origin; the coordinates of the four microphones are M1(0, -d, 0), M2(-d, 0, 0), M3(d, 0, 0) and M4(0, d, 0).

[0019] b-2). Establish the geometric relationship of the sound source; let the coordinates of the sound source S be (x, y, z), and establish the geometric position relationship between the sound source S and the four microphones as shown in formula (1):

[0020]

[0021] In the formula, r, r1, r2, r3, and r4 are the distances of the sound source from the origin and the distances of the microphones M1, M2, M3, and M4, respectively.

[0022] The distance difference between the sound source and the microphone can be represented by the time delay difference. Taking microphone M1 as the reference, the following relationship between the distances of the other three microphones and the sound source is obtained:

[0023]

[0024] In the formula, τ i1 For microphone M i The time delay difference relative to microphone M1, c is the speed of sound propagation, taken as 343m / s;

[0025] b-3). Calculate the location of the sound source; combine formulas (1) and (2) to eliminate x, y, and z, and obtain the distance r of the sound source S:

[0026]

[0027] The azimuth of the sound source is obtained from spatial geometric relationships. And pitch angle θ:

[0028]

[0029] The coordinates of the sound source location are converted into polar coordinates, as shown in formula (5):

[0030]

[0031] Simplify to obtain the azimuth angle And pitch angle θ:

[0032]

[0033] b-4). Sound source identification; based on the distance r of the sound source S from the origin and the azimuth angle... The pitch angle θ determines whether the sound signal comes from the user of the airbag pillow, so as to eliminate interfering sounds from bed partners or other people;

[0034] b-5). Preprocessing of the sound source signal: First, the input signal is decomposed into three levels using the sym8 wavelet basis to obtain wavelet coefficients at different levels. Then, by selecting the number of wavelet function decomposition levels, threshold, and threshold function, an adaptive thresholding method is used to process the signal to remove noise and retain useful information. Finally, the wavelet coefficients after thresholding are reconstructed by inverse wavelet transform to obtain the denoised signal.

[0035] b-6). Snoring recognition; The noise-reduced audio was converted into a Mel cepstral plot as the input to the model. The MobileNet V2 model based on transfer learning was selected for snoring recognition to identify snoring and non-snoring sound source signals.

[0036] The anti-snoring method of the multifunctional airbag pillow integrating snoring monitoring and head posture recognition of the present invention, wherein the head posture recognition in step c) is specifically achieved through the following steps:

[0037] c-1). Pressure signal acquisition; Eight flexible pressure sensors on the pillow shell are set at equal intervals of 30mm. The pressure inside the airbags collected by the neck airbag, left head airbag, and right head airbag are marked as channel 1, channel 2, and channel 3, respectively. The pressure inside the airbags collected by the left auxiliary airbag and right auxiliary airbag is marked as channel 4.

[0038] Fourteen subjects performed each posture—lying flat, lying on their left side, lying on their right side, with their head turned to the left, and with their head turned to the right—40 times each. Head pressure signals from eight flexible pressure sensors and pressure signals from four airbags from pneumatic sensors were collected, totaling 2800 sets of data. The head center of gravity position was calculated using the head pressure signals from the eight flexible pressure sensors. The head center of gravity position was calculated by dividing the sum of the products of the flexible pressure sensor positions and their corresponding pressure values ​​by the sum of the pressure values ​​detected by all sensors.

[0039] c-2). Head position determination: When the pressure value detected by the 8 flexible pressure sensors is lower than the set threshold, it is determined to be in the position of being off the pillow, and no head posture recognition is performed; when the pressure value detected by the 8 flexible pressure sensors is higher than the set threshold, it is determined to be in the position of being on the pillow; at the same time, the side-lying turning process is also considered, that is, when the number of pressure plates changing in the data of the 8 flexible pressure sensors at the same time does not exceed 6, the head posture can be determined to be normal; if it exceeds 6, this is determined to be the turning process, and no head posture recognition is performed.

[0040] c-3). Model Training: Build a CNN model incorporating the SE attention mechanism. Use the 2800 sets of stress data collected in step c-1) to train the head posture model. At the beginning of training, the 2800 sets of sleeping posture recognition data collected above are randomly shuffled and divided into training set: test set: validation set in a ratio of 7:2:1 to ensure the representativeness and generalization ability of the model's training, validation, and testing processes. The training set consists of 1960 data points for model parameter learning; the test set consists of 560 data points for evaluating the model's performance on unseen data; the validation set consists of 280 data points. The trained head posture recognition model is then obtained.

[0041] The anti-snoring method of the multifunctional airbag pillow integrating snoring monitoring and head posture recognition of the present invention, wherein the anti-snoring intervention and wake-up described in step d) are specifically implemented through the following steps:

[0042] d-1). Determine if snoring occurs; collect audio signals from the user's head area in real time and determine whether the user is snoring according to step b); if snoring occurs, proceed to step d-2);

[0043] d-2). Head posture determination; Based on step c), determine whether the user is currently lying flat or on their side. If lying flat, proceed to step d-3); if on their side, proceed to step d-4.

[0044] d-3). Lying flat to stop snoring: First, set a low air pressure value for airbags 1, 2, and 3. Then, inflate airbags 1 and 3 for 20 seconds while deflating airbag 2 for 20 seconds to guide the head to turn to the right and raise the neck support. After completion, the system waits for 30 seconds and re-detects snoring. If snoring disappears, the airbags return to their initial state and the cycle ends.

[0045] If snoring persists, inflate the third airbag for 30 seconds to increase the head tilt angle to the right; wait another 30 seconds after execution and perform a snoring detection; if the snoring disappears, exit the cycle; if snoring persists, continue inflating the third airbag to the maximum limit and hold for 60 seconds, then exit the cycle; after the entire cycle is completed, the system will check again after 5 minutes to see if repeated anti-snoring intervention is needed; if snoring continues for a long time without significant reduction, a threshold can be set to trigger the active wake-up function to ensure the user's sleep quality;

[0046] d-4). Side-lying anti-snoring: First, set a low air pressure value for the fourth airbag, then inflate the fourth airbag for 20 seconds to slightly elevate the head and optimize airway patency; after completion, the system waits for 30 seconds and detects the snoring. If the snoring disappears, the airbag returns to its initial state and the cycle ends. If the snoring still exists, inflate the fourth airbag for another 30 seconds to increase the head elevation.

[0047] After execution, wait another 30 seconds and re-detect snoring. If snoring disappears, exit the cycle. If snoring persists, continue inflating the fourth airbag to the maximum limit and hold for 60 seconds, then exit the cycle. After the entire cycle is completed, the system will detect again after 5 minutes whether the anti-snoring intervention needs to be restarted.

[0048] The beneficial effects of this invention are as follows: The multifunctional airbag pillow integrating snoring monitoring and head posture recognition consists of a pillow shell and a neck airbag, left and right head airbags, and left and right auxiliary airbags disposed therein. Microphones for collecting sound source signals are disposed at the four corners of the pillow shell, and eight flexible pressure sensors are disposed on the pillow shell at even intervals. Each airbag inlet and outlet is provided with an air pressure sensor. Thus, the control circuit in the control box collects the sound source signal of the user's head through the microphones and determines whether it is the user's snoring signal. After the user snores, the pressure of the user's head on the pillow is obtained by collecting the eight flexible pressure sensors and the air pressure sensors at the airbag inlet and outlet to determine the user's sleeping posture. After the user snores and the head posture is detected, the corresponding airbags are intermittently inflated and deflated to activate the supine anti-snoring or side-lying anti-snoring strategy, so as to maximize the improvement of the airway obstruction of the user, promote smooth breathing, and maximize the user's sleep quality. Attached Figure Description

[0049] Figure 1 This is a front view of the multifunctional airbag pillow of the present invention;

[0050] Figure 2 This is a top view of the multifunctional airbag pillow of the present invention;

[0051] Figure 3 This is a perspective view of the multifunctional airbag pillow of the present invention;

[0052] Figure 4 This is a perspective view of the multifunctional airbag pillow of the present invention after part of the pillow shell has been removed;

[0053] Figure 5 This is a schematic diagram of the distribution structure of each airbag in the present invention;

[0054] Figure 6 This is a schematic diagram illustrating the layout of the four microphones in this invention.

[0055] Figure 7 This is a schematic diagram illustrating the layout principle of the eight flexible pressure sensors in this invention.

[0056] Figure 8 This is a schematic diagram showing the connection between the external pressure sensing module and the flexible pressure sensor in this invention.

[0057] Figure 9 This is a schematic diagram showing the division of the 5 airbags into 4 channels in this invention;

[0058] Figure 10 This is a schematic diagram of the four-way inflation / deflation circuit and the internal pressure sensing module in this invention.

[0059] Figure 11This is a coordinate diagram for determining the location of the sound source in this invention;

[0060] Figure 12 This is a schematic diagram illustrating the layer layout of the multifunctional airbag pillow of the present invention.

[0061] Figure 13 This is a flowchart of the anti-snoring strategy of the present invention.

[0062] In the image: 1 Pillow outer shell, 2 Neck airbag, 3 Left head airbag, 4 Right head airbag, 5 Left auxiliary airbag, 6 Right auxiliary airbag, 7 Microphone, 8 Flexible pad, 9 Flexible pressure sensor, 10 External pressure sensing module, 11 Inflation pump, 12 Solenoid valve, 13 Internal pressure sensing module, 14 Soft sponge layer, 15 Array pressure sensor layer, 16 Hard sponge insulation layer, 17 Airbag assembly layer, 18 Acrylic support plate layer, 19 Bottom sponge layer. Detailed Implementation

[0063] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0064] like Figures 1 to 4 The diagram shows the front view, top view, perspective view, and perspective view of the multifunctional airbag pillow of the present invention after removing part of the pillow shell. The multifunctional airbag pillow consists of a control box, a pillow shell 1, and a neck airbag 2, a left head airbag 3, a right head airbag 4, a left auxiliary airbag 5, and a right auxiliary airbag 6 disposed within the pillow shell 1. The control box and the pillow are separate components, i.e., the control box and the pillow are spaced a certain distance apart to avoid the noise generated by the air pump 11 and solenoid valve 12 in the control box from adversely affecting the user's sleep. The neck airbag 2 is positioned near the lower edge of the pillow shell 1. The left head airbag 3 and the right head airbag 4 are positioned above the neck airbag 2. The left auxiliary airbag 5 and the right auxiliary airbag 6 are respectively positioned outside the left head airbag 3 and the right head airbag 4.

[0065] like Figure 5 The diagram shows the distribution structure of the airbags in this invention. The pillow shell 1 is cylindrical, while the left head airbag 3 and right head airbag 4 are L-shaped and both are arranged in three layers. The left auxiliary airbag 5 and right auxiliary airbag 6 are rectangular. By inflating the airbags with gas at a certain pressure, the comfort of the airbag pillow can be increased, and the inflation and deflation of the corresponding airbags can also help the user stop snoring or wake up.

[0066] like Figure 6The diagram illustrates the layout of the four microphones in this invention. Microphones 7 are positioned at each of the four corners of the pillow shell 1. These four microphones are located at the four vertices of a rectangle on the same plane. The control circuit collects audio signals from the four microphones to the sound source, thereby calculating the sound source location and determining whether the sound is emitted by the user of the air-cushioned pillow. Furthermore, by analyzing the collected signals, it can also determine whether the sound is the user's snoring, thus distinguishing between snoring and non-snoring sounds.

[0067] like Figure 7 The diagram illustrates the arrangement of eight flexible pressure sensors in this invention. Eight flexible pressure sensors 9 are mounted on the pillow shell 1, and are equally spaced and fixed to a flexible pad 8. These sensors are then secured to the pillow shell 1 above the left head airbag 3 and right head airbag 4 via the flexible pad 8. The eight flexible pressure sensors 9 form a pressure sensor array. By detecting the pressure applied by the user to the pillow shell 1 using the flexible pressure sensors 9, the user's head posture can be determined accordingly.

[0068] like Figure 8 The diagram shows the connection between the external pressure sensing module and the flexible pressure sensor in this invention. Figure 9 A schematic diagram of the five airbags divided into four sections is provided in this invention. Figure 10 The schematic diagram of the four inflation / deflation circuits and the internal pressure sensing module in this invention is provided. The control box contains a control circuit, air pipes, an inflation pump 11, an external pressure sensing module 10, an internal pressure sensing module 13, an audio sensing module, and eight solenoid valves 12. The inflation pump 11, solenoid valves 12, and air pipes constitute the four inflation / deflation circuits. The neck airbag 2, the left head airbag 3, and the right head airbag 4 each use one inflation / deflation circuit, designated as circuit 1, circuit 2, and circuit 3, respectively. The left auxiliary airbag 5 and the right auxiliary airbag 6 share one inflation / deflation circuit, designated as circuit 4. Pressure sensors are installed on the air inlets and outlets of the neck airbag 2, the left and right head airbags, and the left and right auxiliary airbags. The microphone, the flexible pressure sensor 9, and the pressure sensor are connected to the audio sensing module, the external pressure sensing module 10, and the internal pressure sensing module 13, respectively.

[0069] The control circuit measures the sound signal, head pressure signal, and airbag pressure signal through the sound sensing module, external pressure sensing module 10, and internal pressure sensing module 13, respectively. The control circuit determines whether the sleeper using the airbag pillow is snoring based on the collected sound signal, and determines whether the patient's current sleeping position is supine or lateral based on the collected head pressure signal and airbag pressure signal. The control circuit uses an air pump to intermittently inflate each airbag according to the patient's current sleeping position to activate supine or lateral anti-snoring.

[0070] like Figure 12The diagram shows the arrangement of the layers of the multifunctional airbag pillow of the present invention. The neck airbag 2, the left head airbag 3, the right head airbag 4, the left auxiliary airbag 5, and the right auxiliary airbag 6 are in the same plane and form an airbag group layer 17. The pillow shell 1 above the airbag group layer 17 consists of a soft sponge layer 14, an array pressure sensor layer 15, and a hard sponge isolation layer 16 from the outside to the inside. The array pressure sensor layer is formed by a flexible pad 8 and eight flexible pressure sensors 9 fixed on the flexible pad. The pillow shell below the airbag group layer 17 consists of a bottom sponge layer 19 and an acrylic support plate layer 18 from the outside to the inside.

[0071] The anti-snoring method of the airbag pillow integrating snoring monitoring, head posture recognition, and anti-snoring functions of the present invention is achieved through the following steps:

[0072] a) Signal acquisition; The control circuit acquires external sound source signals when the user is sleeping through four microphones set on the pillow shell, acquires head pressure signals applied to the flexible pressure sensor by the external pressure sensor module, and acquires internal pressure signals of each airbag by the internal pressure sensor module.

[0073] b) Sound source localization and snoring recognition; based on the geometric spatial relationship of the four microphones, the distance of the sound source from the set origin, as well as the pitch and azimuth angles of the sound source, are calculated to achieve spatial localization of the sound source, so as to distinguish whether the sound source is emitted by the user or the bed partner.

[0074] After determining that the sound source is the user by locating the sound source, the original audio is first subjected to wavelet threshold noise reduction, and the Mel frequency spectrum is extracted as the model input. Then, the MobileNet V2 model based on transfer learning is selected for snoring recognition to achieve binary classification of whether the current sound source is snoring or not.

[0075] like Figure 11 As shown, a coordinate diagram for determining the sound source location in this invention is presented. Step b) of sound source localization and snoring recognition is specifically achieved through the following steps:

[0076] b-1). Establish a rectangular coordinate system; denote the four microphone positions on the pillow shell as M1, M2, M3 and M4. The four microphones on the pillow shell are located in the same plane and at the four vertices of the same rectangle. Establish an xyz coordinate system with the center of the rectangle where the four microphones are located as the origin, the line connecting the origin and M4 as the y-axis, the line connecting the origin and M3 as the x-axis, and the z-axis as the direction. Let d be the distance of M1, M2, M3 and M4 from the origin; the coordinates of the four microphones are M1(0, -d, 0), M2(-d, 0, 0), M3(d, 0, 0) and M4(0, d, 0).

[0077] b-2). Establish the geometric relationship of the sound source; let the coordinates of the sound source S be (x, y, z), and establish the geometric position relationship between the sound source S and the four microphones as shown in formula (1):

[0078]

[0079] In the formula, r, r1, r2, r3, and r4 are the distances of the sound source from the origin and the distances of the microphones M1, M2, M3, and M4, respectively.

[0080] The distance difference between the sound source and the microphone can be represented by the time delay difference. Taking microphone M1 as the reference, the following relationship between the distances of the other three microphones and the sound source is obtained:

[0081]

[0082] In the formula, τ i1 For microphone M i The time delay difference relative to microphone M1, c is the speed of sound propagation, taken as 343m / s;

[0083] b-3). Calculate the location of the sound source; combine formulas (1) and (2) to eliminate x, y, and z, and obtain the distance r of the sound source S:

[0084]

[0085] The azimuth of the sound source is obtained from spatial geometric relationships. And pitch angle θ:

[0086]

[0087] The coordinates of the sound source location are converted into polar coordinates, as shown in formula (5):

[0088]

[0089] Simplify to obtain the azimuth angle And pitch angle θ:

[0090]

[0091] b-4). Sound source identification; based on the distance r of the sound source S from the origin and the azimuth angle... The pitch angle θ determines whether the sound signal comes from the user of the airbag pillow, so as to eliminate interfering sounds from bed partners or other people;

[0092] b-5). Preprocessing of the sound source signal: First, the input signal is decomposed into three levels using the sym8 wavelet basis to obtain wavelet coefficients at different levels. Then, by selecting the number of wavelet function decomposition levels, threshold, and threshold function, an adaptive thresholding method is used to process the signal to remove noise and retain useful information. Finally, the wavelet coefficients after thresholding are reconstructed by inverse wavelet transform to obtain the denoised signal.

[0093] b-6). Snoring recognition; The noise-reduced audio was converted into a Mel cepstral plot as the input to the model. The MobileNet V2 model based on transfer learning was selected for snoring recognition to identify snoring and non-snoring sound source signals.

[0094] c) Head posture recognition; First, head pressure signal data collected by a flexible pressure sensor and airbag pressure signal data collected by a pneumatic pressure sensor are fused to construct a 13-dimensional feature vector containing center of gravity shift; then, a one-dimensional convolutional neural network combined with an SE attention mechanism is used to achieve real-time recognition of six head postures, including no one, lying flat, left side lying, right side lying, left head turning, and right head turning; at the same time, a side-lying rolling judgment mechanism is designed to avoid misjudgment caused by initial value refresh during posture switching.

[0095] Head posture recognition primarily relies on the fusion of data from an array of pressure sensors and the internal pressure of the airbag to make judgments. The collected data includes external pressure data from the pressure sensors and internal pressure data from the airbag. Through the organic combination of the airbag internal pressure and the array of pressure sensors, the system can more accurately capture the pressure distribution and changes of the head on the pillow, thereby achieving precise identification of different head postures. The external pressure data from the array of pressure sensors provides information on the pressure on the contact surface between the head and the pillow, as well as the range of the head, while the airbag internal pressure data reflects the pressure changes within the airbag caused by changes in head posture. This combination can supplement potentially missing data, providing more comprehensive and accurate data support for head posture recognition.

[0096] When a user changes from lying flat to lying on their side, the pressure sensor detects an increase in pressure on one side of the head and a decrease in pressure on the other side. In contrast, the change in the center of gravity is smaller when the head is turned; the center of gravity decreases when turned to the left and increases when turned to the right; and the center of gravity remains basically unchanged when there is no turning.

[0097] Step c) The head pose recognition is specifically implemented through the following steps:

[0098] c-1). Pressure signal acquisition; Eight flexible pressure sensors on the pillow shell are set at equal intervals of 30mm. The pressure inside the airbags collected by the neck airbag, left head airbag, and right head airbag are marked as channel 1, channel 2, and channel 3, respectively. The pressure inside the airbags collected by the left auxiliary airbag and right auxiliary airbag is marked as channel 4.

[0099] Fourteen subjects performed each posture—lying flat, lying on their left side, lying on their right side, with their head turned to the left, and with their head turned to the right—40 times each. Head pressure signals from eight flexible pressure sensors and pressure signals from four airbags from pneumatic sensors were collected, totaling 2800 sets of data. The head center of gravity position was calculated using the head pressure signals from the eight flexible pressure sensors. The head center of gravity position was calculated by dividing the sum of the products of the flexible pressure sensor positions and their corresponding pressure values ​​by the sum of the pressure values ​​detected by all sensors.

[0100] c-2). Head position determination: When the pressure value detected by the 8 flexible pressure sensors is lower than the set threshold, it is determined to be in the position of being off the pillow, and no head posture recognition is performed; when the pressure value detected by the 8 flexible pressure sensors is higher than the set threshold, it is determined to be in the position of being on the pillow; at the same time, the side-lying turning process is also considered, that is, when the number of pressure plates changing in the data of the 8 flexible pressure sensors at the same time does not exceed 6, the head posture can be determined to be normal; if it exceeds 6, this is determined to be the turning process, and no head posture recognition is performed.

[0101] c-3). Model Training: Build a CNN model incorporating the SE attention mechanism. Use the 2800 sets of stress data collected in step c-1) to train the head posture model. At the beginning of training, the 2800 sets of sleeping posture recognition data collected above are randomly shuffled and divided into training set: test set: validation set in a ratio of 7:2:1 to ensure the representativeness and generalization ability of the model's training, validation, and testing processes. The training set consists of 1960 data points for model parameter learning; the test set consists of 560 data points for evaluating the model's performance on unseen data; the validation set consists of 280 data points. The trained head posture recognition model is then obtained.

[0102] d) Anti-snoring intervention and wake-up: Based on the detected head posture, the system automatically selects either a supine or side-lying anti-snoring strategy. By controlling the inflation and deflation of different airbags, it guides the head to turn slightly, improving airway patency. If snoring continues to exceed the set threshold, the system will activate the active wake-up function, using the alternating inflation and deflation of the left and right airbags to push the head to move, achieving a gentle wake-up. A 5-minute intervention cooling-off period is set to avoid frequent interventions affecting sleep quality.

[0103] Step d) The anti-snoring intervention and wake-up process described in step d) is implemented through the following steps:

[0104] d-1). Determine if snoring occurs; collect audio signals from the user's head area in real time and determine whether the user is snoring according to step b); if snoring occurs, proceed to step d-2);

[0105] d-2). Head posture determination; Based on step c), determine whether the user is currently lying flat or on their side. If lying flat, proceed to step d-3); if on their side, proceed to step d-4.

[0106] d-3). Lying flat to stop snoring: First, set a low air pressure value for airbags 1, 2, and 3. Then, inflate airbags 1 and 3 for 20 seconds while deflating airbag 2 for 20 seconds to guide the head to turn to the right and raise the neck support. After completion, the system waits for 30 seconds and re-detects snoring. If snoring disappears, the airbags return to their initial state and the cycle ends.

[0107] If snoring persists, inflate the third airbag for 30 seconds to increase the head tilt angle to the right; wait another 30 seconds after execution and perform a snoring detection; if the snoring disappears, exit the cycle; if snoring persists, continue inflating the third airbag to the maximum limit and hold for 60 seconds, then exit the cycle; after the entire cycle is completed, the system will check again after 5 minutes to see if repeated anti-snoring intervention is needed; if snoring continues for a long time without significant reduction, a threshold can be set to trigger the active wake-up function to ensure the user's sleep quality;

[0108] d-4). Side-lying anti-snoring: First, set a low air pressure value for the fourth airbag, then inflate the fourth airbag for 20 seconds to slightly elevate the head and optimize airway patency; after completion, the system waits for 30 seconds and detects the snoring. If the snoring disappears, the airbag returns to its initial state and the cycle ends. If the snoring still exists, inflate the fourth airbag for another 30 seconds to increase the head elevation.

[0109] After execution, wait another 30 seconds and re-detect snoring. If snoring disappears, exit the cycle. If snoring persists, continue inflating the fourth airbag to the maximum limit and hold for 60 seconds, then exit the cycle. After the entire cycle is completed, the system will detect again after 5 minutes whether the anti-snoring intervention needs to be restarted.

Claims

1. An airbag pillow integrating snoring monitoring, head posture recognition, and anti-snoring functions, comprising a pillow shell (1) and a neck airbag (2), a left head airbag (3), a right head airbag (4), a left auxiliary airbag (5), a right auxiliary airbag (6), and a control box disposed within the pillow shell; the neck airbag supports the user's neck, the left and right head airbags support the head, and the left and right auxiliary airbags are located outside the left and right head airbags, respectively; the control box is separately disposed from the pillow shell; characterized in that: Microphones (7) for collecting sound signals are fixed at the four corners of the pillow shell. Eight flexible pressure sensors (9) are evenly arranged on the pillow shell along its length. The flexible pressure sensors measure the pressure signal on the user's head. The control box contains a control circuit, an air pipe, an air pump (11), an external pressure sensing module (10), an internal pressure sensing module (13), an audio sensing module, and eight solenoid valves (12). The air pump, solenoid valves, and air pipe constitute four inflation / deflation circuits. The neck airbag, the left head airbag, and the right head airbag each use one inflation / deflation circuit, while the left auxiliary airbag and the right auxiliary airbag share one inflation / deflation circuit. Air pressure sensors are installed on the air inlets and outlets of the neck airbag, the left and right head airbags, and the left and right auxiliary airbags. The microphone, the flexible pressure sensor, and the air pressure sensor are connected to the audio sensing module, the external pressure sensing module, and the internal pressure sensing module, respectively. The control circuit measures the sound signal, head pressure signal, and airbag pressure signal through the sound sensing module, external pressure sensing module, and internal pressure sensing module, respectively. Based on the collected sound signal, the control circuit determines whether the sleeper using the airbag pillow is snoring. Based on the collected head pressure signal and airbag pressure signal, the control circuit determines whether the patient's current sleeping position is supine or lateral. Based on the patient's current sleeping position, the control circuit uses an air pump to intermittently inflate each airbag to activate supine or lateral anti-snoring.

2. The airbag pillow integrating snoring monitoring, head posture recognition, and anti-snoring functions according to claim 1, characterized in that: The neck airbag (2), left head airbag (3), right head airbag (4), left auxiliary airbag (5) and right auxiliary airbag (6) are in the same plane and form an airbag assembly layer (17). The pillow shell (1) above the airbag assembly layer consists of a soft sponge layer (14), an array pressure sensor layer (15) and a hard sponge isolation layer (16) from the outside to the inside. The array pressure sensor layer is formed by a flexible pad (8) and eight flexible pressure sensors (9) fixed on the flexible pad. The pillow shell below the airbag assembly layer consists of a bottom sponge layer (19) and an acrylic support plate layer (18) from the outside to the inside.

3. The airbag pillow integrating snoring monitoring, head posture recognition, and anti-snoring functions according to claim 1 or 2, characterized in that: The air outlet of the air pump (11) is connected to the air inlet and outlet of the neck airbag (2), left head airbag (3), right head airbag (4), left auxiliary airbag (5) and right auxiliary airbag (6) via an air tube and four solenoid valves (12). The air inlet and outlet of the neck airbag, left head airbag, right head airbag, left auxiliary airbag and right auxiliary airbag are connected to the outside through the four solenoid valves.

4. A method for stopping snoring based on the airbag pillow with integrated snoring monitoring, head posture recognition, and anti-snoring functions as described in claim 1, characterized in that, This can be achieved through the following steps: a) Signal acquisition; The control circuit acquires external sound source signals when the user is sleeping through four microphones set on the pillow shell, acquires head pressure signals applied to the flexible pressure sensor by the external pressure sensor module, and acquires internal pressure signals of each airbag by the internal pressure sensor module. b) Sound source localization and snoring recognition; based on the geometric spatial relationship of the four microphones, the distance of the sound source from the set origin, as well as the pitch and azimuth angles of the sound source, are calculated to achieve spatial localization of the sound source, so as to distinguish whether the sound source is emitted by the user or the bed partner. After determining that the sound source is the user by locating the sound source, the original audio is first subjected to wavelet threshold noise reduction, and the Mel frequency spectrum is extracted as the model input. Then, the MobileNet V2 model based on transfer learning is selected for snoring recognition to achieve binary classification of whether the current sound source is snoring or not. c) Head posture recognition; First, head pressure signal data collected by a flexible pressure sensor and airbag pressure signal data collected by a pneumatic pressure sensor are fused to construct a 13-dimensional feature vector containing center of gravity shift; then, a one-dimensional convolutional neural network combined with an SE attention mechanism is used to achieve real-time recognition of six head postures, including no one, lying flat, left side lying, right side lying, left head turning, and right head turning; at the same time, a side-lying rolling judgment mechanism is designed to avoid misjudgment caused by initial value refresh during posture switching. d) Anti-snoring intervention and wake-up: Based on the detected head posture, the system automatically selects either a supine or side-lying anti-snoring strategy. By controlling the inflation and deflation of different airbags, it guides the head to turn slightly, improving airway patency. If snoring continues to exceed the set threshold, the system will activate the active wake-up function, using the alternating inflation and deflation of the left and right airbags to push the head to move, achieving a gentle wake-up. A 5-minute intervention cooling-off period is set to avoid frequent interventions affecting sleep quality.

5. The method for stopping snoring using an airbag pillow integrating snoring monitoring, head posture recognition, and anti-snoring functions according to claim 4, characterized in that, Step b) involves sound source localization and snoring recognition, which are achieved through the following steps: b-1). Establish a rectangular coordinate system; denote the four microphone positions on the pillow shell as M1, M2, M3 and M4. The four microphones on the pillow shell are located in the same plane and at the four vertices of the same rectangle. Establish an xyz coordinate system with the center of the rectangle where the four microphones are located as the origin, the line connecting the origin and M4 as the y-axis, the line connecting the origin and M3 as the x-axis, and the z-axis as the direction. Let d be the distance of M1, M2, M3 and M4 from the origin; the coordinates of the four microphones are M1(0, -d, 0), M2(-d, 0, 0), M3(d, 0, 0) and M4(0, d, 0). b-2). Establish the geometric relationship of the sound source; let the coordinates of the sound source S be (x, y, z), and establish the geometric position relationship between the sound source S and the four microphones as shown in formula (1): In the formula, r, r1, r2, r3, and r4 are the distances of the sound source from the origin and the distances of the microphones M1, M2, M3, and M4, respectively. The distance difference between the sound source and the microphone can be represented by the time delay difference. Taking microphone M1 as the reference, the following relationship between the distances of the other three microphones and the sound source is obtained: In the formula, τ i1 For microphone M i The time delay difference relative to microphone M1, c is the speed of sound propagation, taken as 343m / s; b-3). Calculate the location of the sound source; combine formulas (1) and (2) to eliminate x, y, and z, and obtain the distance r of the sound source S: The azimuth of the sound source is obtained from spatial geometric relationships. And pitch angle θ: The coordinates of the sound source location are converted into polar coordinates, as shown in formula (5): Simplify to obtain the azimuth angle And pitch angle θ: b-4). Sound source identification; based on the distance r of the sound source S from the origin and the azimuth angle... The pitch angle θ determines whether the sound signal comes from the user of the airbag pillow, so as to eliminate interfering sounds from bed partners or other people; b-5). Preprocessing of the sound source signal: First, the input signal is decomposed into three levels using the sym8 wavelet basis to obtain wavelet coefficients at different levels. Then, by selecting the number of wavelet function decomposition levels, threshold, and threshold function, an adaptive thresholding method is used to process the signal to remove noise and retain useful information. Finally, the wavelet coefficients after thresholding are reconstructed by inverse wavelet transform to obtain the denoised signal. b-6). Snoring recognition; The noise-reduced audio was converted into a Mel cepstral plot as the input to the model. The MobileNet V2 model based on transfer learning was selected for snoring recognition to identify snoring and non-snoring sound source signals.

6. The snoring prevention method of the airbag pillow integrating snoring monitoring, head posture recognition, and anti-snoring functions according to claim 4 or 5, characterized in that, Step c) The head pose recognition is specifically implemented through the following steps: c-1). Pressure signal acquisition; Eight flexible pressure sensors on the pillow shell are set at equal intervals of 30mm. The pressure inside the airbags collected by the neck airbag, left head airbag, and right head airbag are marked as channel 1, channel 2, and channel 3, respectively. The pressure inside the airbags collected by the left auxiliary airbag and right auxiliary airbag is marked as channel 4. Fourteen subjects performed each posture—lying flat, lying on their left side, lying on their right side, with their head turned to the left, and with their head turned to the right—40 times each. Simultaneously, head pressure signals from eight flexible pressure sensors and pressure signals from four airbags from pneumatic sensors were collected, totaling 2800 sets of data. The head center of gravity position of the user was calculated using the head pressure signals from the eight flexible pressure sensors. The head center of gravity position was calculated by dividing the sum of the products of the position of the flexible pressure sensor and its corresponding pressure value by the sum of the pressure values ​​detected by all sensors. c-2). Head position determination: When the pressure value detected by the 8 flexible pressure sensors is lower than the set threshold, it is determined to be in the position of being off the pillow, and no head posture recognition is performed; when the pressure value detected by the 8 flexible pressure sensors is higher than the set threshold, it is determined to be in the position of being on the pillow; at the same time, the side-lying turning process is also considered, that is, when the number of pressure plates changing in the data of the 8 flexible pressure sensors at the same time does not exceed 6, the head posture can be determined to be normal; if it exceeds 6, this is determined to be the turning process, and no head posture recognition is performed. c-3). Model Training; Build a CNN model combining SE attention mechanism, and train the head posture model using the 2800 sets of pressure data collected in step c-1). At the beginning of training, the 2800 sets of sleeping posture recognition data collected above are randomly shuffled and divided into training set: test set: validation set in a ratio of 7:2:1 to ensure that the training, validation and testing process of the model is representative and has generalization ability; 1960 data points are used for model parameter learning. The test set consists of 560 data points and is used to evaluate the model's performance on unseen data; the validation set consists of 280 data points. The head pose recognition model was obtained through training.

7. The method for stopping snoring using an airbag pillow integrating snoring monitoring, head posture recognition, and anti-snoring functions as described in claim 6, characterized in that... Step d) The anti-snoring intervention and wake-up process described in step d) is implemented through the following steps: d-1). Determine if snoring occurs; collect audio signals from the user's head area in real time and determine whether the user is snoring according to step b); if snoring occurs, proceed to step d-2); d-2). Head posture determination; Based on step c), determine whether the user is currently lying flat or on their side. If lying flat, proceed to step d-3); if on their side, proceed to step d-4. d-3). Lying flat to stop snoring: First, set a low air pressure value for airbags 1, 2, and 3. Then, inflate airbags 1 and 3 for 20 seconds while deflating airbag 2 for 20 seconds to guide the head to turn to the right and raise the neck support. After completion, the system waits for 30 seconds and re-detects snoring. If snoring disappears, the airbags return to their initial state and the cycle ends. If snoring persists, inflate the third airbag for 30 seconds to increase the head tilt angle to the right; wait another 30 seconds after execution and perform a snoring detection; if the snoring disappears, exit the cycle; if snoring persists, continue inflating the third airbag to the maximum limit and hold for 60 seconds, then exit the cycle; after the entire cycle is completed, the system will check again after 5 minutes to see if repeated anti-snoring intervention is needed; if snoring continues for a long time without significant reduction, a threshold can be set to trigger the active wake-up function to ensure the user's sleep quality; d-4). Side-lying anti-snoring: First, set a low air pressure value for the fourth airbag, then inflate the fourth airbag for 20 seconds to slightly elevate the head and optimize airway patency; after completion, the system waits for 30 seconds and detects the snoring. If the snoring disappears, the airbag returns to its initial state and the cycle ends. If the snoring still exists, inflate the fourth airbag for another 30 seconds to increase the head elevation. After execution, wait another 30 seconds and re-detect snoring. If snoring disappears, exit the cycle. If snoring persists, continue inflating the fourth airbag to the maximum limit and hold for 60 seconds, then exit the cycle. After the entire cycle is completed, the system will detect again after 5 minutes whether the anti-snoring intervention needs to be restarted.

Images