Sleep closed-loop management method, head-mounted sleep closed-loop management system and electronic equipment

Abstract

The invention discloses a sleep closed-loop management method, which is applied to a head-mounted sleep closed-loop management device, and comprises the following steps: performing real-time sleep monitoring according to an electroencephalogram signal, an electro-oculogram signal, a heart rate and blood oxygen saturation; when the sleep monitoring result meets a preset condition, an intervention component is controlled to perform sleep intervention; the sleep monitoring comprises statistics of sleep stages and duration of the current sleep stage, and the sleep monitoring result comprises the current sleep stage and the duration of the current sleep stage. The invention further discloses a head-mounted sleep closed-loop management system and electronic equipment. According to the invention, sleep staging and current sleep stage duration statistics are carried out through a deep learning algorithm based on multi-mode signals, real-time and accurate sleep state monitoring is realized, effective dynamic intervention can be carried out on sleep based on a sleep monitoring result, and closed-loop management from sleep monitoring to sleep intervention is formed.

Description

Technical Field

[0001] The present invention belongs to the technical field of sleep state monitoring, and in particular relates to a head-mounted sleep detection and sleep intervention technology based on multimodal signals. Background Art

[0002] Sleep is essential to human health, and a normal human spends about one-third of his or her life sleeping. Studies have shown that sleep is essential to the cardiovascular system, and long-term lack of sleep is associated with heart diseases such as hypertension, coronary heart disease, and heart failure. Secondly, sleep is important for emotional regulation and mental health, and lack of sleep increases the risk of anxiety, depression, and other psychological problems. Insomnia is a common sleep disorder, which is defined as a sleep latency or wakefulness time after falling asleep > 20 minutes (sleep latency refers to the time it takes for the user to turn off the lights and prepare to go to sleep to the time he or she falls asleep). Globally, approximately 10% to 30% of the population, as many as 237 million people, are affected by insomnia. In adults, chronic insomnia may impair social or occupational function and reduce quality of life. People with more severe insomnia may have an increased risk of motor vehicle and workplace accidents as well as mental and cardiovascular diseases.

[0003] Through the characteristics of various physiological signals, according to the AASM standard, the sleep state can be divided into wakefulness (W stage), non-rapid eye movement stage I (N1 stage), non-rapid eye movement stage II (N2 stage), non-rapid eye movement stage III (N3 stage) and rapid eye movement stage (REM stage), among which stage N3 is also called slow wave sleep, usually referring to deep sleep. Polysomnography (PSG) is the gold standard method for evaluating sleep quality and diagnosing sleep disorders (including insomnia, apnea, etc.), and is also the main reference for equipment verification. This is a comprehensive sleep measurement method that measures scalp electroencephalogram (EEG), electrooculogram (EOG), mandibular electromyography, electrocardiogram, breathing, leg movement, nasal pressure, oxygen saturation (SPO2, measured using a PPG sensor, so it is also called a PPG signal) and body position, which is used for sleep monitoring and characterization of sleep disorders. Polysomnography is the gold standard device for sleep monitoring. It can detect the above-mentioned multiple physiological signals and is often used by doctors for sleep staging. However, since it requires a dedicated PSG acquisition system and professionals to record, score and interpret PSG data, its use is usually limited to sleep laboratory research and clinical environments, which is time-consuming and expensive. In addition to the cost, the large number of sensors and their distribution throughout the body, as well as the need to enter the sleep laboratory for all-night dynamic monitoring, can make patients feel obvious discomfort.

[0004] To solve the above problems, many researchers have developed wearable portable devices that can monitor sleep in non-laboratory environments, and some can also perform sleep staging. The first generation of wearable sleep monitoring devices uses the built-in accelerometer of the device to measure body movement based on the activity recorder to infer sleep and wakefulness, but this method cannot accurately distinguish between sleeping and non-movement wakefulness. Since the head can detect electrophysiological signals that are critical for sleep staging and disease diagnosis, the second generation of devices uses the head as a key position to monitor physiological signals such as PPG and EEG, further improving the accuracy of sleep monitoring. During rapid eye movement sleep (REM), the human body will experience rapid eye movement. Therefore, the REM stage can be accurately identified by EOG signals, but due to the complex layout and positioning of EOG electrodes, they are located on both sides of the face and are easily compressed when sleeping on the side. Currently, there are few head-mounted devices that detect EOG signals. EEG signals are weak voltage signals at the μV level, and their amplitude and waveform are easily affected by the position of the acquisition electrode and the reference electrode. However, many current second-generation devices have non-standardization problems due to the different positions of the reference electrode and the acquisition electrode. Most of the acquired signals do not meet the medical sleep staging standards in terms of amplitude and waveform, and cannot be interpreted by professional doctors. The accuracy of the sleep staging algorithm developed based on unknown signal quality cannot be determined. In addition, most of the current sleep staging algorithms analyze data throughout the night, making it difficult to achieve real-time sleep staging.

[0005] In terms of sleep intervention, a variety of devices have shown potential, such as transcranial magnetic stimulator (TMS), transcranial electrical stimulator (tES), 40Hz light stimulation glasses, forehead cooling devices, and smart headbands with integrated EEG monitoring and acoustic feedback. These devices have been proven to have a positive impact on sleep quality through different mechanisms of action (such as neural regulation, light therapy, temperature regulation, and acoustic stimulation). However, these technologies have only made progress in a single sleep intervention function.

[0006] Therefore, there is still a lack of a portable, wearable, integrated device that can not only achieve real-time and accurate sleep monitoring, but also achieve effective sleep intervention on this basis. Summary of the invention

[0007] To solve the above problems, the present invention proposes a sleep closed-loop management method and a head-mounted sleep closed-loop management system and electronic device, which are based on multimodal signals (electroencephalogram signals, electrooculogram signals, heart rate and blood oxygen saturation) and use a deep learning algorithm to perform sleep staging and current sleep stage duration statistics, thereby achieving real-time and accurate sleep status monitoring, and based on the sleep monitoring results, can effectively and dynamically intervene in sleep, forming a closed-loop management from sleep monitoring to sleep intervention.

[0008] The technical solution of the present invention includes: A first aspect discloses a sleep closed-loop management method, which can be applied to a head-mounted sleep closed-loop management device, comprising: Real-time sleep monitoring based on EEG signals, EOG signals, heart rate and blood oxygen saturation; and controlling the intervention component to perform sleep intervention when the sleep monitoring result meets the preset conditions; Sleep monitoring includes sleep staging and current sleep stage duration statistics. The sleep monitoring results include the current sleep stage and its duration. The sleep stages include W stage, N1 stage, N2 stage, N3 stage and REM stage.

[0009] As an optional solution, the method further includes: Acquire EEG signals, EOG signals and photoplethysmogram signals; Heart rate and blood oxygen saturation are calculated based on the photoplethysmographic signal.

[0010] As an optional solution, the sleep staging is implemented by a sleep staging convolutional neural network, which includes: The feature extraction and fusion module is configured as follows: the signal features of EEG signals, EOG signals, heart rate and blood oxygen saturation are extracted through four feature extraction branches, and the extracted four signal features are fused; the feature extraction branch optimizes the feature expression through convolution layer, batch normalization, ReLU activation function and maximum sampling, and combines Dropout to prevent overfitting; The feature enhancement module is configured to: dynamically assign feature weights to the fused feature array through convolutional layers and tanh activation functions, amplify key features, and filter redundant information; The classifier is configured to map the enhanced feature array to five sleep stages: W, N1, N2, N3, and REM through a fully connected layer and a Softmax layer.

[0011] As an optional approach, sleep intervention can be performed when the preset conditions are met, including: When the current sleep stage is stage W and the statistical duration of stage W exceeds the first preset time, sleep-aiding intervention is performed; and when the current sleep stage changes from stage W to other sleep stages other than stage W and the statistical duration of other sleep stages exceeds the second preset time, sleep-aiding intervention is stopped.

[0012] As an optional solution, sleep intervention is performed when the preset conditions are met, and also includes: When the current sleep stage is stage N3, deep sleep intervention is performed; and when the current sleep stage changes from stage N3 to other sleep stages, the deep sleep intervention instruction is stopped.

[0013] As an optional solution, the sleep intervention method includes at least one of light intervention, audio intervention and forehead cooling intervention.

[0014] As an optional solution, the sleep monitoring also includes counting the total duration of each sleep stage.

[0015] The second aspect discloses a head-mounted sleep closed-loop management system, comprising a head-mounted sleep closed-loop management device and a computer program; the head-mounted sleep closed-loop management device comprises a head-mounted wearable mechanism, and a signal acquisition component, an internal circuit and an intervention component arranged on the head-mounted wearable mechanism; the signal acquisition component comprises an electrooculogram electrode, an electroencephalogram electrode, a reference electrode and a PPG sensor; wherein the electrooculogram electrode is used to collect electrooculogram signals; the electroencephalogram electrode is used in conjunction with the reference electrode to collect electroencephalogram signals; the PPG sensor is used to obtain human photoelectric volume pulse wave signals; the internal circuit is used to pre-process the signals collected by the signal acquisition component, and to control the intervention component to perform sleep intervention according to the received intervention instructions; the computer program has one or more, stored in the internal circuit and / or an electronic device that is communicatively connected to the head-mounted sleep closed-loop management device; when the computer program is executed, it is used to implement the sleep closed-loop management method described in the first aspect of the present invention and any one of the optional schemes.

[0016] As an optional solution, there is a pair of EOG electrodes, which are respectively located below the outer canthus of the left eye and above the outer canthus of the right eye when the head-mounted wearable mechanism is worn; there is a pair of EEG electrodes, which are respectively located on the left and right sides of the frontal lobe of the human body when the head-mounted wearable mechanism is worn; there is a pair of reference electrodes, which are respectively in contact with the mastoid processes at the junction of the left and right external auricles and cheeks of the human body when the head-mounted wearable mechanism is worn.

[0017] As an optional solution, when the head-mounted wearable mechanism is worn, a pair of EEG electrodes are arranged 4-6 cm to the left and right of the inner midline of the eye mask and 0.5-2.5 cm from the upper edge; a pair of EOG electrodes are arranged 6-8 cm to the left and right, 0.5-2.5 cm from the lower edge and 4-6 cm from the upper edge.

[0018] As an optional solution, the EOG and EEG electrodes are detachably connected to the head-mounted wearable device via a flexible Velcro tape; the Velcro tape includes a velvet surface and a hook surface, the velvet surface is fixed to the inner side of the head-mounted wearable device and has an area larger than the hook surface, and the EOG and EEG electrodes are bonded to the velvet surface area via the hook surface.

[0019] As an optional solution, the internal circuit includes an MCU, and a signal processing module, a first communication module, an intervention drive module and a power supply module electrically connected to the MCU; the signal acquisition module is used to pre-process the electrooculogram, electroencephalogram and human photoelectric volume pulse wave signals input into the internal circuit, and the preprocessing includes signal amplification, digital-to-analog conversion and analog-to-digital conversion; the first communication module is used to realize the communication connection between the MCU and the electronic device; the intervention drive module is electrically connected to the intervention component.

[0020] As an optional solution, the intervention component includes at least one of a stroboscopic lighting module and a bone conduction earphone; the intervention driving module includes at least one of a light driving module electrically connected to the stroboscopic lighting module and an earphone driving module electrically connected to the bone conduction earphone.

[0021] As an optional solution, the intervention component also includes a semiconductor refrigeration module; the intervention drive module includes a refrigeration drive module electrically connected to the semiconductor refrigeration module; the signal acquisition component also includes a temperature sensor for collecting human body temperature signals; the computer program is also used to achieve: during the operation of the semiconductor refrigeration module, when it is detected that the human body temperature signal is lower than the temperature threshold, the semiconductor refrigeration module is controlled to stop sleep intervention.

[0022] As an optional solution, the wearable mechanism is an eye mask; the eye mask has an eye mask body and an eye mask strap; the electrooculogram electrodes, the electroencephalogram electrodes and the PPG sensor are arranged on the side of the eye mask body close to the skin; the internal circuit is arranged inside the eye mask body; the reference electrode is arranged on the eye mask strap; when the intervention component includes at least one of a stroboscopic lighting module, a bone conduction earphone and a semiconductor refrigeration module, the stroboscopic lighting module and the semiconductor refrigeration module are arranged inside the eye mask body; and the bone conduction earphone is arranged on the eye mask strap.

[0023] As an optional solution, the wearable mechanism is a flexible structure; the internal circuit is a flexible circuit board; the EOG, EEG and reference electrodes are flexible electrodes, wherein the EOG and EEG electrodes are also constructed as serpentine patterns with stretchable lines; the wires used to connect various electrodes, sensors, intervention components and internal circuits are all constructed as serpentine traces with stretchable lines.

[0024] The second aspect discloses an electronic device, including a processor, a memory, and a second communication module; the electronic device is communicatively connected to a head-mounted sleep closed-loop management device through the second communication module; the processor is used to call a computer program stored in the memory to implement the sleep closed-loop management method described in the first aspect of the present invention and any optional scheme.

[0025] The present invention has the following beneficial effects: (1) The present invention uses the electroencephalogram (EEG) signals, electrooculogram (EOG) signals and photoplethysmogram (PPE) signals of the head through a head-mounted sleep closed-loop management device to calculate heart rate and blood oxygen saturation, and divides sleep into stages based on the EEG signals, EOG signals, heart rate and blood oxygen saturation, and performs real-time statistics on the duration of the current sleep stage, and performs real-time and accurate sleep status monitoring. At the same time, based on the sleep monitoring results, the present invention controls the intervention components integrated in the head-mounted sleep closed-loop management device to perform effective sleep intervention (sleep-aid intervention and deep sleep intervention), thereby forming a closed-loop management from sleep monitoring to sleep intervention, which can effectively improve insomnia and promote deep sleep.

[0026] (2) The sleep staging method based on multimodal signals (EEG, EOG, heart rate, blood oxygen signal) of the present invention has significant advantages in accuracy and real-time performance. By integrating multiple physiological signals, it can more comprehensively reflect the sleep state and significantly improve the accuracy of staging. In particular, multimodal signals can capture subtle changes in EEG slow waves, heart rate variability, and blood oxygen levels, achieve high-precision identification of stage N3 (deep sleep), reduce errors caused by improper electrode position, noise, etc. in a single signal, and improve the robustness and accuracy of sleep staging. In addition, multi-dimensional analysis enhances real-time monitoring capabilities, provides more reliable technical support for sleep health management, and is suitable for scenarios such as wearable devices.

[0027] (3) The present invention adopts a sleep staging convolutional neural network. In the sleep staging task, the convolution kernel is used to extract the local features of the signal (such as changes in brain wave rhythm), and a multi-layer structure is used to learn the sleep pattern features from low-level to high-level, which significantly improves the staging accuracy. Its parameter sharing and sparse connections reduce the computational complexity and are suitable for processing high-dimensional physiological signal data. CNN is translationally invariant and can adapt to individual differences. The convolution operation can be highly parallelized, with high training and inference efficiency, making it suitable for real-time applications. In addition, CNN supports end-to-end learning, without the need for manual feature design, which simplifies the complex process of traditional sleep staging.

[0028] (4) The present invention provides a variety of optional intervention methods for photoacoustic temperature control. The method of combining photoacoustic temperature control to implement sleep intervention can be highly organically integrated with wearable devices, providing innovative solutions for sleep management. By combining light stimulation of a specific frequency (such as 40Hz light), acoustic feedback (such as pink noise or sleep-aiding audio), and forehead temperature control feedback, wearable devices can accurately adjust the user's sleep state, such as promoting deep sleep or alleviating insomnia. Photoacoustic temperature control intervention has the characteristics of non-invasiveness, low power consumption and high compatibility, and is very suitable for integration into wearable devices such as smart headbands, eye masks or headphones. This multimodal intervention method not only improves the portability and user experience of the device, but also realizes dynamic closed-loop management through real-time monitoring of sleep data, providing users with personalized and adaptive sleep optimization solutions, thereby significantly improving sleep quality.

[0029] (5) The present invention combines a semiconductor refrigeration device with a temperature sensor to achieve closed-loop control of core body temperature at the forehead, and uses precise adjustment of forehead temperature to intervene in sleep. By real-time monitoring of forehead temperature, the system automatically adjusts the cooling intensity according to the set safety threshold to ensure that the temperature change does not exceed the safety range and avoid adverse reactions, thereby improving insomnia and improving sleep quality. Compared with traditional water-cooling cycle cooling methods, this technology has higher accuracy, response speed and wearability.

[0030] (6) The present invention optimizes the position of the reference electrode, placing it on the inside of the eye mask strap, where the auricle intersects the cheek (near the mastoid) when worn, to provide a reference potential signal. The potential difference between the potential signals collected by the two EEG electrodes on the left and right sides of the forehead and the reference signals of the left and right reference electrodes forms two EEG signals. At the same time, the reference electrode is located at a location with thin skin and thick bones, and is far from the cerebral cortex, so it is not easily disturbed by sweat, body movement, brain activity and other factors, and can provide a stable reference signal.

[0031] (7) The present invention also optimizes the position of EEG and EOG electrodes on the head and face, and optimizes the position of the electrodes according to the forehead length, width, eye distance and eye width of Asian adults, so that the electrode positions meet the international PSG wearing standards and ensure the high consistency and interpretability of the signal characteristics and clinical equipment.

[0032] (8) The present invention uses flexible Velcro to achieve the coordinated use of the velvet surface and the hook surface, which enables flexible adjustment and accurate positioning of the electrode position, and accommodates the deviation caused by differences in human facial structure. When wearing it for the first time, the user can accurately adjust the electrode position according to the characteristics of his or her facial structure, and no adjustment is required for subsequent wear.

[0033] (9) The present invention can design the head-mounted closed-loop sleep management device into a flexible structure. By using fabric, flexible circuit boards, and flexible stretchable electrodes and wires, the wearing comfort of the device can be improved, the impact on the patient's sleep can be reduced, and the accuracy of sleep staging can be further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a schematic diagram of the structure of the smart eye mask; Figure 2 This is a schematic diagram of the explosion of the smart eye mask; Figure 3 :(a) is the illustration of wearing the smart eye mask Figure 1 ; (b) is a diagram of how to wear the smart eye mask Figure 2 ; Figure 4 : (a) is a schematic diagram of the electrode structure of the smart eye mask; (b) is a schematic diagram of the connecting wire structure of the smart eye mask; Figure 5This is a schematic diagram of the connection of electrophysiological monitoring electrodes; Figure 6 Schematic diagram of the photoplethysmogram measured within a 5-second time window; Figure 7 This is a blood oxygen calculation flow chart; Figure 8 This is a circuit connection diagram of the smart eye mask system; Fig. 9 This is a flowchart of the real-time sleep staging algorithm; Fig.10 This is a schematic diagram of the sleep intervention process; Fig.11 This is a schematic diagram of the closed-loop sleep management process. DETAILED DESCRIPTION

[0035] The technical solution of the present invention will be clearly and completely described below in conjunction with specific embodiments and drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] In the description of the present invention, if there are terms indicating orientation or positional relationship such as "upper", "lower", "left", "right", "front", "back", "inside", "outside", "top", "bottom", etc., they are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, if there are terms such as "first", "second", etc., they are used to distinguish similar objects, and are not necessarily used to describe a specific order or relative importance. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, but may include other units that are not clearly listed or inherent to these products or devices.

[0037] Combination Figure 1 and Figure 2 As shown, Embodiment 1 of the present invention discloses a wearable non-intrusive smart eye mask (referred to as "smart eye mask"), which is mainly composed of an eye mask body 2 and an eye mask strap 1 in appearance. The eye mask body 2 is mainly composed of an eye mask lining 21, an eye mask shell 22 and a supporting sponge 23, and is provided with a flexible low-voltage LED light bar (referred to as "LED light bar") 24, an internal circuit 25, a semiconductor refrigeration module 26 and a power supply module (not shown in the figure), as well as a PPG sensor 27, a temperature sensor 28 and an electrophysiological monitoring electrode arranged on the eye mask lining 21.

[0038] The eye mask strap 1 mainly includes an elastic elastic band 15, and the two ends of the elastic elastic band 15 are fixedly connected to the eye mask shell 22. Reference electrodes (left reference electrode 11 and right reference electrode 14) and bone conduction earphones (left bone conduction earphone 12 and right bone conduction earphone 13) are arranged on the elastic elastic band 15. In the eye mask body 2, the eye mask shell 22 is located at the outermost side (the side away from the skin when worn), and the eye mask lining 21 is attached to the support sponge 23, and its edge can be sewn to the opening at the edge of the support sponge 23 with sutures, and the whole formed can be called the eye mask lining. The eye mask lining 21 and the eye mask shell 22 can be connected by means of snapping, bonding, plugging, etc. When worn, the eye mask lining 21 is close to the facial skin and is located at the innermost side.

[0039] The eye mask lining 21, the eye mask shell 22 and the eye mask strap 1 together constitute a wearable mechanism for carrying components such as the LED light bar 24, the internal circuit 25, the PPG sensor 27, the temperature sensor 28, the reference electrode, the electrophysiological monitoring electrode, the semiconductor refrigeration module 26, the bone conduction earphone, and the power supply module. In terms of material selection, the material of the eye mask lining 21 is preferably medical non-woven fabric, and the eye mask shell 22 can use materials with good thermal conductivity, such as carbon fiber; the support sponge 23 is preferably HKGK high-density sponge.

[0040] The electrophysiological monitoring electrodes are arranged on the inner side of the eye mask lining 21 (attached to one side of the skin when worn), mainly including the left electrooculogram electrode (also called "LEOG electrode") 212, the right electrooculogram electrode (also called "REOG electrode") 215, the left electroencephalogram electrode (also called "EEG-FP1 electrode" or "FP1 electrode") 213, and the right electroencephalogram electrode (also called "EEG-FP2 electrode" or "FP2 electrode") 214. Among them, the LEOG electrode 212 is located at the lower left of the eye mask lining 21, that is, the lower left of the PPG sensor 27, and is attached to 1 cm below the left outer corner of the human body when worn; the EEG-FP1 electrode 213 is located at the upper left of the eye mask lining, and is attached to the left forehead when worn; the EEG-FP2 electrode 214 is located at the upper right of the eye mask lining, and is attached to the right forehead when worn; the REOG electrode 215 is located at the upper right of the eye mask lining 21, and is attached to 1 cm above the right corner of the human body when worn. The EEG-FP1 electrode 213 and the EEG-FP2 electrode 214 are axially symmetrically distributed with the midline of the eye mask as the axis. The potential difference between the LEOG electrode 212 and the REOG electrode 215 is the electrooculographic signal (also called "EOG signal"). The potential difference between the EEG-FP1 electrode 213 and the left reference electrode 11 forms an electroencephalogram signal (also called "EEG signal"); the potential difference between the EEG-FP2 electrode 214 and the right reference electrode 14 forms another electroencephalogram signal.

[0041] In the reference electrodes, the left reference electrode 11 and the left bone conduction earphone 12 are arranged on the left side of the elastic band 15, and the right reference electrode 14 and the right bone conduction earphone 13 are arranged on the right side of the elastic band 15. When worn, the left and right reference electrodes are respectively in contact with the mastoids at the junction of the left and right auricles and cheeks to provide a stable, low-amplitude reference potential. Figure 3 As shown, when worn, the eye mask strap 1 surrounds the head for a whole week, the left reference electrode 11 and the right reference electrode 14 are attached to the top of the left and right auricles respectively, and the left bone conduction earphone 12 and the right bone conduction earphone 13 are attached to the temporal part in front of the ear.

[0042] It is worth noting that the present invention optimizes the position of the reference electrode and arranges the reference electrode at the mastoid process where the left and right auricles meet the cheeks. Compared with the occipital region, the area near the mastoid process is less affected by head movement and is free of complex external signal interference (such as electrocardiogram signals). Compared with the forehead and the tip of the nose, the skin near the mastoid process is thinner and sweats less, ensuring the long-term stability of the electrode-skin interface. At the same time, the mastoid process behind the ear is a raised bone, and the thicker bone structure is conducive to isolating the brain's activity signals. The mastoid process is located in the middle and rear of the entire skull, and is far away from the main electrical activity area of ​​the cerebral cortex, which can provide a reference potential that is not affected by brain activity as a reference signal.

[0043] It is worth noting that, in a preferred embodiment, the present invention can design specific electrophysiological monitoring electrode arrangement positions according to the anatomical characteristics of the Asian human body, and the electrode position can be adjusted by using flexible Velcro to ensure that after setting the position once, it does not need to be set again, and the position is always correct. Since the length of the human forehead (from the hairline to the upper edge of the eye) is about 6-8 cm, the width of the forehead (from the hairline on both sides to the inner side of the eye) is about 10-12 cm, the distance between the eyes (the horizontal distance between the inner corners of the two eyes) is about 2.5-3.5 cm, and the length of the eye (the horizontal distance from the corner of the eye to the corner of the eye) is about 2.2-3.0 cm. Therefore, according to the facial anatomical characteristics of Asian adults, 4 Velcro velvets can be provided on the midline of the eye mask lining 21, 4-6 cm to the left and right and 0.5-2.5 cm from the upper edge, 6-8 cm to the left and right and 0.5-2.5 cm from the lower edge, and 4-6 cm from the upper edge, to arrange the corresponding electrodes, each with a size of 2×2 cm 2 Various types of electrophysiological monitoring electrodes are attached to the eye mask lining 21 through Velcro, and are electrically connected to the internal circuit 25 through wires passing through the eye mask lining 21.

[0044] like Figure 5 As shown, the Velcro is divided into a velvet surface 291 and a hook surface 292. The size of the electrodes is 1×1 cm 2 Velcro suede 291 size is 2×2cm 2The Velcro hook surface 292 is the same size as the electrode, 1×1 cm 2 The Velcro velvet surface 291 is bonded to the eye mask lining 21 via the Velcro adhesive, and the electrode is bonded to the Velcro hook surface 292 via the Velcro adhesive. The Velcro velvet surface 291 and the Velcro hook surface 292 are used in conjunction to adjust and fix the electrode position. When wearing for the first time, the user can adjust the electrode to the corresponding position of the Velcro velvet surface 291 via the Velcro hook surface 292 according to the user's facial structure characteristics and the instructions, and no adjustment is required for subsequent wearing.

[0045] In a preferred embodiment, in order to ensure good contact between the electrodes and the skin, various electrodes (EEG electrodes, EOG electrodes and reference electrodes) need to be flexible and able to deform like the fabric of the eye mask and adapt to different wearing conditions. In addition, the eye mask strap 1 needs to be stretched repeatedly when worn, so the reference electrode and the connecting wire also need to have sufficient stretchability to avoid breakage during the stretching process of the strap, thereby ensuring stable signal collection and transmission. Figure 4 As shown in (a) in the figure, these electrodes can be made by high-precision laser cutting based on ultra-thin metal foils with a serpentine pattern. Compared with hard electrodes such as silver / silver chloride and gold cup electrodes, flexible electrodes have higher skin conformality and comfort, and can effectively avoid the problem of poor contact due to uneven skin. Figure 4 As shown in (b), the wires between each electrode and the internal circuit can also be serpentine, which can effectively avoid circuit breakage caused by stretching and deformation of the eye mask. Furthermore, ultra-thin and stretchable polyimide film can be used to insulate and encapsulate the serpentine line through hot melting and hot pressing to ensure stable and reliable signal transmission.

[0046] Based on this, various types of electrophysiological monitoring electrodes can be made of gold foil, and the main production steps are as follows: first, use CAD software to design a specific serpentine pattern to improve the flexibility of the electrode; then place the gold foil on the PDMS substrate and perform laser cutting (for example, 80% power cutting 40 times); then use tweezers to remove the edge of the gold foil material; to ensure that the serpentine pattern does not deform, use PU film to stick the remaining part; finally, place the obtained electrode on polyurethane (PU) hot melt adhesive for high-temperature embossing and then form it; finally, weld the formed electrode to one end of the wire, and connect the other end of the wire to the internal circuit 25.

[0047] The reference electrode is mainly made of polyurethane (PU) material, Ag / AgCl slurry and copper foil, and can be made by the following steps: first, the copper foil is cut into a width that is basically the same as the width of the strap and a length of about 5 mm; double-sided tape is used to stick it on the non-sticky surface of the PU material; Ag / AgCl slurry is applied to the non-sticky surface of the flat and wrinkle-free PU material, covering 1 / 2 of the area of ​​the copper foil, and placed in an oven for drying at 60°C for 1 hour; then the dried PU material coated with Ag / AgCl is cut to a width that is basically the same as the eye mask strap; the adhesive backing is pasted to the elastic band 15, one end of the wire is soldered to the copper foil, and the other end is connected to the internal circuit 25.

[0048] The semiconductor refrigeration module 26 is embedded in the support sponge 23 and can be in sheet form, but in other embodiments, it can also be set to other structural forms according to the structural characteristics of the eye mask, and the present invention is not limited to this. The semiconductor refrigeration module 26 works based on the thermoelectric effect, mainly transferring heat from one side to the other side through the thermoelectric effect of the thermoelectric material, thereby achieving refrigeration. Therefore, the semiconductor refrigeration module 26 is divided into a front side and a back side. After the front side contacts the support sponge 23, it can cool the forehead of the human body, and the back side is close to the eye mask shell 22 and dissipates heat through the eye mask shell 22. The semiconductor refrigeration module 26 is connected to the internal circuit 25 through a wire.

[0049] The PPG sensor 27 and the temperature sensor 28 are also arranged on the inner side of the eye mask lining 21. The PPG sensor 27 and the temperature sensor 28 are embedded in the upper left side of the eye mask lining 21, and can also be electrically connected to the internal circuit 25 through the wires passing through the eye mask lining 21. When the smart eye mask is worn, the PPG sensor 27 and the temperature sensor 28 are attached near the left supraorbital artery of the human body.

[0050] The PPG sensor 27 has a light source and a signal receiver. It acquires the photoelectric volumetric pulse wave signal (referred to as "pulse wave signal" or "PPG signal") of the human body by emitting and receiving red light and infrared light, and converts it into a digital signal for output. For example, the common commercial sensor MAX30102 can be used. The pulse wave signal output by the PPG sensor 27 can be used to calculate the heart rate, heart rate variability and blood oxygen data of the human body. The specific calculation can be performed in the internal circuit 25, and the data after the calculation is completed is transmitted to the host computer. Alternatively, the above calculation can also be completed in the host computer, so that the acquired pulse wave signal can be directly transmitted to the host computer.

[0051] In an optional solution, the heart rate is calculated by the beat-to-beat blood oxygen calculation method. The calculation of heart rate and heart rate variability mainly uses the infrared light signal in the pulse wave signal, and the heart rate can be determined by calculating the average value of the RR interval (the time interval between the R waves of two consecutive heartbeats). , the formula is:

[0052] In the formula, 60 represents 60 seconds, and the unit of RR interval is also seconds.

[0053] Combination Figure 6 As shown, in another optional solution, the time window calculation method is used as an example to calculate the heart rate. The specific calculation process includes setting a time window ΔT of a specific length (taking 5 seconds as an example), storing a new 1-second photoelectric volume pulse wave every 1 second in this time window, and releasing the 1-second photoelectric volume pulse wave 5 seconds ago. In this segment of data, find the main trough value of the photoelectric volume pulse wave and record its time. Calculate the average value of the RR interval based on multiple peak times, and then calculate the heart rate as the heart rate value of that second. The formula is as follows: (2) In the formula, Indicates the pulse wave number within the time window, represents the time of the trough of the i-th pulse wave, represents the time of the (i+1)th pulse wave trough, and N represents the total number of pulse waves in the time window. In this embodiment, N is 3.

[0054] Based on the pulse wave signal collected by the PPG sensor 27, in an optional solution, taking the beat-to-beat blood oxygen calculation method as an example, the flow chart of the blood oxygen data calculation is as follows: Figure 7 As shown. The red light and infrared light in the pulse wave signal are first low-pass filtered at 5Hz, and then the peaks and troughs of the two signal waveforms are found. The absolute value of the difference between the peak and trough of a certain beat is taken as the AC value of this beat, and the average of all numbers from one peak to the next peak is taken as the DC value of this beat. The blood oxygen saturation of this beat can be obtained using the following formula (3): (3) In the formula, Indicates blood oxygen saturation, represents the red light AC component, Indicates the infrared light AC component, represents the DC component of red light, Represents the DC component of infrared light.

[0055] In another alternative, in order to increase the sampling rate of the blood oxygen saturation signal, the time window method can also be used to calculate the blood oxygen saturation every second to ensure a stable sampling rate (1Hz), so that the speed of obtaining blood oxygen data is independent of the subject's heart rate. Specifically, first set a time window of a specific length (Take 5 seconds as an example), the newly collected photoelectric volume pulse wave is stored every 1 second in this time window, and the photoelectric volume pulse wave of the first 1 second is released. In the data of the time window, the main peak and trough values ​​of the photoelectric volume pulse wave are found, and the absolute value of the difference between the peak and trough of each beat is averaged as the AC value in this time window, and the mean of all numbers is taken as the DC value of the beat in this time window, and then the blood oxygen saturation is calculated according to formula (3).

[0056] The temperature sensor 28 is mainly used to monitor the core body temperature of the human body in real time. In addition, monitoring the body temperature throughout the night can also determine the individual's circadian rhythm and provide the user with circadian rhythm information. The temperature sensor 28 is connected to the internal circuit 25 via a wire. The ionization temperature sensor has high sensitivity, high selectivity and biomimicry characteristics, can detect temperature differences less than 10 millikelvin, and responds to temperature changes independently of mechanical deformation. It is suitable for wearable devices and flexible electronic applications and can be used as a preferred embodiment. It can be understood that in other embodiments, temperature sensors based on other principles such as capacitive and resistive can also be used, and the present invention does not limit this.

[0057] like Figure 7As shown, the internal circuit 25 mainly includes MCU251, and functional modules such as a signal processing module 252, a power management module 253, an LED driving module 254, an earphone driving module 255 and a Bluetooth communication module 257 electrically connected to the MCU251. The EEG electrode, the EOG electrode, the PPG sensor 27 and the temperature sensor 28 are electrically connected to the MCU251 through the signal processing module 252. The reference electrode is directly connected to the reference electrode port of the MCU251 through a wire. Among them, the signal processing module 252 is mainly used for amplification of the input signal and digital-to-analog and analog-to-digital conversion, for example, an ADS1299 chip can be selected. The signals collected by the EEG electrode and the EOG electrode are analog signals, which are converted into digital signals through the ADS1299 chip, that is, digitized to become digital signals that can be transmitted. Since the pulse wave signal sent by the PPG sensor 27 is a digital signal, in order to realize the data transmission between the PPG sensor 27 and the MCU251, the signal processing module 252 provides a dedicated digital-to-analog conversion circuit for the PPG sensor 27, converts the pulse wave signal collected by the PPG sensor 27 into an analog voltage first, and then inputs it into the ADS1299 chip for digitization, thereby realizing the unified digitization of all signals. The power management module 253 can use a 3.7V-1000mAh lithium battery, which is connected to the MCU251, the LED driver module 254, the earphone driver module 255 and the refrigeration driver module 256 via a wire, which can ensure the signal collection and management throughout the night. The LED driver module 254, the earphone driver module 255, and the refrigeration driver module 256 are electrically connected to the LED light bar 24, the bone conduction earphone and the semiconductor refrigeration module 26 respectively. The Bluetooth communication module 257 is connected to the host computer for communication. In a preferred embodiment, the internal circuit 25 can be designed based on a flexible circuit board to better adapt and be installed inside the eye mask body 2.

[0058] Bone conduction headphones can be directly selected from commercial bone conduction headphones. The bone conduction headphones are connected to the MCU251 of the internal circuit 25 via the headphone driver module 255. When the headphones are running, the MCU251 first turns on the headphone driver module 255 and sends an electrical signal carrying audio information to the headphone driver module 255, which converts the electrical signal into a vibration signal, which is finally converted into audio through the auditory system of the human ear.

[0059] The total length of the eye mask shell 22 is 18 cm, and the width is 8 cm. A groove for installing the internal circuit 25 and the LED light bar 24 is provided inside the eye mask shell 22. The internal circuit 25 and the LED light bar 24 are installed in the groove. The LED light bar 24 can be powered by 5V DC, or by a PWM wave with an amplitude of 5V to change the brightness. The LED density is 60LEDs / m. After being bent, it is adhered to the inside of the eye mask shell 22 by glue or other sticky materials. After installation, it corresponds to the eye depression of the support sponge 23 and is covered by the eye mask lining 21. The LED light bar 24 is connected to the internal circuit 25 via a wire and is controlled by the internal circuit 25. It can emit 40Hz monochromatic light and has a sleep-aiding effect. Optionally, the edge of the eye mask shell 22 can also be provided with a device switch button 221, a music play / pause button 222, an LED light bar on / off button 223, a brightness increase button 224, a brightness decrease button 225, a cooling on / off button 226 and other function buttons according to needs.

[0060] Further, embodiment 2 of the present invention discloses a wearable non-intrusive smart eye mask system, which includes, in addition to the smart eye mask described in embodiment 1 above, a host computer connected to the smart eye mask. The host computer can be selected from electronic devices such as smart phones, tablet computers, dedicated terminal devices, PCs, smart watches, smart bracelets, etc. These electronic devices are usually equipped with a processor, a memory, a communication module, and a human-computer interaction interface. Among them, the processor mainly performs sleep monitoring, and the memory can store music and collected physiological data. The communication module is mainly used to receive data collected by the smart eye mask and send relevant instructions generated by the processor to the smart eye mask (including turning on or off sleep-aiding audio / sleep-aiding lights, etc.), and has a Bluetooth communication function. On the human-computer interaction interface, there are option buttons such as start signal acquisition, turn on / off lights, turn on / off music, system presets, and temperature control buttons. The user clicks the button to realize the corresponding function. The processor is used to receive data such as EEG signals, electrooculographic signals, heart rate, and blood oxygen saturation sent by the smart eye mask, perform sleep stages and the duration of each stage, and control the LED light bar 24, bone conduction headphones, and conductor cooling module 26 for sleep intervention. It can be understood that one or more computer programs are stored in the memory of the host computer, and when the computer programs are run, they are used to achieve: calculating heart rate and blood oxygen saturation based on photoelectric volume pulse wave signals; real-time sleep monitoring based on EEG signals, EOG signals, heart rate and blood oxygen saturation; and controlling the intervention component to perform sleep intervention when the sleep monitoring results meet the preset conditions. Among them, sleep monitoring includes sleep staging and current sleep stage duration statistics, and accordingly, the sleep monitoring results include the current sleep stage and its duration. The sleep stages include five stages: W, N1, N2, N3 and REM, so each sleep stage can be accurately distinguished.

[0061] Further, embodiment 3 of the present invention discloses a sleep monitoring method, which can be applied to the wearable non-intrusive smart eye mask system described in embodiment 2. The sleep monitoring method mainly includes three stages, namely, the preprocessing stage, the sleep staging stage and the sleep duration statistics stage. According to the international sleep staging standard, 30s of sleep time is selected as a cycle for classification, and the sampling rates of the EEG signal, the EOG signal and the core body temperature signal are set to 100Hz, and the sampling rate of the PPG signal is 40Hz. In the preprocessing stage, the host computer first performs a 0.3-35Hz bandpass filter on the EEG signal uploaded by the smart eye mask, and performs a 0-15Hz low-pass filter on the EOG signal, and normalizes them respectively; the photoelectric volume pulse wave signal uploaded by the smart eye mask is low-pass filtered at 0-5Hz, and then the heart rate and blood oxygen saturation are calculated according to the preset algorithm. It can be understood that if the heart rate and blood oxygen saturation have been calculated by the smart eye mask, the calculated heart rate and blood oxygen saturation data can be directly uploaded to the host computer.

[0062] In the sleep staging stage, the host computer uses a pre-trained deep learning algorithm to achieve accurate and real-time sleep staging. Figure 8 As shown in the figure, the network architecture of the sleep staging model based on deep learning algorithm is mainly divided into three parts: the first part is the feature extraction and fusion module, which includes four feature extraction branches, which are used to extract the features of four signals, namely, EEG signals, EOG signals, heart rate and blood oxygen saturation, and fuse these four signals; the second part is the feature enhancement module, which amplifies and filters the fused features; the third part is a classifier, which predicts the sleep stage through features.

[0063] Among them, the feature extraction and fusion module has four feature extraction branches, mainly including the series-connected convolution layer, Dropout layer, Reshape layer and Concate layer. In each branch, the signal passes through three series-connected convolution layers respectively. In each convolution layer, the high-dimensional eigenvalue is first calculated through a one-dimensional convolution operation (Conv1D); then the eigenvalue is transformed to a standard normal distribution with a mean of 0 and a variance of 1 by batch normalization (BN); then ReLU is used as the activation function to perform nonlinearization; then the local maximum value of the feature is extracted through maximum sampling (MaxPool) to reduce the size of the feature; finally, the Dropout layer is used to randomly discard a part of the eigenvalue to prevent overfitting problems. Due to the different sizes of the input signal and the parameters of the convolution layer, the sizes of the features extracted by the four feature extraction branches are also different. The Reshape layer maps the features extracted by different feature extraction branches to the same size, and combines the features through the Concate layer to obtain the feature array Feature1.

[0064] Among them, the feature enhancement module mainly includes a convolution layer, an expansion layer (Flatten) and a than activation function. The feature array Feature1 passes through the convolution layer in the feature enhancement module, and the convolution layer also includes a one-dimensional convolution operation (Conv1D) to further increase the feature dimension of Feature1; then the convolution operation Feature1 is reduced to a one-dimensional array through Flatten; then the value range of the one-dimensional array is scaled to the interval of -1 to 1 by the than activation function, so as to obtain the weight value of each feature; then the feature array Feature1 is multiplied by the weight value to obtain the enhanced feature Feature2, so as to realize the amplification and filtering of different features and reflect the difference in importance between features; finally, the enhanced feature Feature2 is input into the classifier in the third part.

[0065] Among them, the classifier mainly includes an expansion layer (Flatten), two series-connected fully connected layers (FC+ReLU) with ReLU activation function, and a Softmax layer. The enhanced feature Feature2 expands the multi-dimensional array into a one-dimensional array through Flatten; the one-dimensional array is sequentially mapped to a classification array through two fully connected layers with ReLU activation function, that is, an array of size (1, 5); finally, the Softmax layer scales the numerical range of the classification array to 0-1, which is used to represent the classification probability of the five sleep stages (including W stage, N1 stage, N2 stage, N3 stage and REM stage). The sleep stage corresponding to the maximum probability is the prediction result of the sleep stage. The above sleep staging method uses convolution operation, which is fast and efficient, and can accurately and real-time output the monitoring results of sleep stages.

[0066] Based on the above, it can be seen that in the sleep staging stage, the present invention uses four signal inputs (electroencephalogram signal, electrooculogram signal, heart rate and blood oxygen saturation) for sleep staging, which can reduce the impact of various interferences (such as improper electrode position, noise, etc.) on a single signal and greatly reduce the staging error. In other words, multimodal signals can complement each other, reduce the impact of a certain signal error or noise on the entire staging system, thereby improving the robustness and accuracy of sleep staging.

[0067] It is worth noting that the present invention uses a sleep staging model based on a deep learning algorithm. The model network structure design realizes high-precision and real-time sleep staging through multimodal feature extraction, fusion and enhancement, as well as efficient classifiers. In the sleep staging model, the feature extraction and fusion module uses four independent feature extraction branches to extract the features of EEG, EOG, heart rate and blood oxygen signals respectively, optimizes feature expression through convolution layer (Conv1D), batch normalization (BN), ReLU activation function and maximum sampling (MaxPool), and combines Dropout to prevent overfitting; the feature enhancement module dynamically allocates feature weights through convolution layer and tanh activation function, amplifies key features (such as slow wave activity in stage N3), and filters redundant information; the classifier uses a fully connected layer (FC) and a Softmax layer to map features to five sleep stages and output probability distribution. In particular, the high efficiency and parallel computing power of the convolution operation enable the model to process data in real time and meet the low latency requirements of wearable devices. The complementarity of multimodal signals (such as EEG slow waves and heart rate variability) and the feature enhancement mechanism significantly improved the recognition accuracy of the N3 stage. At the same time, Dropout and batch normalization enhanced the generalization ability of the model and adapted to individual differences and noise interference.

[0068] During the sleep duration statistics stage, the duration of the current sleep stage can be counted in real time based on the sleep staging results. Specifically, if the sleep stage remains unchanged, the duration of the sleep stage increases by one sleep cycle (30s), and the duration of other sleep stages is 0; if the sleep stage changes, the duration of the previous sleep stage is reset to zero, and the duration of the current sleep stage is modified to one sleep cycle. For example, if the current sleep stage is stage W and lasts for 150 s, and the monitoring result shows stage N1 at this time, the state is modified to N1 and has been maintained for 30s; conversely, if the predicted result is still stage W, the state is still stage W, and the time is maintained for 180s. By counting the duration of specific stages N3 and W, the control of sleep intervention duration can be achieved.

[0069] In addition, in order to evaluate whether the sleep quality of a certain stage has improved, in addition to paying attention to the duration of the current sleep stage, the total duration of all stages can also be counted. In an optional embodiment, the state is first initialized: a dictionary is created to record the cumulative duration of each sleep stage (W, N1, N2, N3, REM), and the duration of all stages is zero at the beginning. At the same time, the current sleep stage and the duration of the current stage are defined. Every time there is a new sleep stage monitoring result, it is processed according to whether the stage changes: if the current stage is the same as the previous stage, the duration of the stage is increased by the duration of a sleep cycle (30 seconds); if the current stage is different from the previous stage, the duration of the previous stage is recorded in the total duration of the corresponding stage, and the duration of the current stage is reset to a new cycle duration (30 seconds), and the current sleep stage is updated to a new stage. At each stage change, the duration of the current stage is promptly added to the total duration of the corresponding stage to ensure that the duration of each stage is accurate in the end. Finally, at the end of the entire sleep cycle (or at the end of the stage change), the total duration of each sleep stage is returned, that is, the total duration of each stage is the cumulative duration from the beginning of sleep to the end of sleep, in seconds. The changes in the statistical duration data collected based on each sleep stage over multiple days can be used to analyze the wearer's sleep structure characteristics, evaluate their sleep quality and improvement, etc., and can also be used as data for the diagnosis and treatment of insomnia patients.

[0070] Under the condition of known sleep stages, doctors can judge the patient's sleep structure and whether there is insomnia according to the patient's sleep state, and intervene and treat insomnia. In addition to drugs, there are many physical methods without side effects in the existing insomnia intervention methods, currently including light stimulation, sound stimulation, transcranial electrical stimulation, thermal stimulation, transcranial magnetic stimulation, etc. Among them, the transcranial magnetic stimulation scheme cannot be miniaturized, and the design required for transcranial electrical stimulation is more complicated. The principle mechanism of 40Hz light stimulation and sound stimulation is clear, and it is easy to realize wearability, which is the optimal solution for closed-loop sleep management. In addition, there are currently some sound stimulation schemes that play synchronized pink noise pulses during slow-wave sleep to significantly increase the time and quality of deep sleep. Forehead cooling stimulation has also been shown to promote the subject to enter a sleep state and increase sleep time. Based on this, the present invention can also organically combine sleep monitoring with sleep intervention, conduct targeted intervention according to the stage of sleep, form a closed-loop sleep management, and can be personalized according to needs.

[0071] Furthermore, Example 4 of the present invention discloses a sleep intervention method based on the sleep staging results of implementation 3. Before using the wearable non-intrusive smart eye mask system, the user wears it for the first time, determines the initial position of each electrode as much as possible according to the standard, and changes the electrode position through Velcro. After wearing it, the user can first preset the function on the host computer, including whether to turn on the pre-sleep audio and / or light and / or temperature control intervention means, whether to turn on deep sleep regulation, and audio selection, etc. It can be understood that after the above settings are completed, the parameters can be adjusted on the host computer at any time according to needs.

[0072] Usually, the intervention duration of sleep-aiding audio, lighting and temperature control ranges from 10 to 30 minutes to ensure that the intervention process is both effective and does not disturb the normal sleep cycle. Whether to perform deep sleep regulation can also be flexibly selected according to user needs, and the duration of deep sleep regulation is determined by the user's actual deep sleep time. When the user chooses to turn on deep sleep regulation, the host computer sends relevant instructions to the smart eye mask through the communication module, which is received by the Bluetooth communication module 257 in the internal circuit 25 and transmitted to MCU251. MCU251 modifies the internal logic and executes the regulation of the light PWM wave, the regulation of the PWM wave of the cooling drive module, and the regulation of the headphone driver module 255. Furthermore, it can also be set in the initial state that the sleep-aiding mode is automatically turned on by default after the user reaches the insomnia standard; the deep sleep regulation mode is automatically turned on after entering the deep sleep stage.

[0073] When the wearable non-disturbing smart eye mask system is used, when the user presses the device switch button 221 on the smart eye mask, the system starts running. First, the host computer and the user predetermine the intervention information, including the selection of sleep-aiding intervention means, whether to turn on deep sleep regulation, and the selection of sleep-aiding and deep sleep audio. After the intervention information is determined, MCU251 starts signal acquisition and data transmission, and the host computer starts real-time sleep staging. When the system monitors the current stage W (continuous awake state) through sleep monitoring and the user's stage W is counted to be more than 20 minutes, it is judged that the insomnia standard is met. The system will automatically trigger the selected sleep-aiding intervention means according to the preset intervention information to help the user fall asleep faster. The intervention means mainly include 40Hz stroboscopic lighting (realized by controlling the LED light bar 24 through the internal circuit 25), forehead cooling (realized by controlling the semiconductor refrigeration module 26 through the internal circuit 25) and sleep-aiding audio (realized by controlling the bone conduction earphones 12 and bone conduction earphones 13 through the internal circuit 25). Among them, 40Hz stroboscopic lighting simulates light pulses of a specific frequency, which may promote sleep by regulating neural activity; sleep-aiding audio includes white noise, natural sounds (such as rain and waves), meditation music, ASMR, etc., to relax the body and mind and promote sleep; the semiconductor refrigeration module 26 can be combined with the temperature sensor 28 to form a closed loop of body temperature detection and control.

[0074] Combination Fig.10 and Fig.11As shown, when the host computer detects that the current state is W (continuous awake state) and counts that the user's W state exceeds 20 minutes, it sends a sleep-aid intervention instruction to the smart eye mask. After receiving the sleep-aid intervention instruction, MCU251 first determines whether to turn on lighting, cooling or sleep-aid audio according to the preset intervention means. If it is necessary to turn on the light, activate the enable state of the PWM wave, and MCU251 sends a 40Hz PWM wave. After passing through the LED drive module 254, its current and voltage are amplified to reach the level of driving the LED light bar 24. During the lighting process, the user can press the brightness increase button 224 or the brightness decrease button 225 on the shell to adjust the duty cycle of the PWM wave. When the brightness needs to be increased, the duty cycle of the PWM wave increases; when the brightness needs to be reduced, the duty cycle of the PWM wave decreases. When the user presses the LED light bar on / off button 223, if the current duty cycle is 0, it means that the current LED light bar 24 is in the off state, and the duty cycle is set to 30% to turn on the LED light bar; if the current duty cycle is not 0, it means that the current LED light bar is in the on state, and the duty cycle is set to 0% to turn off the LED light bar 24. If the illumination is not turned on, the enabling state of the PWM wave is turned off.

[0075] Similarly, if the sleep-aiding audio is to be turned on, the enabled state of the headphone driver module 254 is activated, and at the same time, an electrical signal carrying audio information is sent to the headphone driver module 254, and the headphone driver module 254 amplifies the current and voltage of the electrical signal carrying audio information to reach the level of driving the bone conduction headphone. If the sleep-aiding audio is not to be turned on, the enabled state of the headphone driver module 254 is turned off.

[0076] It should be noted that if the semiconductor refrigeration module 26 is to be turned on, the PWM wave enable state is activated, and the MCU sends a PWM wave of a specific frequency, which is controlled by the refrigeration drive module 256. The larger the duty cycle of the PWM wave, the stronger the refrigeration effect. Considering the user experience and the impact on sleep, the semiconductor refrigeration module 26 needs to be used in conjunction with the temperature sensor 28. After turning on the PWM wave of the semiconductor refrigeration module 26, the temperature sensor 28 continuously collects body temperature signals. If it is detected that the user's forehead temperature drops by more than 0.5°C, even if the user has not entered the sleep state, the MCU251 will turn off the PWM wave enable of the refrigeration drive module 256 in advance to end the cooling.

[0077] According to the user's choice, when the intervention duration exceeds the user-set duration, all intervention means are turned off. When the host computer detects that the user has entered a sleeping state (the sleep staging result is non-W stage and remains for 3 minutes), a stop sleep-aiding intervention instruction is issued to the smart eye mask. After receiving the instruction, the MCU251 in the smart eye mask turns off the enable state of the PWM wave, turns off the enable state of the headphone driver module, and stops assisting sleep. Studies have shown that pink noise helps consolidate the deep sleep stage and improve sleep quality. Specifically, the audio of pink noise can be designed as a 50ms pulse, and each pulse has a 5ms rising and falling edge to avoid adverse stimulation to sleep, while promoting slow wave activity and prolonging the duration of deep sleep. Therefore, based on the precise staging of sleep in Example 3, sleep intervention can be performed on a specific deep sleep stage.

[0078] In an optional scheme, deep sleep intervention can also be included. If the preset intervention information selects to enable the deep sleep control mode, the system will automatically start the dedicated audio intervention mechanism after detecting that the user has entered the N3 stage (deep sleep stage). Specifically, the host computer first sends a deep sleep intervention start instruction to the smart eye mask, and the MCU251 in the smart eye mask activates the enable state of the headphone driver module 254 after receiving the instruction, and at the same time sends an electrical signal carrying pink noise audio information to the headphone driver module 255, and the headphone driver module 255 converts the electrical signal into a vibration signal, and finally converts it into a specific deep sleep extension audio (pink noise) through the auditory system of the human ear. This process is based on the principle of not interfering with the natural sleep rhythm, and is played at intervals. When the N3 stage is detected to end (usually changed to the N2 stage), the host computer sends a deep sleep intervention stop instruction. After receiving the instruction, the MCU251 turns off the enable state of the PWM wave and stops the deep sleep intervention.

[0079] Finally, it should be noted that although the embodiments of the present invention are described above in conjunction with the accompanying drawings, the present invention is not limited to the above-mentioned specific embodiments and application fields, and the above-mentioned specific embodiments are only illustrative and instructive, rather than restrictive. Under the guidance of this specification, a person of ordinary skill in the art can also make many forms without departing from the scope of protection of the claims of the present invention, all of which belong to the protection of the present invention.

Claims

1. A closed-loop sleep management method, characterized in that: Applied to head-mounted sleep closed-loop management devices, including: Real-time sleep monitoring based on EEG signals, EOG signals, heart rate and blood oxygen saturation; and controlling the intervention component to perform sleep intervention when the sleep monitoring result meets the preset conditions; The sleep monitoring includes sleep staging and current sleep stage duration statistics, and the sleep monitoring results include the current sleep stage and its duration. The sleep stages include W stage, N1 stage, N2 stage, N3 stage and REM stage.

2. The closed-loop sleep management method according to claim 1, characterized in that: Also includes: Acquire EEG signals, EOG signals and photoplethysmogram signals; The heart rate and blood oxygen saturation are calculated according to the photoplethysmography signal.

3. The closed-loop sleep management method according to claim 1, characterized in that: The sleep staging is implemented by a sleep staging convolutional neural network, and the sleep staging convolutional neural network includes: The feature extraction and fusion module is configured as follows: extracting the signal features of EEG signals, EOG signals, heart rate and blood oxygen saturation respectively through four feature extraction branches, and fusing the extracted four signal features; the feature extraction branch optimizes feature expression through convolution layer, batch normalization, ReLU activation function and maximum sampling, and combines Dropout to prevent overfitting; The feature enhancement module is configured to: dynamically assign feature weights to the fused feature array through convolutional layers and tanh activation functions, amplify key features, and filter redundant information; The classifier is configured to map the enhanced feature array to five sleep stages: W, N1, N2, N3, and REM through a fully connected layer and a Softmax layer.

4. The closed-loop sleep management method according to claim 1, characterized in that: The sleep intervention is performed when the preset conditions are met, including: When the current sleep stage is stage W and the statistical duration of stage W exceeds the first preset time, sleep-aiding intervention is performed; and when the current sleep stage changes from stage W to other sleep stages other than stage W and the statistical duration of other sleep stages exceeds the second preset time, sleep-aiding intervention is stopped.

5. The closed-loop sleep management method according to claim 1, characterized in that: The sleep intervention when the preset conditions are met also includes: When the current sleep stage is stage N3, deep sleep intervention is performed; and when the current sleep stage changes from stage N3 to other sleep stages, the deep sleep intervention instruction is stopped.

6. The closed-loop sleep management method according to claim 1, characterized in that: The sleep intervention means includes at least one of light intervention, audio intervention and forehead cooling intervention.

7. The closed-loop sleep management method according to any one of claims 1 to 6, characterized in that: The sleep monitoring also includes counting the total duration of each sleep stage.

8. A head-mounted closed-loop sleep management system, characterized in that: Includes a head-mounted closed-loop sleep management device and a computer program; The head-mounted sleep closed-loop management device includes a head-mounted wearable mechanism, and a signal collection component, an internal circuit and an intervention component arranged on the head-mounted wearable mechanism; The signal acquisition component includes an electrooculogram electrode, an electroencephalogram electrode, a reference electrode, and a PPG sensor; wherein the electrooculogram electrode is used to collect electrooculogram signals; the electroencephalogram electrode is used in conjunction with the reference electrode to collect electroencephalogram signals; and the PPG sensor is used to obtain human photoelectric volumetric pulse wave signals; The internal circuit is used to pre-process the signal collected by the signal collection component, and control the intervention component to perform sleep intervention according to the received intervention instruction; The computer program has one or more programs stored in an internal circuit and / or an electronic device that is communicatively connected to the head-mounted closed-loop sleep management device; When the computer program is executed, it is used to implement the closed-loop sleep management method according to any one of claims 1 to 7.

9. The head-mounted closed-loop sleep management system according to claim 8, characterized in that: The electrooculographic electrodes have a pair, which are respectively located below the outer canthus of the left eye and above the outer canthus of the right eye of the human body when the head-mounted wearable mechanism is worn; The EEG electrodes have a pair, which are respectively located on the left and right sides of the frontal lobe of the human body when the head-mounted wearable mechanism is worn; The reference electrodes have a pair, and when the head-mounted wearable structure is worn, they respectively contact the mastoids at the junction of the left and right external auricles and cheeks of the human body.

10. The head-mounted closed-loop sleep management system according to claim 9, characterized in that: When the head-mounted wearable mechanism is worn, a pair of EEG electrodes are arranged 4-6 cm to the left and right of the inner midline of the eye mask and 0.5-2.5 cm from the upper edge; a pair of electrooculographic electrodes are arranged 6-8 cm to the left and right and 0.5-2.5 cm from the lower edge and 4-6 cm from the upper edge.

11. The head-mounted closed-loop sleep management system according to claim 9, characterized in that: The EOG and EEG electrodes are detachably connected to the head-mounted wearable device via a flexible Velcro strip; the Velcro strip includes a velvet surface and a hook surface, the velvet surface is fixed to the inner side of the head-mounted wearable device and has an area larger than the hook surface, and the EOG and EEG electrodes are bonded to the velvet surface area via the hook surface.

12. The head-mounted closed-loop sleep management system according to any one of claims 8 to 11, characterized in that: The internal circuit includes an MCU, and a signal processing module, a first communication module, an intervention drive module and a power supply module electrically connected to the MCU; The signal acquisition module is used to pre-process the electrooculogram, electroencephalogram and human photoelectric volume pulse wave signals input into the internal circuit, and the pre-processing includes signal amplification, digital-to-analog conversion and analog-to-digital conversion; The first communication module is used to realize communication connection between the MCU and the electronic device; The intervention driving module is electrically connected to the intervention component.

13. The head-mounted closed-loop sleep management system according to claim 12, characterized in that: The intervention component includes at least one of a stroboscopic lighting module and a bone conduction headset; The intervention driving module includes at least one of a light driving module electrically connected to the stroboscopic lighting module and an earphone driving module electrically connected to the bone conduction earphone.

14. The head-mounted closed-loop sleep management system according to claim 13, characterized in that: The intervention component also includes a semiconductor refrigeration module; The intervention drive module includes a refrigeration drive module electrically connected to the semiconductor refrigeration module; The signal acquisition component also includes a temperature sensor for acquiring human body temperature signals; When the computer program is executed, it is also used to implement: During the operation of the semiconductor refrigeration module, when it is detected that the human body temperature signal is lower than the temperature threshold, the semiconductor refrigeration module is controlled to stop sleep intervention.

15. The head-mounted closed-loop sleep management system according to any one of claims 8 to 11, characterized in that: The wearable mechanism is a flexible structure; the internal circuit is a flexible circuit board; the EOG, EEG and reference electrodes are flexible electrodes, wherein the EOG and EEG electrodes are also constructed as serpentine patterns with stretchable lines; the wires used to connect various electrodes, sensors, intervention components and internal circuits are all constructed as serpentine traces with stretchable lines.

16. The head-mounted closed-loop sleep management system according to any one of claims 8 to 11, characterized in that: The wearable mechanism is an eye mask; the eye mask comprises an eye mask body and an eye mask strap; the electrooculogram, electroencephalogram electrodes and PPG sensor are arranged on the side of the eye mask body close to the skin; the internal circuit is arranged inside the eye mask body; The reference electrode is arranged on the eye mask strap; When the intervention component includes at least one of a stroboscopic lighting module, a bone conduction earphone and a semiconductor refrigeration module, the stroboscopic lighting module and the semiconductor refrigeration module are arranged inside the eye mask body; and the bone conduction earphone is arranged on the eye mask strap.

17. An electronic device, characterized in that: It comprises a processor, a memory and a second communication module; the electronic device is communicatively connected to a head-mounted sleep closed-loop management device via the second communication module; the processor is used to call a computer program stored in the memory to implement the sleep closed-loop management method described in any one of claims 1 to 7.

Images