An intelligent sleep-aiding system and closed-loop control method based on multi-modal collaborative induction

Abstract

This application discloses an intelligent sleep aid system and closed-loop control method based on multimodal collaborative induction, relating to the field of intelligent sleep aid device technology. The system includes an acoustic module, a sensor module, a lighting module, a voice interaction module, and a control module. The acoustic module generates a 30-100 Hz directional low-frequency sound curtain through a speaker array, combined with a bone conduction vibrator and an RGB LED light ring to construct a three-channel stimulation. The closed-loop control method includes three-channel collaborative induction, real-time respiratory closed-loop matching, multi-sensor fusion sleep monitoring, adaptive output adjustment, and night awakening intervention. This invention actively shortens the sleep latency through multi-sensory rhythm coordination and respiratory feedback closed-loop, achieves zero companion interference through the directional sound curtain and bone conduction, and ensures privacy and security through offline voice and physical switches, making it suitable for daily use in family bedrooms.

Description

Technical Field

[0001] This invention relates to the field of intelligent sleep aid technology, specifically to an intelligent sleep aid system and closed-loop control method based on multimodal collaborative induction. Background Technology

[0002] Insomnia and sleep deprivation have become widespread health problems in modern society, with difficulty falling asleep being one of the most prominent clinical manifestations. Currently, sleep aids on the market can be mainly categorized as follows:

[0003] Category 1: White noise / natural sound players. These products mask ambient noise by playing sounds like rain or ocean waves. Their technical drawbacks are: passive masking rather than active induction, resulting in limited effectiveness for users with long sleep latency; the sound travels through the air, easily disturbing bed partners; and the lack of sleep state awareness means that continued playback after the user has fallen asleep can actually disrupt deep sleep.

[0004] Category 2: Intelligent monitoring devices. Some publicly available solutions use bone conduction devices to highlight the low-frequency part of music, but they only use single bone conduction stimulation and lack multi-sensory coordination; heart rate / respiration detection uses contact sensors, resulting in a poor user experience; and the problem of sound leakage interference has not been resolved.

[0005] Category 3: Deep brain modulation devices. Some ultrasound and infrasound coupling modulation schemes fall under the category of medical devices. They are technically complex and costly, making them unsuitable for everyday home use. Furthermore, the long-term safety of using ultrasound and infrasound requires further verification.

[0006] Category 4: Directional sound devices. Some sleep aids create a sense of spatial immersion by adjusting the direction of sound waves, rather than reducing sound interference; and they do not synergize with sleep monitoring or multimodal stimulation.

[0007] In summary, existing technologies share the following common technical deficiencies: limited stimulation dimensions and lack of synergistic induction through multiple sensory channels; insufficient closed-loop feedback, failing to dynamically adjust stimulation parameters based on the user's real-time physiological state; unresolved partner interference issues; and a lack of screenless interaction, with increased blue light exposure from using mobile apps before bedtime. Summary of the Invention

[0008] To address the technical problems of existing technologies, such as limited stimulation dimensions, insufficient closed-loop feedback, interference from bed partners, and inconvenient pre-sleep interaction, the present invention aims to provide an intelligent sleep aid system and closed-loop control method based on multimodal collaborative induction. This system employs a technical architecture of three-channel collaborative induction, real-time breathing matching, and sleep stage adaptation. Through auditory, tactile, and visual three-channel synergistic rhythmic induction, it actively induces brain waves to transition to a sleep state, thereby shortening the average sleep latency, increasing the deep sleep detection rate, and effectively improving the problem of difficulty falling asleep.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] In a first aspect, the present invention provides an intelligent sleep aid system based on multimodal collaborative induction, including a main unit housing, and an acoustic module, a sensor module, a lighting module, a voice interaction module and a control module disposed inside the main unit housing;

[0011] The acoustic module includes a speaker array, a full-range speaker, and a bone conduction vibrator. The speaker array consists of at least four ultralinear micro-speakers, used to generate a directional low-frequency sound curtain through electronic beamforming technology. The full-range speaker is used to play sleep aid audio, meditation music, wake-up sound effects, and nighttime awakening intervention audio. The bone conduction vibrator is connected to an independent external bone conduction pillow pad via a detachable medical-grade silicone sound guide tube. The bone conduction pillow pad is a flat silicone structure with an internal vibrating diaphragm for placing under the user's pillow to transmit low-frequency vibrations.

[0012] The sensor module includes an infrared thermopile sensor, a capacitive contact sensor, and an accelerometer; the infrared thermopile sensor is used for non-contact detection of the user's breathing rate; the capacitive contact sensor is disposed inside the bone conduction pillow and is used to detect the contact state between the user's head and the bone conduction pillow; the accelerometer is used to detect the intensity of the user's body movement and to identify the interactive signal of the user tapping the host.

[0013] The lighting module is an RGB LED light ring surrounding the edge of the host housing. The voice interaction module includes a dual MEMS microphone array and a local voice wake-up chip for offline recognition of preset voice commands. The voice interaction module has an independent physical microphone disable switch. The control module is electrically connected to the acoustic module, sensor module, lighting module, and voice interaction module respectively.

[0014] As a further embodiment of the present invention, the frequency range of the directional low-frequency sound curtain is 30-100 Hz, the horizontal beam angle is within ±15°, and the sidelobe suppression is not less than 12dB; the operating frequency range of the bone conduction oscillator is 20-200 Hz.

[0015] As a further aspect of the present invention, the effective detection distance of the infrared thermopile sensor is no greater than 50 cm.

[0016] As a further embodiment of the present invention, the lighting module includes at least 24 RGB LED beads, with at least 5 levels of adjustable brightness and at least 2 levels of adjustable color temperature.

[0017] As a further aspect of the present invention, the control module is configured to execute the following closed-loop control logic:

[0018] When the system enters sleep aid mode, it simultaneously activates three stimulation channels for coordinated induction:

[0019] The first stimulation channel controls the speaker array to output the directional low-frequency sound curtain;

[0020] The second stimulation channel controls the bone conduction oscillator to play heartbeat sounds or output low-frequency vibrations in a manner that starts at 60 BPM and decreases to 50 BPM.

[0021] The third stimulation channel controls the LED light ring to provide pulsating illumination in a preset respiratory cycle, wherein the preset respiratory cycle is 4 seconds of inhalation and 4-6 seconds of exhalation, and the pulsation frequency is matched to the brain slow oscillation frequency band of 0.1 Hz to 0.25 Hz.

[0022] The control module acquires the user's respiratory rate detected by the infrared thermopile sensor in real time, and dynamically adjusts the output rhythm of the bone conduction oscillator after filtering and smoothing to keep the bone conduction rhythm synchronized with the user's current respiratory rate.

[0023] When the sensor module detects that the user has entered a light sleep state, it controls the second and third stimulation channels to gradually fade out within 3 minutes, while maintaining the continuous operation of the first stimulation channel.

[0024] When the sensor module detects that the user has entered a deep sleep state, it controls all stimulation channels to enter a silent standby state, and at the same time controls the sensor module to switch to a low-power polling mode.

[0025] As a further aspect of the present invention, the output timing of the directional low-frequency sound curtain, the rhythmic output timing of the bone conduction oscillator, and the pulsating illumination timing of the LED light ring are kept synchronized in time, and are output in coordination with a unified common rhythmic cycle to form a rhythmic entrainment effect of multi-sensory channels.

[0026] As a further aspect of the present invention, the control module determines that the user has entered a light sleep state under the following conditions: the capacitive contact sensor detects that the user's head has been in continuous contact with the sensor for more than 5 minutes, the intensity of the continuous motion signal detected by the accelerometer is lower than a first preset threshold, and the respiratory rate detected by the infrared thermopile sensor is more than 15% lower than the user's resting baseline.

[0027] As a further aspect of the present invention, the control module determines that the user has entered a deep sleep state under the following conditions: the capacitive contact sensor detects that the user's head has been in continuous contact with the sensor for more than 10 minutes, the intensity of the motion signal detected by the accelerometer is lower than a second preset threshold, and the fluctuation of the breathing rate detected by the infrared thermopile sensor does not exceed 1 breath / minute within 5 consecutive minutes.

[0028] As a further aspect of the present invention, the control module has a built-in night-wake intervention unit, used for:

[0029] When the accelerometer detects a sudden increase in motion signal intensity that exceeds a third preset threshold, and the infrared thermopile sensor detects that the user's breathing rate increases by more than 30% within 30 seconds, it is determined that the user has woken up at night.

[0030] Automatic activation of night wake-up intervention mode: Controls the full-range speaker to play white noise or natural ambient sound, while controlling the LED light ring to light up warm light at the lowest brightness level and 2200K color temperature;

[0031] The white noise or natural ambient sound gradually fades out in 10-minute cycles, and the LED ring automatically turns off 5 minutes after the audio fades out.

[0032] As a further aspect of the present invention, the control module also has a built-in wake-up unit for:

[0033] Start the wake-up mode 10-30 minutes before the preset wake-up time;

[0034] The LED light ring is controlled to gradually change from dark purple to warm yellow, with the gradation process lasting 10 minutes. The color temperature gradually changes from 1600K to 4000K, and the brightness gradually increases from 10% to 60%.

[0035] Synchronously control the full-range speaker to play a natural wake-up sound effect, with the volume gradually increasing from 10dB SPL to 45dB SPL;

[0036] Upon detecting a user tapping the host or inputting a voice command, the wake-up mode is immediately terminated.

[0037] As a further aspect of the present invention, the voice interaction module supports 16 preset commands for offline recognition, including wake-up words and control commands. The control commands include: turn on the light, turn off the light, brighten it, dim it, warm light, cool light, night light mode, reading mode, meditation mode, sleep aid mode, stop everything, help me think, play it again, longer, shorter. After the command is successfully recognized, the control module controls the LED light ring to output corresponding visual feedback, wherein the feedback for successful command execution is a slow green flash once, the feedback for wake-up response is a fast white flash once, and the feedback for microphone disabling is a fast red flash three times.

[0038] As a further embodiment of the present invention, the physical microphone disable switch is a sliding two-position switch located at the bottom of the main unit housing; when the switch is turned to the off position, the power supply circuit of the dual MEMS microphone array is completely cut off, and the LED light ring flashes red three times as a disable confirmation feedback.

[0039] As a further aspect of the present invention, it also includes an air quality sensor, which is a PM2.5 sensor and / or a carbon dioxide sensor; the control module automatically adjusts the output intensity of the directional low-frequency sound curtain and the output rhythm of the bone conduction oscillator according to the indoor air quality parameters detected by the air quality sensor.

[0040] As a further embodiment of the present invention, the bone conduction pillow pad comprises, from the surface layer to the bottom layer, a skin-friendly silicone layer, a PET vibration diaphragm layer, an air conduction cavity layer, and an EPDM sponge backing layer; the silicone sound tube connects the bone conduction vibrator and the air conduction cavity layer, transmitting vibration to the PET vibration diaphragm layer.

[0041] As a further aspect of the present invention, the control module has a built-in individualized learning unit, configured as follows:

[0042] Local storage of the user's sleep data for the past 30 days, including sleep latency, duration of each sleep stage, number of nighttime awakenings, and records of user-initiated interventions;

[0043] The frequency parameters of the directional low-frequency sound curtain, the initial rhythm parameters of the bone conduction oscillator, and the duration parameters of the night awakening intervention are automatically optimized based on historical sleep data.

[0044] All sleep data and optimization parameters are stored on local storage media and are not uploaded to external network servers.

[0045] As a further aspect of the present invention, the control module is built on the FreeRTOS real-time operating system and includes audio control tasks, sensor fusion tasks, LED lighting effect tasks, voice processing tasks, state machine management tasks, data storage tasks, and power management tasks.

[0046] The state machine management task is used to manage the state transitions between standby mode, sleep aid mode, meditation mode, wake-up mode, and night light mode.

[0047] As a further embodiment of the present invention, the main unit housing is a square rounded corner structure with dimensions of 120mm×120mm×40mm; the main unit housing is provided with four PCB boards, which are arranged from top to bottom as a top layer, a ground layer, a power layer and a bottom layer.

[0048] The main unit also houses a 6000mAh lithium polymer battery and a power management chip, which supports 18W PD fast charging and power path management.

[0049] Secondly, the present invention also provides a closed-loop control method for an intelligent sleep aid system based on multimodal collaborative induction, comprising the following steps:

[0050] S1. Mode Activation and Selection: Users can trigger the system to enter the corresponding working mode through voice commands or physical tapping operations. The working modes include sleep aid mode, meditation mode, night light mode and wake-up mode.

[0051] S2. Multimodal Collaborative Induction: When entering sleep aid mode, the control module simultaneously activates three stimulation channels for output:

[0052] The first channel controls the loudspeaker array to generate a directional low-frequency sound curtain of 30-100 Hz through electronic beamforming, with the horizontal beam angle controlled within ±15°;

[0053] The second channel controls the bone conduction oscillator to play heartbeat sounds or output low-frequency vibrations starting at 60 BPM, and gradually decreases to 50 BPM.

[0054] The third channel controls the RGB LED light ring to provide pulsating illumination with a breathing cycle of 4 seconds of inhalation and 4-6 seconds of exhalation, with the pulsation frequency matching the 0.1 Hz to 0.25 Hz frequency band;

[0055] S3. Real-time closed-loop matching of breathing: The infrared thermopile sensor detects the user's breathing rate in real time with a sampling period of no more than 2 seconds. After smoothing by Kalman filtering, the control module uses the breathing rate as a feedback signal to dynamically adjust the output rhythm of the second channel bone conduction oscillator so that the bone conduction rhythm is synchronized with the user's current breathing rate.

[0056] S4. Multi-sensor fusion sleep stage determination: The control module integrates the detection data from capacitive contact sensors, accelerometers, and infrared thermopile sensors to determine whether the user is currently awake, in light sleep, in deep sleep, or awake at night.

[0057] S5. Sleep stage adaptive adjustment:

[0058] If the sleep state is determined to be light sleep, control the second and third channels to gradually fade out within 3 minutes, while maintaining the continuous operation of the first channel directional sound curtain;

[0059] If the state is determined to be deep sleep, all stimulation channels are controlled to enter silent standby mode, and the sensor module is switched to low-power polling mode.

[0060] If the condition is determined to be a nighttime awakening, initiate the nighttime awakening intervention procedure;

[0061] S6. Night Wake-up Intervention Execution: Automatically control the full-range speaker to play white noise or natural ambient sounds, while controlling the LED light ring to light up the lowest brightness 2200K warm light; the white noise gradually fades out in 10-minute cycles, and the LED light ring automatically turns off 5 minutes after the audio fades out.

[0062] As a further aspect of the present invention, in step S2, the preset breathing cycle includes two modes: inhalation for 4 seconds, exhalation for 4 seconds, and a complete cycle of 8 seconds, corresponding to a frequency of 0.125 Hz; inhalation for 4 seconds, exhalation for 6 seconds, and a complete cycle of 10 seconds, corresponding to a frequency of 0.1 Hz; the control module automatically selects the matching breathing frequency mode based on the user's historical sleep data.

[0063] As a further aspect of the present invention, a personalized learning step is also included:

[0064] The control module locally records the user's daily sleep data, including sleep latency, deep sleep duration, number of nighttime awakenings, and user-manually adjusted data.

[0065] If the sleep latency exceeds 30 minutes for three consecutive days, gradually reduce the center frequency of the bone conduction initiation rhythm and the directional sound curtain.

[0066] If the number of nighttime awakenings exceeds 2 times for 3 consecutive days, the audio fade-out duration of the nighttime awakening intervention will be extended.

[0067] If the user frequently and manually increases the brightness, the default brightness level of the night light mode will be increased.

[0068] All parameter optimization processes are completed locally, and data is not transmitted to external networks.

[0069] As a further aspect of the present invention, a meditation mode control step is also included:

[0070] When a user enters meditation mode via voice command or tapping, the control module shuts down the first channel directional sound curtain and the second channel bone conduction output.

[0071] Control the full-range speaker to play preset meditation audio, including pure alpha wave tones, guided natural sounds, or breath-guided chimes;

[0072] The LED light ring outputs a soft, flowing blue light effect with a brightness of 20% and no flicker.

[0073] When the meditation mode ends, the full-range speaker will play a chime as a cue, and the LED light ring will provide three breathing pulsations as feedback.

[0074] As a further aspect of the present invention, in the sleep aid mode or meditation mode, the control module generates binaural beat audio with a difference frequency range of 0.25 Hz to 8 Hz; the binaural beat is output to the user's left and right ear areas respectively through a bone conduction oscillator and a full-range speaker to form binaural difference frequency stimulation.

[0075] As a further aspect of the present invention, in step S5, the low-power polling mode in the deep sleep state specifically involves: reducing the detection frequency of the infrared thermopile sensor to once every 10 seconds, reducing the detection frequency of the accelerometer to once every 5 seconds, and maintaining the detection frequency of the capacitive contact sensor at once every 2 seconds.

[0076] As a further embodiment of the present invention, the user can manually terminate or switch the current operating mode at any time by tapping the top of the host or by using voice commands.

[0077] The criteria for determining the tapping action are: the accelerometer detects a pulse signal with a duration of no more than 200ms and a peak acceleration of no less than 1.5g on any of the three axes.

[0078] Compared with existing technologies, the present invention provides an intelligent sleep aid system and closed-loop control method based on multimodal collaborative induction, which solves the technical problems of single stimulation, lack of closed-loop feedback, partner interference, and inconvenient interaction in existing technologies, and specifically has the following beneficial effects:

[0079] 1. Significantly shortens sleep latency: Through the rhythmic entrainment effect of the three channels of hearing, touch and vision, it actively induces brain waves to transition to a sleep state, which can shorten the sleep latency by an average of more than 15 minutes, which is superior to sleep aids with single stimulation.

[0080] 2. Achieve extremely low interference to your partner: The directional low-frequency sound curtain confines the sound energy within a narrow beam of ±15°. Combined with the solid-borne sound transmission characteristics of bone conduction, the sound leakage at 1 meter is ≤25dB SPL, which is far lower than traditional white noise equipment and will not affect your partner's rest.

[0081] 3. Closed-loop adaptive adjustment: Real-time matching of breathing rate is used for sleep aid closed-loop control to achieve dynamic synchronization between stimulation parameters and user physiological state. Stimulation parameters are dynamically synchronized with user physiological state. Combined with sleep stage, the output intensity is automatically adjusted to avoid continuous stimulation after falling asleep, which may interfere with deep sleep.

[0082] 4. Screenless interaction to avoid blue light exposure before bed: Fully offline voice recognition and tap-based physical interaction, no need to operate with a mobile app, avoiding blue light exposure before bed; integrated voice-controlled night light function to meet the low-brightness lighting needs when getting up at night.

[0083] 5. Privacy Protection: Local voice processing and local data storage; all sleep data is not uploaded to the cloud; a physical microphone disable switch is configured to cut off microphone power at the hardware level, eliminating privacy concerns.

[0084] These or other aspects of the invention will become more apparent from the following description of embodiments. It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0085] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. In the drawings:

[0086] Figure 1 This is a block diagram of the overall architecture of an intelligent sleep aid system based on multimodal collaborative induction, provided for an embodiment of the present invention.

[0087] Figure 2 The above is an overall flowchart of a closed-loop control method for an intelligent sleep aid system based on multimodal collaborative induction, provided in an embodiment of the present invention.

[0088] Figure 3 The flowchart of the multi-sensor fusion sleep stage determination sub-flowchart in the closed-loop control method of an intelligent sleep aid system based on multimodal collaborative induction provided in an embodiment of the present invention.

[0089] Figure 4 The offline voice interaction recognition subprocess is provided in the closed-loop control method of an intelligent sleep aid system based on multimodal collaborative induction, as an embodiment of the present invention. Detailed Implementation

[0090] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with embodiments. Obviously, the described embodiments are only some embodiments of this invention, not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0091] To address the technical problems of existing technologies, such as limited stimulation dimensions, insufficient closed-loop feedback, interference from bed partners, and inconvenient pre-sleep interaction, this invention provides an intelligent sleep aid system and closed-loop control method based on multimodal collaborative induction. It employs a technical architecture of three-channel collaborative induction, real-time breathing matching, and sleep stage adaptation. Through the coordinated rhythmic induction of auditory, tactile, and visual channels, it actively induces brainwaves to transition to a sleep state, thereby shortening the average sleep latency, increasing the deep sleep detection rate, and effectively improving difficulty falling asleep.

[0092] See Figure 1 As shown, this embodiment of the invention provides an intelligent sleep aid system based on multimodal collaborative induction, including a main unit housing, and an acoustic module, a sensor module, a lighting module, a voice interaction module, and a control module disposed inside the main unit housing.

[0093] The acoustic module includes a speaker array, a full-range speaker, and a bone conduction transducer. The speaker array consists of at least four ultralinear micro-speakers, used to generate a directional low-frequency sound curtain with a frequency range of 30-100 Hz through electronic beamforming technology. The horizontal beam angle is within ±15°, and the sidelobe suppression is not less than 12dB, concentrating sound energy in the user's head area and significantly reducing sound leakage beyond 1 meter. The full-range speaker is used to play sleep aid audio, meditation music, wake-up sound effects, and night awakening intervention audio. The bone conduction transducer operates in a frequency range of 20-200 Hz and is connected to an independently placed external bone conduction pillow pad via a detachable medical-grade silicone sound guide tube. The bone conduction pillow pad is a flat silicone structure with an internal vibrating diaphragm, placed under the user's pillow to transmit low-frequency vibrations, further reducing airborne sound leakage.

[0094] The sensor module includes an infrared thermopile sensor, a capacitive contact sensor, and an accelerometer, used to achieve non-contact physiological detection and interactive sensing. The infrared thermopile sensor extracts respiratory waveforms by detecting periodic changes in nasal temperature, and non-contactly detects the user's respiratory rate. The effective detection distance of the infrared thermopile sensor is no more than 50 cm, with an accuracy of ±1 breaths / minute. The capacitive contact sensor is located inside the bone conduction pillow and is used to detect the contact state between the user's head and the bone conduction pillow. The accelerometer is used to detect the intensity of the user's body movements and to identify interactive signals from the user tapping the host.

[0095] The lighting module is an RGB LED ring surrounding the edge of the host housing. The lighting module contains at least 24 RGB LED beads, has at least 5 levels of adjustable brightness and at least 2 levels of adjustable color temperature, and guides the user's breathing rhythm through pulsating light effects. The voice interaction module includes a dual MEMS microphone array and a local voice wake-up chip for offline recognition of preset voice commands. The voice interaction module has an independent physical microphone disable switch to physically cut off microphone power, ensuring privacy and security. It supports offline recognition of 16 preset commands and can complete mode switching and device control without a network connection. The control module is electrically connected to the acoustic module, sensor module, lighting module, and voice interaction module respectively; wherein, the control module is configured to execute the following closed-loop control logic:

[0096] When the system enters sleep aid mode, it simultaneously activates three stimulation channels for coordinated induction:

[0097] The first stimulation channel controls the speaker array to output the directional low-frequency sound curtain;

[0098] The second stimulation channel controls the bone conduction oscillator to play heartbeat sounds or output low-frequency vibrations in a manner that starts at 60 BPM and decreases to 50 BPM.

[0099] The third stimulation channel controls the LED light ring to provide pulsating illumination in a preset respiratory cycle, wherein the preset respiratory cycle is 4 seconds of inhalation and 4-6 seconds of exhalation, and the pulsation frequency is matched to the brain slow oscillation frequency band of 0.1 Hz to 0.25 Hz.

[0100] The control module acquires the user's respiratory rate detected by the infrared thermopile sensor in real time, and dynamically adjusts the output rhythm of the bone conduction oscillator after filtering and smoothing to keep the bone conduction rhythm synchronized with the user's current respiratory rate.

[0101] When the sensor module detects that the user has entered a light sleep state, it controls the second and third stimulation channels to gradually fade out within 3 minutes, while maintaining the continuous operation of the first stimulation channel.

[0102] When the sensor module detects that the user has entered a deep sleep state, it controls all stimulation channels to enter a silent standby state, and at the same time controls the sensor module to switch to a low-power polling mode.

[0103] In this embodiment, the output timing of the directional low-frequency sound curtain, the rhythmic output timing of the bone conduction oscillator, and the pulsating illumination timing of the LED ring are synchronized in time, and are output in coordination with a unified common rhythmic cycle to form a rhythmic entrainment effect of multi-sensory channels.

[0104] In this embodiment, the control module determines that the user has entered a light sleep state as follows: the capacitive contact sensor detects that the user's head has been in continuous contact with the sensor for more than 5 minutes, the intensity of the continuous motion signal detected by the accelerometer is lower than a first preset threshold, and the breathing rate detected by the infrared thermopile sensor is more than 15% lower than the user's resting baseline.

[0105] In this embodiment, the control module determines that the user has entered a deep sleep state under the following conditions: the capacitive contact sensor detects that the user's head has been in continuous contact with the sensor for more than 10 minutes, the intensity of the motion signal detected by the accelerometer is lower than the second preset threshold, and the fluctuation of the breathing rate detected by the infrared thermopile sensor does not exceed 1 breath / minute within 5 minutes.

[0106] In this embodiment, the control module has a built-in night-wake intervention unit, used for:

[0107] When the accelerometer detects a sudden increase in motion signal intensity exceeding a third preset threshold, and the infrared thermopile sensor detects that the user's breathing rate increases by more than 30% within 30 seconds, it is determined that the user has woken up at night; wherein, the third preset threshold is a normalized motion intensity of 50.

[0108] Automatic activation of night wake-up intervention mode: Controls the full-range speaker to play white noise or natural ambient sound, while controlling the LED light ring to light up warm light at the lowest brightness level and 2200K color temperature;

[0109] The white noise or natural ambient sound gradually fades out in 10-minute cycles, and the LED ring automatically turns off 5 minutes after the audio fades out.

[0110] In this embodiment, the control module also has a built-in wake-up unit, used for:

[0111] Start the wake-up mode 10-30 minutes before the preset wake-up time;

[0112] The LED light ring is controlled to gradually change from dark purple to warm yellow, with the gradation process lasting 10 minutes. The color temperature gradually changes from 1600K to 4000K, and the brightness gradually increases from 10% to 60%.

[0113] Synchronously control the full-range speaker to play a natural wake-up sound effect, with the volume gradually increasing from 10dB SPL to 45dB SPL;

[0114] Upon detecting a user tapping the host or inputting a voice command, the wake-up mode is immediately terminated.

[0115] In this embodiment, the voice interaction module supports 16 preset commands for offline recognition, including wake-up words and control commands. The control commands include: turn on the light, turn off the light, brighten it, dim it, warm light, cool light, night light mode, reading mode, meditation mode, sleep aid mode, stop everything, help me think, play it again, longer, shorter. After the command is successfully recognized, the control module controls the LED light ring to output corresponding visual feedback, where the feedback for successful command execution is a slow green flash once, the feedback for wake-up response is a fast white flash once, and the feedback for microphone disabling is a fast red flash three times.

[0116] In this embodiment, the physical microphone disable switch is a sliding two-position switch located at the bottom of the main unit housing; when the switch is turned to the off position, the power supply circuit of the dual MEMS microphone array is completely cut off, and the LED light ring flashes red three times as a disable confirmation feedback.

[0117] In this embodiment, an air quality sensor is also included, which is a PM2.5 sensor and / or a carbon dioxide sensor; the control module automatically adjusts the output intensity of the directional low-frequency sound curtain and the output rhythm of the bone conduction oscillator according to the indoor air quality parameters detected by the air quality sensor.

[0118] In this embodiment, the bone conduction pillow pad includes, from the surface to the bottom, a skin-friendly silicone layer, a PET vibration diaphragm layer, an air-conducting cavity layer, and an EPDM sponge backing layer; the silicone sound-conducting tube connects the bone conduction vibrator and the air-conducting cavity layer, transmitting vibration to the PET vibration diaphragm layer.

[0119] In this embodiment, the control module has a built-in individualized learning unit, which is configured as follows:

[0120] Local storage of the user's sleep data for the past 30 days, including sleep latency, duration of each sleep stage, number of nighttime awakenings, and records of user-initiated interventions;

[0121] The frequency parameters of the directional low-frequency sound curtain, the initial rhythm parameters of the bone conduction oscillator, and the duration parameters of the night awakening intervention are automatically optimized based on historical sleep data.

[0122] All sleep data and optimization parameters are stored on local storage media and are not uploaded to external network servers.

[0123] In this embodiment, the control module is built on the FreeRTOS real-time operating system and includes audio control tasks, sensor fusion tasks, LED lighting effect tasks, voice processing tasks, state machine management tasks, data storage tasks, and power management tasks.

[0124] The state machine management task is used to manage the state transitions between standby mode, sleep aid mode, meditation mode, wake-up mode, and night light mode.

[0125] In this embodiment, the main unit housing is a square rounded corner structure with dimensions of 120mm×120mm×40mm; the main unit housing has four PCB layers inside, which are, from top to bottom, a top layer, a ground layer, a power layer and a bottom layer;

[0126] The main unit also houses a 6000mAh lithium polymer battery and a power management chip, which supports 18W PD fast charging and power path management.

[0127] This embodiment details the hardware composition and structural design of an intelligent sleep aid system. In this embodiment, the main body of an intelligent sleep aid system based on multimodal collaborative induction is a square, rounded-corner main unit housing with dimensions of 120mm × 120mm × 40mm. The housing is made of PC+ABS material and is available in three colors: Moon Rock Gray, Cloud White, and Night Black. From top to bottom, the main unit housing consists of an acoustic fabric top cover, a semi-transparent PC light guide cover, an LED light ring flexible PCB, a middle frame, a main PCB board, a battery, a speaker assembly, a bottom shell, and silicone anti-slip pads.

[0128] The main unit employs a four-layer PCB rigid board design with a stacked structure of TOP-GND-PWR-BOTTOM. Audio trace impedance is controlled within 8Ω±10%, I2S trace length error is less than 5mm, and the distance from the crystal oscillator to the MCU is less than 15mm, ensuring signal integrity. ESD protection devices and common-mode chokes are installed at the power input and interface to meet EMC design requirements. The control module is based on an STM32U5A9 main control MCU with a Cortex-M33 core, a main frequency of 160MHz, and built-in 2MB Flash and 786KB RAM. It is paired with an ADAU1467 audio DSP responsible for beamforming algorithms and audio signal processing; and features 32MB of local Flash storage for preset audio tracks and sleep data storage.

[0129] The power management module uses the TI BQ25601 PMIC chip, supporting 18W PD fast charging and power path management, and is equipped with a 6000mAh lithium polymer battery; continuous working time in sleep aid mode is ≥8 hours, and low-power standby time is ≥30 days. The power system is divided into multiple independent power supplies: VCC_SYS, VCC_MCU, VCC_AUDIO, VCC_LED, VCC_SENSOR, and VCC_VOICE. The voice module power supply is independently controlled by a physical switch, enabling hardware-level microphone disabling. The acoustic module includes four 15mm ultra-linear miniature titanium diaphragm speakers arranged in a square array with an 80mm spacing, paired with a four-channel MAX98357 audio amplifier to achieve electronic beamforming; it also features a 40mm wool cone full-range speaker, paired with a TPA3132 amplifier, responsible for full-range audio playback; and a custom bone conduction vibrator, paired with a DRV2700 piezoelectric driver chip, connected to an external pillow pad via a 2-meter-long medical-grade silicone sound tube.

[0130] In this embodiment, the sensor module includes a Melexis MLX90614 infrared thermopile sensor, a custom capacitive contact electrode, and an ST LIS3DH triaxial accelerometer; the high-end version additionally integrates PM2.5 and CO2 air quality sensors to achieve environmental parameter linkage. The voice interaction module uses two Knowles SPH1642 MEMS microphones to form a dual-microphone array, paired with a VIT SN-2002 local voice wake-up chip, supporting offline voice recognition with standby power consumption of less than 10mW.

[0131] This embodiment details the design of the directional sound curtain, bone conduction pillow, and audio materials. In the directional low-frequency sound curtain, four ultra-linear miniature speakers are arranged at the four corners of a square, and a beamforming algorithm is executed via an audio DSP to target the desired beam angle. Calculate the first Signal delay of road loudspeakers ,in The speed of sound (343 m / s). For the first The distance from the loudspeaker to the array reference point is determined; the delay is converted into the number of sampling points at a 48kHz sampling rate, and the delay and gain of each channel are adjusted using an FIR filter. The final performance specifications are: operating frequency 30-100Hz, -6dB beamwidth ±15°, sidelobe suppression ≥12dB, maximum sound pressure level of 65dB SPL at 0.5m, directivity factor Q≥8; sound leakage ≤25dB SPL at 1m, achieving extremely low interference to bed partners.

[0132] The bone conduction pillow uses a layered composite structure, consisting of the following layers from top to bottom: a 0.5mm thick skin-friendly silicone surface layer, a 0.1mm thick PET vibrating diaphragm, a 2mm thick air-conducting cavity, and a 3mm thick EPDM sponge backing layer. A silicone sound-conducting tube connects the bone conduction vibrator at the main unit to the air-conducting cavity within the pillow, transmitting vibrations to the PET diaphragm and then through the pillow to the user's head. The pillow's performance specifications are: frequency response 30-200 Hz ± 3dB, maximum vibration acceleration 0.2g, and sound leakage ≤30dB at 1 meter; the vibration acceleration is far below the harmful threshold for the human body, ensuring safe use.

[0133] The system has multiple pre-installed audio materials: 6 60-minute sleep aid audio tracks (FLAC format, 44.1k Hz / 16bit), 5 30-minute meditation music tracks (FLAC format, 44.1k Hz / 16bit), 4 10-15 minute guided audio tracks (MP3 format, 32k Hz / 128kbps), 3 30-minute binaural beat audio tracks (FLAC format, 48k Hz / 24bit), and 10 short sound effects (WAV format, 44.1k Hz / 16bit), covering all scenarios of sleep aid, meditation, and interactive feedback.

[0134] See Figure 2 and Figure 3 As shown, the present invention also provides a closed-loop control method for an intelligent sleep aid system based on multimodal collaborative induction, comprising the following steps:

[0135] S1. Mode Activation and Selection: Users can trigger the system to enter the corresponding working mode through voice commands or physical tapping operations. The working modes include sleep aid mode, meditation mode, night light mode and wake-up mode.

[0136] S2. Multimodal Collaborative Induction: When entering sleep aid mode, the control module simultaneously activates three stimulation channels for output:

[0137] The first channel controls the loudspeaker array to generate a directional low-frequency sound curtain of 30-100 Hz through electronic beamforming, with the horizontal beam angle controlled within ±15°;

[0138] The second channel controls the bone conduction oscillator to play heartbeat sounds or output low-frequency vibrations starting at 60 BPM, and gradually decreases to 50 BPM.

[0139] The third channel controls the RGB LED light ring to provide pulsating illumination with a breathing cycle of 4 seconds of inhalation and 4-6 seconds of exhalation, with the pulsation frequency matching the 0.1 Hz to 0.25 Hz frequency band;

[0140] S3. Real-time closed-loop matching of breathing: The infrared thermopile sensor detects the user's breathing rate in real time with a sampling period of no more than 2 seconds. After smoothing by Kalman filtering, the control module uses the breathing rate as a feedback signal to dynamically adjust the output rhythm of the second channel bone conduction oscillator so that the bone conduction rhythm is synchronized with the user's current breathing rate.

[0141] S4. Multi-sensor fusion sleep stage determination: The control module integrates the detection data from capacitive contact sensors, accelerometers, and infrared thermopile sensors to determine whether the user is currently awake, in light sleep, in deep sleep, or awake at night.

[0142] S5. Sleep stage adaptive adjustment:

[0143] If the sleep state is determined to be light sleep, control the second and third channels to gradually fade out within 3 minutes, while maintaining the continuous operation of the first channel directional sound curtain;

[0144] If the state is determined to be deep sleep, all stimulation channels are controlled to enter silent standby mode, and the sensor module is switched to low-power polling mode.

[0145] If the condition is determined to be a nighttime awakening, initiate the nighttime awakening intervention procedure;

[0146] S6. Night Wake-up Intervention Execution: Automatically control the full-range speaker to play white noise or natural ambient sounds, while controlling the LED light ring to light up the lowest brightness 2200K warm light; the white noise gradually fades out in 10-minute cycles, and the LED light ring automatically turns off 5 minutes after the audio fades out.

[0147] In this embodiment, after the nighttime awakening intervention in step S6 ends, the process returns to step S4 to continuously monitor the user's sleep status, forming a complete control closed loop.

[0148] In this embodiment, in step S2, the preset breathing cycle includes two modes: 4 seconds of inhalation and 4 seconds of exhalation, with a complete cycle of 8 seconds, corresponding to a frequency of 0.125 Hz; and 4 seconds of inhalation and 6 seconds of exhalation, with a complete cycle of 10 seconds, corresponding to a frequency of 0.1 Hz. The control module automatically selects the matching breathing frequency mode based on the user's historical sleep data.

[0149] This embodiment also includes a personalized learning step:

[0150] The control module locally records the user's daily sleep data, including sleep latency, deep sleep duration, number of nighttime awakenings, and user-manually adjusted data.

[0151] If the sleep latency exceeds 30 minutes for three consecutive days, gradually reduce the center frequency of the bone conduction initiation rhythm and the directional sound curtain.

[0152] If the number of nighttime awakenings exceeds 2 times for 3 consecutive days, the audio fade-out duration of the nighttime awakening intervention will be extended.

[0153] If the user frequently and manually increases the brightness, the default brightness level of the night light mode will be increased.

[0154] All parameter optimization processes are completed locally, and data is not transmitted to external networks.

[0155] This embodiment also includes a meditation mode control step:

[0156] When a user enters meditation mode via voice command or tapping, the control module shuts down the first channel directional sound curtain and the second channel bone conduction output.

[0157] Control the full-range speaker to play preset meditation audio, including alpha wave pure tones, natural voice guided words, or breath-guided bell sounds;

[0158] The LED light ring outputs a soft, flowing blue light effect with a brightness of 20% and no flicker.

[0159] When the meditation mode ends, the full-range speaker will play a chime as a cue, and the LED light ring will provide three breathing pulsations as feedback.

[0160] In this embodiment, in sleep aid mode or meditation mode, the control module generates binaural beat audio with a difference frequency range of 0.25 Hz to 8 Hz. The binaural beat is output to the user's left and right ear areas through bone conduction oscillators and full-range speakers respectively, forming binaural difference frequency stimulation.

[0161] In this embodiment, in step S5, the low-power polling mode in deep sleep state is specifically as follows: the detection frequency of the infrared thermopile sensor is reduced to once every 10 seconds, the detection frequency of the accelerometer is reduced to once every 5 seconds, and the detection frequency of the capacitive contact sensor is maintained once every 2 seconds.

[0162] In this embodiment, the user can manually terminate or switch the current operating mode at any time by tapping the top of the host or by using voice commands.

[0163] The criteria for determining the tapping action are: the accelerometer detects a pulse signal with a duration of no more than 200ms and a peak acceleration of no less than 1.5g on any of the three axes.

[0164] This embodiment uses the principle of respiratory detection and sleep stage determination through multi-sensor fusion as the sensor module and sleep monitoring implementation steps. The specific steps are as follows:

[0165] (1) The infrared thermopile sensor collects the user's nasal temperature data at a sampling rate of 10 Hz, and extracts the respiratory rate through the following steps:

[0166] Bandpass filtering: A 0.1-0.8 Hz bandpass filter is used to remove high-frequency noise and DC drift, corresponding to a breathing range of 6-48 breaths / minute;

[0167] Peak detection: Respiratory cycles are identified based on waveform peak values. The minimum peak interval is set to 30 sampling points to avoid false detections.

[0168] Rate calculation: Instantaneous respiratory rate is calculated through the interpeak period;

[0169] Smoothing: Median filtering with a window size of 5 is used to remove outliers and output stable respiratory rate results.

[0170] The algorithm has an effective detection distance of 30-50 cm and a detection accuracy of ±1 time / minute, which is highly consistent with the detection results of contact breathing belts.

[0171] (2) Multi-sensor fusion sleep stage determination: The control module adopts a weighted fusion strategy, combining three types of signals—capacitive contact, acceleration, and respiratory rate—to determine the sleep stage.

[0172] Baseline initialization: After the first detection of continuous head contact, the average respiratory rate over 30 seconds is collected as the user's resting baseline.

[0173] Light sleep is defined as follows: if the head is in continuous contact with the skin for ≥5 minutes, the normalized value of the accelerometer motion intensity is <15, and the respiratory rate is <85% of the resting baseline, the person is considered to be in a light sleep state.

[0174] Deep sleep is defined as follows: the head is in continuous contact with the lungs for ≥10 minutes, the normalized value of the accelerometer motion intensity is <5, and the respiratory rate fluctuates by <1 breath / minute within 5 consecutive minutes.

[0175] Night awakening criteria: If the intensity of exercise suddenly increases beyond a threshold, and the respiratory rate increases by more than 30% within 30 seconds, and the person was previously in a state of continuous contact, the person is considered to be awake at night.

[0176] Validated by 20 subjects, the algorithm achieved a deep sleep detection rate of ≥85% and a light sleep detection rate of ≥80%, meeting the detection accuracy requirements for consumer-grade sleep aid products.

[0177] In this embodiment, the logic for determining the sleep stage through multi-sensor fusion is described below. Figure 3 As shown; see also Figure 4 As shown, the voice interaction uses a local offline solution. After the sleep aid mode is activated, the three stimulation channels output synchronously with a common rhythm cycle, forming a rhythmic entrainment effect:

[0178] Taking a common cycle of 0.25 Hz as an example, at t=0, the directional sound curtain starts to output a low frequency of 37 Hz, the bone conduction oscillator starts to play a heartbeat sound, and the LED light ring starts to activate an orange breathing light effect; 0-2 seconds is the inhalation phase, and the LED brightness gradually increases from 0% to 30%; 2-4 seconds is the exhalation phase, and the LED brightness gradually decreases from 30% to 5%; the above cycle is repeated from 4-8 seconds.

[0179] After 16 seconds of induction, the system enters the real-time breathing matching phase: the infrared sensor updates the user's breathing rate every 2 seconds, which is then smoothed by Kalman filtering. The target BPM for bone conduction is calculated according to a mapping relationship of "breathing rate × 4", and the target BPM is limited to the range of 50-65 BPM. Linear interpolation is used to achieve a smooth transition from the current BPM to the target BPM, avoiding abrupt changes in rhythm. At the same time, the pulsation period of the LED light ring is adjusted synchronously to achieve a two-way matching of "active guidance + passive following", more naturally inducing the user to relax and fall asleep.

[0180] The system of this invention implements mode management based on a state machine, including five core states: standby, sleep aid, meditation, wake-up, and night light.

[0181] 1. Standby mode: LED flashes orange slowly (0.1 Hz), sensor polls at low power, voice module only listens for wake word; user triggers mode switching by voice or tap.

[0182] 2. Sleep Aid Mode: Implements three-channel synergistic induction and adaptive control of sleep stages. It automatically adjusts the output after detecting that the user has fallen asleep; it switches to wake-up mode when the wake-up time is detected.

[0183] 3. Meditation Mode: Turns off directional sound screen and bone conduction channel, full-range speaker plays alpha waves or guided words, LED outputs blue flowing light effect; exits after timer ends or user command, plays chime feedback.

[0184] 4. Wake-up Mode: Starts before the preset time, the light gradually changes from dark purple to warm yellow, and the birdsong effect gradually increases, simulating the wake-up effect of sunrise; the user can stop it in advance by tapping or giving a voice command.

[0185] 5. Night light mode: The LED stays on with a warm light at the lowest brightness. It automatically returns to standby mode after 5 minutes of inactivity. Voice adjustment of brightness and color temperature is supported.

[0186] The status switching is triggered by three methods: voice command, tap interaction, and automatic detection, without the need for mobile phone operation, achieving screenless interaction.

[0187] In this embodiment, see Figure 3 As shown, the voice interaction uses a local offline solution, and the recognition process is as follows:

[0188] The dual-microphone array continuously collects ambient audio, and the voice wake-up chip calculates the matching confidence of the wake-up word "Hello Dreeem" in real time.

[0189] When the confidence level is greater than 0.85, the main control MCU is woken up, and the following 2 seconds of audio is sent to the NPU for instruction classification;

[0190] The NPU outputs a command category, and executes the corresponding operation when the confidence level is >0.8, providing feedback through LED lighting effects;

[0191] After execution, the main control unit enters low power mode, while the voice chip continues to listen for the wake word.

[0192] The system supports 16 offline commands, covering all scenarios including mode switching, lighting control, and playback control; the recognition accuracy is ≥95% in quiet environments and ≥85% in 30dB background noise.

[0193] In terms of privacy protection, a physical sliding microphone switch is installed at the bottom. When switched to the OFF position, the microphone power supply circuit is completely cut off, eliminating the risk of recording from a hardware perspective. At the same time, a red LED flashes three times to confirm that the microphone is disabled, making the status clearly perceptible. All voice processing and sleep data are completed locally, with no network uploads, ensuring user privacy and security.

[0194] The control module stores sleep data from the past 30 days in local Flash memory, including sleep latency, duration of each sleep stage, number of nighttime awakenings, and user-manually adjusted data. The individualized learning unit automatically optimizes system parameters based on historical data.

[0195] If the sleep latency is greater than 30 minutes for 3 consecutive days, gradually reduce the bone conduction initiation BPM (from 60 to 55) and the directional sound curtain center frequency (from 37 Hz to 32 Hz) to enhance the induction intensity;

[0196] If the number of nighttime awakenings exceeds 2 times for 3 consecutive days, the white noise fade-out time of the nighttime awakening intervention will be extended from 10 minutes to 15 minutes to improve the intervention effect.

[0197] If the user frequently adjusts the night light brightness manually, the default brightness level of the night light mode will be automatically increased to adapt to the user's usage habits.

[0198] All data and parameters are stored locally and are not uploaded to any cloud servers, balancing personalized experience with privacy and security.

[0199] In summary, the complete execution steps of the closed-loop control method of the present invention are as follows:

[0200] 1. Mode Activation: The user says "Hello Dreeem, sleep aid mode" or taps the main unit, and the system enters sleep aid mode from standby mode, playing a power-on confirmation tone.

[0201] 2. Synergistic Induction: Synchronously activate directional low-frequency sound curtain, 60BPM bone conduction heartbeat sound, and 0.25 Hz breathing light, with three channels outputting in a time-synchronized manner to begin active induction.

[0202] 3. Breathing Matching: The infrared sensor detects the breathing rate every 2 seconds and dynamically adjusts the bone conduction rhythm and light cycle to gradually synchronize with the user's breathing.

[0203] 4. Status determination: The sensor fusion algorithm continuously monitors the user's status, updating motion data every 50ms and breathing data every 2 seconds to determine the sleep stage in real time.

[0204] 5. Adaptive adjustment: After light sleep is detected, bone conduction and light will slowly fade out within 3 minutes, while retaining the directional sound curtain; after deep sleep is detected, all stimuli will be turned off and the system will enter low-power standby mode; after night awakening is detected, white noise and warm light intervention will be activated.

[0205] 6. Intelligent wake-up: Before the preset wake-up time is reached, gradually activate the light gradient and bird chirping sound effects to wake the user during the light sleep period.

[0206] 7. Data recording: Sleep data for the entire night is stored in a local log for individualized learning units to optimize parameters.

[0207] This invention provides an intelligent sleep aid system and closed-loop control method based on multimodal synergistic induction, which solves the technical problems of single stimulation, lack of closed-loop feedback, partner interference, and inconvenient interaction in existing technologies. It significantly shortens the sleep latency by actively inducing brainwaves to transition to a sleep state through the rhythmic entrainment effect of the three channels of hearing, touch, and vision. This can shorten the sleep latency by an average of more than 15 minutes, which is superior to single-stimulation sleep aid products. A directional low-frequency sound curtain confines sound energy within a narrow beam of ±15°. Combined with the solid-state sound transmission characteristics of bone conduction, the sound leakage at 1 meter is ≤25dB SPL, far lower than traditional white noise devices, and does not affect the rest of the bed partner. Real-time matching of breathing rate is used for closed-loop control of the sleep aid, achieving dynamic synchronization between stimulation parameters and the user's physiological state. The stimulation parameters are dynamically synchronized with the user's physiological state; combined with automatic adjustment of output intensity according to the sleep stage, it avoids continuous stimulation after falling asleep, which would interfere with deep sleep. Fully offline voice recognition and physical tap interaction eliminate the need for a mobile app, avoiding blue light exposure before sleep; integrated voice-activated night light function meets low-brightness lighting needs when getting up at night. Local voice processing and local data storage ensure that all sleep data is not uploaded to the cloud; a physical microphone disable switch cuts off microphone power at the hardware level, eliminating privacy concerns.

[0208] The foregoing description of specific exemplary embodiments of the invention is for illustrative and explanatory purposes. These descriptions are not intended to limit the invention to the precise forms disclosed, and it will be apparent that many changes and variations can be made in accordance with the foregoing teachings. The exemplary embodiments were chosen and described in order to explain the specific principles of the invention and its practical application, thereby enabling those skilled in the art to implement and utilize various different exemplary embodiments of the invention, as well as various different choices and variations. The scope of the invention is intended to be defined by the claims and their equivalents.

Claims

1. A smart sleep aid system based on multimodal collaborative induction, characterized in that, It includes a main unit housing, and an acoustic module, a sensor module, a lighting module, a voice interaction module, and a control module disposed inside the main unit housing; The acoustic module includes a speaker array, a full-range speaker, and a bone conduction transducer; the speaker array consists of at least four superlinear microspeakers, used to generate a directional low-frequency sound curtain through electronic beamforming technology. The full-range speaker is used to play sleep aid audio, meditation background music, wake-up sound effects and night wake-up intervention audio; the bone conduction oscillator is used to connect to an independent external bone conduction pillow pad through a detachable medical-grade silicone sound tube. The bone conduction pillow pad is a flat silicone structure with a vibrating diaphragm inside, which is placed under the user's pillow to transmit low-frequency vibrations. The sensor module includes an infrared thermopile sensor, a capacitive contact sensor, and an accelerometer; the infrared thermopile sensor is used for non-contact detection of the user's breathing rate; the capacitive contact sensor is disposed inside the bone conduction pillow and is used to detect the contact state between the user's head and the bone conduction pillow. The accelerometer is used to detect the intensity of the user's body movements and to identify the interactive signal of the user tapping the host. The lighting module is an RGB LED light ring surrounding the edge of the host housing. The voice interaction module includes a dual MEMS microphone array and a local voice wake-up chip for offline recognition of preset voice commands. The voice interaction module has an independent physical microphone disable switch. The control module is electrically connected to the acoustic module, sensor module, lighting module, and voice interaction module respectively.

2. The intelligent sleep aid system based on multimodal collaborative induction as described in claim 1, characterized in that, The directional low-frequency sound curtain has a frequency range of 30-100 Hz, a horizontal beam angle within ±15°, and a sidelobe suppression of not less than 12dB; the bone conduction oscillator operates in a frequency range of 20-200 Hz.

3. The intelligent sleep aid system based on multimodal collaborative induction as described in claim 1, characterized in that, The effective detection distance of the infrared thermopile sensor is no more than 50 cm; the lighting module contains at least 24 RGB LED beads, with at least 5 levels of adjustable brightness and at least 2 levels of adjustable color temperature.

4. The intelligent sleep aid system based on multimodal collaborative induction as described in claim 1, characterized in that, The control module is configured to execute the following closed-loop control logic: When the system enters sleep aid mode, it simultaneously activates three stimulation channels for coordinated induction: The first stimulation channel controls the speaker array to output the directional low-frequency sound curtain; The second stimulation channel controls the bone conduction oscillator to play heartbeat sounds or output low-frequency vibrations in a manner that starts at 60 BPM and decreases to 50 BPM. The third stimulation channel controls the LED light ring to provide pulsating illumination in a preset respiratory cycle, wherein the preset respiratory cycle is 4 seconds of inhalation and 4-6 seconds of exhalation, and the pulsation frequency is matched to the brain slow oscillation frequency band of 0.1 Hz to 0.25 Hz. The control module acquires the user's respiratory rate detected by the infrared thermopile sensor in real time, and dynamically adjusts the output rhythm of the bone conduction oscillator after filtering and smoothing to keep the bone conduction rhythm synchronized with the user's current respiratory rate. When the sensor module detects that the user has entered a light sleep state, it controls the second and third stimulation channels to gradually fade out within 3 minutes, while maintaining the continuous operation of the first stimulation channel. When the sensor module detects that the user has entered a deep sleep state, it controls all stimulation channels to enter a silent standby state, and at the same time controls the sensor module to switch to a low-power polling mode.

5. The intelligent sleep aid system based on multimodal collaborative induction as described in claim 4, characterized in that, The output timing of the directional low-frequency sound curtain, the rhythmic output timing of the bone conduction oscillator, and the pulsating illumination timing of the LED light ring are synchronized in time, and are output in coordination with a unified common rhythmic cycle to form a rhythmic entrainment effect of multi-sensory channels.

6. The intelligent sleep aid system based on multimodal collaborative induction as described in claim 4, characterized in that, The control module determines that the user has entered a light sleep state under the following conditions: the capacitive contact sensor detects that the user's head is in continuous contact with the sensor for more than 5 minutes, the intensity of the continuous motion signal detected by the accelerometer is lower than a first preset threshold, and the respiratory rate detected by the infrared thermopile sensor is more than 15% lower than the user's resting baseline.

7. The intelligent sleep aid system based on multimodal collaborative induction as described in claim 4, characterized in that, The control module determines that the user has entered a deep sleep state under the following conditions: the capacitive contact sensor detects that the user's head is in continuous contact with the sensor for more than 10 minutes, the intensity of the motion signal detected by the accelerometer is lower than the second preset threshold, and the fluctuation of the breathing rate detected by the infrared thermopile sensor does not exceed 1 breath / minute within 5 consecutive minutes.

8. The intelligent sleep aid system based on multimodal collaborative induction as described in claim 4, characterized in that, The control module has a built-in night-wake intervention unit, used for: When the accelerometer detects a sudden increase in motion signal intensity that exceeds a third preset threshold, and the infrared thermopile sensor detects that the user's breathing rate increases by more than 30% within 30 seconds, it is determined that the user has woken up at night. Automatic activation of night wake-up intervention mode: Controls the full-range speaker to play white noise or natural ambient sound, while controlling the LED light ring to light up warm light at the lowest brightness level and 2200K color temperature; The white noise or natural ambient sound gradually fades out in 10-minute cycles, and the LED ring automatically turns off 5 minutes after the audio fades out.

9. The intelligent sleep aid system based on multimodal collaborative induction as described in claim 4, characterized in that, The control module also has a built-in wake-up unit for: Start the wake-up mode 10-30 minutes before the preset wake-up time; The LED light ring is controlled to gradually change from dark purple to warm yellow, with the gradation process lasting 10 minutes. The color temperature gradually changes from 1600K to 4000K, and the brightness gradually increases from 10% to 60%. Synchronously control the full-range speaker to play a natural wake-up sound effect, with the volume gradually increasing from 10dB SPL to 45dB SPL; Upon detecting a user tapping the host or inputting a voice command, the wake-up mode is immediately terminated.

10. A closed-loop control method for an intelligent sleep aid system based on multimodal collaborative induction, applied to the intelligent sleep aid system according to any one of claims 1-9, characterized in that, Includes the following steps: S1. Mode Activation and Selection: Users can trigger the system to enter the corresponding working mode through voice commands or physical tapping operations. The working modes include sleep aid mode, meditation mode, night light mode and wake-up mode. S2. Multimodal Collaborative Induction: When entering sleep aid mode, the control module simultaneously activates three stimulation channels for output: The first channel controls the loudspeaker array to generate a directional low-frequency sound curtain of 30-100 Hz through electronic beamforming, with the horizontal beam angle controlled within ±15°; The second channel controls the bone conduction oscillator to play heartbeat sounds or output low-frequency vibrations starting at 60 BPM, and gradually decreases to 50 BPM. The third channel controls the RGB LED light ring to provide pulsating illumination with a breathing cycle of 4 seconds of inhalation and 4-6 seconds of exhalation, with the pulsation frequency matching the 0.1 Hz to 0.25 Hz frequency band; S3. Real-time closed-loop matching of breathing: The infrared thermopile sensor detects the user's breathing rate in real time with a sampling period of no more than 2 seconds. After smoothing by Kalman filtering, the control module uses the breathing rate as a feedback signal to dynamically adjust the output rhythm of the second channel bone conduction oscillator so that the bone conduction rhythm is synchronized with the user's current breathing rate. S4. Multi-sensor fusion sleep stage determination: The control module integrates the detection data from capacitive contact sensors, accelerometers, and infrared thermopile sensors to determine whether the user is currently awake, in light sleep, in deep sleep, or awake at night. S5. Sleep stage adaptive adjustment: If the sleep state is determined to be light sleep, control the second and third channels to gradually fade out within 3 minutes, while maintaining the continuous operation of the first channel directional sound curtain; If the state is determined to be deep sleep, all stimulation channels are controlled to enter silent standby mode, and the sensor module is switched to low-power polling mode. If the condition is determined to be a nighttime awakening, initiate the nighttime awakening intervention procedure; S6. Night Wake-up Intervention Execution: Automatically control the full-range speaker to play white noise or natural ambient sounds, while controlling the LED light ring to light up the lowest brightness 2200K warm light; the white noise gradually fades out in 10-minute cycles, and the LED light ring automatically turns off 5 minutes after the audio fades out.

Images