Search

CN-121987915-A - Multi-mode sleep-aiding earphone control system driven by biofeedback

CN121987915ACN 121987915 ACN121987915 ACN 121987915ACN-121987915-A

Abstract

The invention discloses a biofeedback driven multimode sleep-aiding earphone control system, which belongs to the technical field of biomedical engineering and artificial intelligence, wherein a sensor array built in an earphone is used for collecting brain electrical signals, heart rate variability signals, respiratory frequency signals and body temperature signals of a user in real time, estimating sleep preparation stages and physiological tension of the user, adaptively adjusting audio content and parameters according to an estimation result, innovatively integrating a respiratory guidance function, inducing respiratory frequency synchronization of the user through audio rhythm, activating a parasympathetic nervous system, continuously learning user history data based on a personalized optimization engine for reinforcement learning, realizing a thousand-face sleep-aiding scheme, and a biofeedback closed-loop control module is used for monitoring intervention effects in real time and dynamically adjusting strategies.

Inventors

  • YANG XINGWANG
  • CHEN YONGYAN
  • LIU ZIJIAN

Assignees

  • 江西佳芯物联有限公司

Dates

Publication Date
20260508
Application Date
20251231

Claims (10)

  1. 1. The multi-mode sleep-aiding earphone control system driven by biofeedback is characterized by comprising: the multi-source physiological signal acquisition module is used for acquiring brain electrical signals, heart rate variability signals, respiratory frequency signals and body temperature signals of a user in real time through a sensor array built in the earphone to generate a physiological state data set of the user containing various physiological parameters; The physiological state evaluation module is used for carrying out feature extraction and pattern recognition on the physiological state data set of the user, obtaining a sleep preparation stage identifier of the user at present, wherein the sleep preparation stage identifier comprises at least one of a waking period, a sleeping transition period, a shallow sleeping period and a deep sleeping period, and calculating and generating a real-time physiological tension score according to amplitude change features and frequency domain features of various physiological signals; The self-adaptive audio adjusting module is used for selecting a target audio content type from a preset audio strategy library according to the sleep preparation stage mark and the real-time physiological tension score, dynamically calculating an audio parameter adjusting scheme based on the real-time physiological tension score, wherein the audio parameter adjusting scheme comprises a tone adjusting coefficient, a rhythm adjusting coefficient, a volume adjusting coefficient and a frequency adjusting coefficient, if the real-time physiological tension score is in a first target interval range, the tone adjusting coefficient and the rhythm adjusting coefficient are reduced to promote relaxation, and if the real-time physiological tension score is in a second target interval range, the volume adjusting coefficient and the frequency adjusting coefficient are further reduced to deepen a sleep inducing effect; The breathing guiding module is used for calculating the current breathing period of the user based on the breathing frequency signal, generating an audio rhythm guiding signal matched with the current breathing period, and gradually inducing the breathing frequency of the user to approach the preset comfort breathing frequency threshold through the audio rhythm guiding signal if the frequency of the current breathing period is detected to be higher than the preset comfort breathing frequency threshold, so that progressive activation of a parasympathetic nervous system is realized; The personalized optimization engine is used for continuously learning historical sleep data and an audio response mode of a user based on a reinforcement learning algorithm, wherein the historical sleep data comprises sleep latency time duration, sleep quality scores and awakening times under different audio parameter combinations, and the generation strategy of the audio parameter adjustment scheme is dynamically updated by constructing a personalized state-action-rewarding mapping relation of the user, so that the optimal audio parameter combination which is more in line with individual characteristics of the user can be generated in the same sleep preparation stage; The biological feedback closed-loop control module is used for continuously monitoring the change of the physiological signal of the user after audio output and calculating a physiological signal improvement degree index, wherein the physiological signal improvement degree index is obtained by comparing the heart rate variability change amount, the brain wave alpha wave and theta wave ratio change amount and the respiratory frequency change amount before and after audio adjustment, and if the physiological signal improvement degree index does not reach a preset improvement threshold value, the self-adaptive audio adjustment module is triggered to regenerate an audio parameter adjustment scheme to form a closed-loop control mechanism of real-time feedback adjustment.
  2. 2. The system of claim 1, wherein the multi-source physiological signal acquisition module comprises: The electroencephalogram signal acquisition unit is used for acquiring electroencephalogram signals of the forehead and the back of the ear of a user through a dry electrode arranged on the inner side of an earphone ear cover, and the sampling frequency is 128Hz to 512Hz; the heart rate variability acquisition unit is used for acquiring a heart rate signal of a user through a photoelectric volume pulse wave sensor arranged at a skin contact part of the earphone and calculating variability parameters of continuous heartbeat intervals; The respiration monitoring unit is used for capturing micro displacement changes caused by the respiration motion of the user through a micro acceleration sensor or a pressure sensor arranged in the earphone and identifying the respiration frequency and the respiration depth; The body temperature monitoring unit is used for continuously monitoring the body surface temperature change trend of the user through the temperature sensor arranged in the contact area of the earphone and the skin.
  3. 3. The system of claim 1, wherein in the physiological state assessment module: The method for determining the sleep preparation stage mark comprises the steps of extracting frequency band power of delta wave, theta wave, alpha wave and beta wave in the electroencephalogram signal, calculating the power proportion of each frequency band, judging as a awake period if the power proportion of the alpha wave is more than 40% and the power proportion of the beta wave is more than 30%, judging as a sleep transition period if the power proportion of the theta wave gradually rises and exceeds 25%, judging as a shallow sleep period if the power proportion of the delta wave exceeds 20%, and judging as a deep sleep period if the power proportion of the delta wave exceeds 50%; The calculation method of the real-time physiological tension score comprises the steps of carrying out time domain analysis and frequency domain analysis on the heart rate variability signal, extracting a standard deviation index SDNN, low-frequency power LF and high-frequency power HF, calculating the ratio of LF to HF as a sympathological nerve activation degree index, combining the amplitude of beta waves in the electroencephalogram signal and the numerical value of the respiratory frequency, and generating the real-time physiological tension score in the range of 0 to 100 in a weighted summation mode, wherein the higher the score numerical value is, the higher the tension degree is.
  4. 4. The system of claim 1, wherein the adaptive audio conditioning module: the preset audio strategy stores a plurality of audio content types including natural environment sound, binaural beats, white noise variants, meditation guiding voices and personalized music fragments, and each audio content type corresponds to different reference audio parameter configurations; The audio parameter adjustment scheme is generated by firstly matching initial audio content from the preset audio strategy library according to the sleep preparation stage mark, then calculating parameter offset according to the real-time physiological tension score, wherein the first target interval range is 60-80, the second target interval range is 40-60, if the score is in the first target interval, the tone adjustment coefficient is set to be 0.7-0.9, the rhythm adjustment coefficient is set to be 50-70 times per minute, if the score is in the second target interval, the volume adjustment coefficient is reduced to be 50-70% of the reference volume, and the frequency adjustment coefficient is adjusted to be a low frequency range of 100-300 Hz.
  5. 5. The system of claim 1, wherein in the breath guide module: the preset soothing respiratory rate threshold is 6 to 10 times per minute; the method for generating the audio rhythm guide signal comprises the steps of calculating a target breathing period according to the current breathing period, wherein the target breathing period is obtained by gradually prolonging the current breathing period, each adjusting period is prolonged by 0.5 to 2 seconds, audio envelope changes synchronous with the target breathing period are generated, the breathing rhythm is simulated through periodic fluctuation of volume, and a user is guided to adjust the breathing mode under auditory prompt; And when the current respiratory cycle and the audio rhythm guiding signal are detected to be kept synchronous for more than 3 continuous respiratory cycles, judging that the guiding is successful, continuously maintaining the current audio rhythm, and if the current respiratory cycle and the audio rhythm guiding signal are detected to be asynchronous, suspending rhythm adjustment and maintaining the current parameters.
  6. 6. The system of claim 1, wherein in the personalized optimization engine: the reinforcement learning algorithm adopts a deep Q network or strategy gradient method, a state space comprises the sleep preparation stage identifier, the real-time physiological tension degree score, a current time period and a historical audio use record, an action space comprises the audio content type selection and specific parameter configuration of the audio parameter adjustment scheme, and a reward signal is comprehensively calculated based on sleep-in latency shortening degree, sleep quality score lifting amplitude and user subjective comfort degree feedback; The user individuation state-action-rewarding mapping relation is established by collecting user sleep data for at least 7 continuous days, and model parameters are updated after each use, so that the individuation optimizing engine can identify response differences of specific users to different audio strategies in a specific state, and an individuation sleep-aiding scheme for thousands of people and thousands of faces is realized.
  7. 7. The system of claim 1, wherein the biofeedback closed loop control module: The method for calculating the physiological signal improvement index comprises the steps of calculating the difference between a current physiological signal and a baseline physiological signal every 30 seconds to 120 seconds after audio adjustment is started, obtaining the heart rate variability variable quantity by comparing the increasing amplitude of SDNN parameters, obtaining the brain wave alpha wave and theta wave specific value variable quantity by calculating the specific value descending degree, obtaining the respiratory frequency variable quantity by calculating the frequency descending amplitude, carrying out normalization processing on the three variable quantities, and then carrying out weighted summation to obtain the physiological signal improvement index; The preset improvement threshold is dynamically set according to the user type, a lower threshold is set for a user with sleep disorder tendency, a higher threshold is set for a user with good sleep quality, and if the preset improvement threshold is not reached in the continuous 3 times of detection, a strategy switching mechanism is triggered, and an alternative audio strategy is selected from the preset audio strategy library to try again.
  8. 8. The system of claim 1, further comprising: The sleep quality assessment module is used for comprehensively analyzing physiological signal data of the whole night after the user finishes sleeping, calculating the sleep latency time, the total sleep time, the awakening times, the deep sleep period occupation ratio and the shallow sleep period occupation ratio, and generating a sleep quality comprehensive score and a personalized improvement suggestion report; The data management module is used for safely storing historical sleep data, audio use records and physiological signal original data of the user, supporting data export and cloud synchronization functions and protecting the encryption storage of privacy data of the user.
  9. 9. The system of claim 1, wherein the adaptive audio conditioning module further comprises an ambient noise compensation function: The ambient noise level is monitored in real time through the ambient noise sensor outside the earphone, when the ambient noise decibel value is detected to exceed the preset noise threshold value, the volume adjustment coefficient is automatically increased, the active noise reduction function is started, and the ambient noise is attenuated to a level which does not influence the sleep-aiding audio effect.
  10. 10. The system of claim 1, further comprising a security protection mechanism: An abnormality detection unit is arranged in the multi-source physiological signal acquisition module and is used for monitoring whether abnormal fluctuation occurs in a physiological signal of a user in real time, if the abnormal fluctuation occurs in the physiological signal of the user or the continuous abnormality occurs in the respiratory frequency or the heart rate exceeds a preset safety range, a reminding signal is sent out, the audio output is automatically reduced, and if necessary, an emergency notification is sent through a matched mobile application; and setting a volume upper limit protection in the self-adaptive audio adjusting module, preventing the audio output volume from exceeding a hearing safety threshold, and protecting the hearing health of a user.

Description

Multi-mode sleep-aiding earphone control system driven by biofeedback Technical Field The invention relates to the technical field of biomedical engineering and artificial intelligence, in particular to a biofeedback driven multimode sleep-aiding earphone control system which is used for assisting a user to improve sleep quality and shorten sleep time by integrating multisource physiological signal monitoring, a self-adaptive audio adjusting technology and a reinforcement learning personalized optimization algorithm. Background Sleep disorders have become a common health problem in modern society, and it is statistically common that about one third of adults worldwide have different degrees of sleep problems, including symptoms such as difficulty in falling asleep, poor sleep quality, early awakening, etc. Not only does sleep insufficiency affect daytime mental state and work efficiency, but long-term increases the risk of cardiovascular diseases, metabolic disorders and reduced immune function. Traditional sleep-aiding methods mainly comprise drug treatment and cognitive behavior therapy, but the drug treatment has problems of dependency and side effects, and the cognitive behavior therapy requires guidance of professionals and takes a long time. In recent years, audio-based sleep-aiding technology is widely focused due to non-invasive and usability, but most of the prior art adopts fixed audio content for playing, and lacks the ability of sensing and dynamically adjusting the real-time physiological state of a user. The Chinese patent application with the application number 202410191378.8 discloses a sleep and breathing evaluation and auxiliary regulation method, a system and a device based on sound, the technical proposal is that by collecting sound signals, acceleration signals and brain state signals in the sleeping process of a user, and identifying the sleep breathing sound event and generating a sleep breathing evaluation report, and further generating a sleep breathing auxiliary regulation strategy through time series prediction analysis. This prior art focuses mainly on the monitoring and assessment of sleep disordered breathing, recognizing respiratory events such as snoring by audible signal analysis, and controlling external conditioning devices such as ventilators or posture conditioning devices to intervene based on the prediction. However, this prior art has the following disadvantages. First, the core of the technical scheme is the recognition and evaluation of sleep respiratory events, and the technical scheme mainly aims at specific people with sleep respiratory disorders and lacks a targeted sleep-aiding solution for the user groups with common insomnia or difficulty in falling asleep. Secondly, the technology relies on acquisition and analysis of acoustic signals and body position signals, and lacks of real-time monitoring of physiological indexes which more directly reflect sleep states, such as brain electrical activity, heart rate variability and the like, so that judgment of sleep preparation stages of users is not accurate enough. Thirdly, the auxiliary regulation strategy of the technology needs the cooperation of external equipment such as a breathing machine or an intelligent mattress, the system integration level is low, the use field Jing Shouxian cannot meet the portable sleep-aiding requirements of users in different environments. Fourth, the technology adopts a time sequence prediction method based on historical data to generate an adjustment strategy, lacks deep learning and self-adaptive optimization capability for individual differences, and is difficult to realize personalized sleep-aiding effect for thousands of people. Fifth, the adjusting mechanism of the technology is mainly open-loop unidirectional control, that is, a control instruction is sent to the adjusting device according to a prediction result, and a real closed-loop control system cannot be formed due to lack of a real-time feedback monitoring and dynamic adjusting mechanism for adjusting effects. With the rapid development of wearable device technology and artificial intelligence algorithms, foreign research institutions and enterprises have made remarkable progress in the field of biofeedback sleep assistance. The Muse S Athena intelligent headband developed by Muse corporation combines an electroencephalogram and a functional near infrared spectrum technology, can monitor brain electric activity and cerebral blood oxygen level at the same time, and guides a user to enter meditation and sleep states through real-time nerve feedback. The MW75 nerve earphone built-in brain sensor introduced by Neurable can track the concentration degree and fatigue degree of the user, and prompts the user to rest time through an artificial intelligent algorithm. The academic world also breaks through in the aspects of multi-mode physiological signal fusion and sleep stage prediction, and mechanisms such as Stanford univers