WO2021103084A1 - 一种用于睡眠调节的深度声音刺激系统和方法 - Google Patents

一种用于睡眠调节的深度声音刺激系统和方法 Download PDF

Info

Publication number
WO2021103084A1
WO2021103084A1 PCT/CN2019/123156 CN2019123156W WO2021103084A1 WO 2021103084 A1 WO2021103084 A1 WO 2021103084A1 CN 2019123156 W CN2019123156 W CN 2019123156W WO 2021103084 A1 WO2021103084 A1 WO 2021103084A1
Authority
WO
WIPO (PCT)
Prior art keywords
sound
sleep
deep
eeg
optimal
Prior art date
Application number
PCT/CN2019/123156
Other languages
English (en)
French (fr)
Inventor
周晖晖
马征
谢津
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Publication of WO2021103084A1 publication Critical patent/WO2021103084A1/zh
Priority to US17/729,871 priority Critical patent/US20220249017A1/en

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61M—DEVICES FOR INTRODUCING MEDIA INTO, OR ONTO, THE BODY; DEVICES FOR TRANSDUCING BODY MEDIA OR FOR TAKING MEDIA FROM THE BODY; DEVICES FOR PRODUCING OR ENDING SLEEP OR STUPOR
    • A61M21/00—Other devices or methods to cause a change in the state of consciousness; Devices for producing or ending sleep by mechanical, optical, or acoustical means, e.g. for hypnosis
    • A61M21/02—Other devices or methods to cause a change in the state of consciousness; Devices for producing or ending sleep by mechanical, optical, or acoustical means, e.g. for hypnosis for inducing sleep or relaxation, e.g. by direct nerve stimulation, hypnosis, analgesia
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A61B5/375—Electroencephalography [EEG] using biofeedback
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/25—Bioelectric electrodes therefor
    • A61B5/279—Bioelectric electrodes therefor specially adapted for particular uses
    • A61B5/291—Bioelectric electrodes therefor specially adapted for particular uses for electroencephalography [EEG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A61B5/372—Analysis of electroencephalograms
    • A61B5/374—Detecting the frequency distribution of signals, e.g. detecting delta, theta, alpha, beta or gamma waves
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/4806—Sleep evaluation
    • A61B5/4812—Detecting sleep stages or cycles
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/4836—Diagnosis combined with treatment in closed-loop systems or methods
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6802—Sensor mounted on worn items
    • A61B5/6803—Head-worn items, e.g. helmets, masks, headphones or goggles
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7253—Details of waveform analysis characterised by using transforms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/74—Details of notification to user or communication with user or patient ; user input means
    • A61B5/7405—Details of notification to user or communication with user or patient ; user input means using sound
    • A61B5/7415—Sound rendering of measured values, e.g. by pitch or volume variation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0059—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence
    • A61B5/0075—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence by spectroscopy, i.e. measuring spectra, e.g. Raman spectroscopy, infrared absorption spectroscopy
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves 
    • A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61M—DEVICES FOR INTRODUCING MEDIA INTO, OR ONTO, THE BODY; DEVICES FOR TRANSDUCING BODY MEDIA OR FOR TAKING MEDIA FROM THE BODY; DEVICES FOR PRODUCING OR ENDING SLEEP OR STUPOR
    • A61M21/00—Other devices or methods to cause a change in the state of consciousness; Devices for producing or ending sleep by mechanical, optical, or acoustical means, e.g. for hypnosis
    • A61M2021/0005—Other devices or methods to cause a change in the state of consciousness; Devices for producing or ending sleep by mechanical, optical, or acoustical means, e.g. for hypnosis by the use of a particular sense, or stimulus
    • A61M2021/0027—Other devices or methods to cause a change in the state of consciousness; Devices for producing or ending sleep by mechanical, optical, or acoustical means, e.g. for hypnosis by the use of a particular sense, or stimulus by the hearing sense
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61M—DEVICES FOR INTRODUCING MEDIA INTO, OR ONTO, THE BODY; DEVICES FOR TRANSDUCING BODY MEDIA OR FOR TAKING MEDIA FROM THE BODY; DEVICES FOR PRODUCING OR ENDING SLEEP OR STUPOR
    • A61M2230/00—Measuring parameters of the user
    • A61M2230/08—Other bio-electrical signals
    • A61M2230/10—Electroencephalographic signals

Definitions

  • the invention relates to a deep sound stimulation system and method for sleep regulation, which regulates slow-wave sleep to enhance slow-wave activity and promote subjects to enter a deep sleep period as soon as possible.
  • sleep is essential for maintaining the normal physiological functions of the human body. Improving sleep quality through sleep regulation is of great value for promoting physical and mental health of the human body, and it is particularly important for people suffering from sleep disorders.
  • the current methods of sleep regulation include: (1) Psychiatric drug regulation, through the use of psychiatric drugs, inhibit the relevant excitatory activities of the central nervous system, thereby promoting sleep, the disadvantage is that it is easy to cause drug dependence and accompanied by greater side effects; (2) Psychotherapy, which relieves sleep disorders caused by psychological problems through mental health consultation.
  • the disadvantage is that the treatment cycle is long and it is difficult to treat symptomatically; (3) Neural circuit therapy, which repairs the nerves that cause sleep disorders through brain surgery Loop, but the related treatment plan is not yet mature, and is currently only used for the treatment of depression, etc.; (4) Acoustic and light stimulation methods, through sound (music, voice, natural sound, white/colored noise, etc.), Light and other external stimuli induce the cerebral cortex to produce neural activities that help sleep, which is currently the most effective and least risky sleep regulation method.
  • the present invention belongs to the category of methods for regulating sleep through sound stimulation.
  • Sleep regulation usually acts on different periods of a sleep cycle.
  • a sleep cycle can be divided into two stages: rapid eye movement sleep (REM) sleep and non-rapid eye movement (non-REM) sleep, and starts from non-REM sleep, to REM sleep, and then to the next cycle of non-REM sleep.
  • REM sleep usually lasts about 90 minutes.
  • Non-REM sleep is slow-wave sleep. It is characterized by specific slow-rhythmic sleep brain waves (electroencephalograph, EEG) recorded by scalp electrodes. It is divided into 4 periods: The first period is the light sleep period, which lasts for several minutes, and is mainly manifested in the EEG.
  • Theta wave of 4 ⁇ 7Hz appears in the second stage; sleep is a little deeper and lasts for 5 ⁇ 15 minutes, mainly manifested as the occasional 8 ⁇ 14Hz spindle wave and K complex wave appearing in the EEG; the third and the first stage Stage 4 is deep sleep, which is mainly manifested as high amplitude delta waves below 4 Hz in EEG. Stage 4 is the deepest sleep stage, which can last for 20-40 minutes. Its EEG rhythm is expressed as high amplitude delta waves below 2 Hz. . REM sleep is characterized by frequent eye movements and fast beta waves above 14 Hz in the EEG.
  • non-REM sleep is related to the elimination of certain neurotoxins (such as ⁇ -amyloid), rather than slow wave activity of REM sleep. Decrease may be related to aging and brain atrophy; REM sleep may play an important role in strengthening memory. Compared with the relatively monotonous slow wave of non-REM sleep, the EEG rhythm of REM sleep is closer to the awakening state, has richer components, and the regulation is more complicated. Therefore, the current sleep regulation technology based on EEG rhythm is mainly focused on non-REM sleep. REM sleep stage. The present invention regulates slow-wave sleep to enhance slow-wave activity and promote subjects to enter the deep sleep period as soon as possible.
  • neurotoxins such as ⁇ -amyloid
  • the existing technology for regulating sleep based on sound stimulation mainly uses the above-mentioned correlation between the sleep EEG and the spectral characteristics and acoustic characteristics of the sound stimulation to select sound stimulation. For example, detect the current sleep period based on the EEG rhythm change, and generate hypnotic sounds ( ⁇ wave, ⁇ wave, high ⁇ wave, low ⁇ wave, etc.) similar to the current EEG rhythm spectrum characteristics (Patent Publication Number: CN107715276A) ; Or select the music with the greatest correlation between acoustic characteristics (rhythm, pitch, pitch, etc.) and the sleep EEG recorded in the current sleep period from the music library (Patent Publication Number: CN105451801A); or directly record the natural world through a sound pickup device White noise with a wide frequency spectrum (wind, rain, water, etc.) is played to the subject (Patent Publication No. CN101773696B).
  • the prior art selects sound stimulation based on the external similarities between sleep EEG and sound stimulation in signal characteristics, instead of starting from the internal correlation between sound stimulation and sleep-related neural activities, choosing sound stimulation to strengthen Related neural activities, so the screening of sound stimuli is relatively rough, and there is a lack of relevant optimization.
  • the method of selecting the same frequency band sound for stimulation only considers the similarity of the sleep EEG and the sound stimulus used in the frequency spectrum, and each sleep EEG rhythm ( ⁇ wave, ⁇ wave, etc.) even if it belongs to the low-frequency slow wave, still has a certain It is still unclear whether the use of sound stimulation in the same frequency band can achieve the best enhancement of sleep-related neural activities, and the method of directly using natural white noise with a wider frequency spectrum for stimulation is even more lacking in sound screening.
  • the method of selecting sound stimuli from the music library based on the correlation between the acoustic characteristics of music and sleep EEG signals cannot be sure of the stimulation effect of the selected sound stimulus on sleep-related neural activities. At the same time, this method depends on the subject. The audience’s reaction to the actual music, so if you choose the most suitable music from the mass of music, it will be difficult to achieve due to the huge workload, so it is difficult to better optimize the sound stimulation.
  • the present invention starts from the brain-like characteristics of deep neural networks, finds the sound stimulus that can maximize the estimated sleep EEG response, broadcasts the selected sound stimulus to the subject, and performs closed-loop optimization.
  • the present invention mainly selects sounds on the trained deep neural network, without the subject to test a large number of sounds, thus greatly reducing the dependence on the subject's own voice selection Burden.
  • the present invention mainly includes two advantages: (1) The prior art only starts from the external similarities in the signal characteristics of sleep EEG and sound stimulation, selects the sound stimulation, and at the same time the sound source Less, so the selection of sound stimuli is rough and lacks relevant optimization; the present invention starts from the inherent correlation between sound stimuli and sleep-related neural activities, and relies on deep neural networks to select sound stimuli from a large number of natural sounds and synthetic sounds to strengthen Related nerve activity can obtain sound stimulation that is more effective in improving sleep quality.
  • the prior art selects the sound stimulus according to the correlation between the subject’s EEG and the acoustic characteristics of the sound stimulus, which is less efficient and difficult to perform large-scale testing; the present invention first uses a deep neural network to select the sound stimulus, and then The individualized optimal sound stimulus is selected through closed-loop optimization, so there is no need to record the subject’s response to a large number of sounds, and large-scale tests can be carried out efficiently, and then the sound stimulus that is most beneficial to sleep regulation can be selected from the large number of sounds.
  • Figure 1 Block diagram of a deep sound stimulation system for sleep regulation. Describe the components of the system of the present invention.
  • Figure 2 is a flow chart of predicting EEG activities based on deep neural networks. Description The method of the present invention is based on a deep neural network to predict EEG activity.
  • Figure 3 Flow chart of sound stimulus selection and optimization. Describe the method of selecting and optimizing natural sound and synthetic sound based on the deep neural network of the present invention.
  • FIG. 4 Single-channel brainwave amplification circuit. A realization circuit of the signal amplifier in the brainwave acquisition module used in the present invention is described.
  • the present invention is based on the deep neural network to select the best sound stimuli from natural sounds and synthesized sounds to form a sound library. It includes two steps: firstly establish a deep neural network mapping model for predicting EEG activity, and then select the best according to the established mapping model Sound stimulation.
  • the same music melody is input to the deep speech recognition network and played to the subject (user), and the changes in the output of the neurons in each layer of the model in the neural network and the changes in the subject’s real EEG signal are simultaneously recorded.
  • the EEG signal can be collected by itself, or public on-going EEG data of people under music stimulation can be used.
  • the music melody input into the deep speech recognition neural network should be the same as the music used to generate the on-going EEG.
  • the music melody used can be soothing piano music, pop music, etc.
  • Deep speech recognition networks can use publicly available acoustic models that have been trained on large-scale speech databases, or they can build their own models (such as HMM-CNN speech recognition models), and use them in large-scale speech databases (such as LibriSpeech, VoxCeleb, etc.). Voice/corpus).
  • the deep network is composed of an input layer, multiple intermediate layers, and an output layer.
  • the present invention focuses on the neuron layer in the middle layer of the deep network.
  • the stimulation of the music melody used in Figure 1 and the suggested recording of sleep EEG can reduce the interference of the early EEG signals caused by advanced cognitive processing in the later stages of the human brain.
  • the mapping relationship between the deep neural network and the EEG activity can be established by, but not limited to, the following method: the output vector of the i-th layer of the deep neural network is denoted as x i (t), and the corresponding EEG mapping weight is denoted as ⁇ i to predict EEG
  • the signal is denoted as y i (t)
  • the actual EEG signal currently recorded is denoted as s(t)
  • y i (t) ⁇ i ,x i (t)>, where ⁇ , ⁇ > represents the vector Inner product.
  • CCA canonical correlation analysis
  • Figure 3 shows the flow chart of selecting the optimal sound stimulus from a large number of natural sounds and synthetic sounds based on the mapping model of the established deep network to EEG activities.
  • Natural sounds include various types of music, speech, natural white noise, etc. that are widely available, and artificially synthesized sounds can be obtained through a deep music generation network.
  • the deep music generation network can use publicly trained acoustic models, or can build models (such as LSTM models) by itself for training.
  • the deep EEG prediction network is the optimal mapping of the deep speech recognition network shown in Figure 2 to EEG activities.
  • the sound stimulus is input to the deep EEG prediction network to obtain the estimation of the EEG signal by the deep network under the current sound stimulus, which is called the EEG estimation signal.
  • the frequency band energy can be estimated by the fast Fourier transform (FFT) method
  • Sort the input natural sounds from high to low according to the energy ratio of these frequency bands select the sound stimuli within the top decile of the sort, as the optimal sound stimuli corresponding to different sleep periods, and store them in the deep optimal sound library after marking among.
  • FFT fast Fourier transform
  • the brainwave acquisition module is composed of dry electrodes, amplifiers, and A/D conversion modules.
  • the dry electrode adopts Ag/AgCl electrode, which is placed on a fixed device such as a hairpin head-mounted device or an electrode cap woven from highly elastic fibers. The latter is worn on the user's head during use to ensure effective contact between the electrodes and the scalp.
  • a fixed device such as a hairpin head-mounted device or an electrode cap woven from highly elastic fibers.
  • dry electrodes do not need to be coated with conductive paste, which is more convenient for users to wear and use at any time.
  • the present invention proposes to use two electrodes F3 and F4 arranged in accordance with the 10/20 international EEG system standard; it can also be extended to use more electrodes, such as adding electrodes such as TP7 and TP8 in the occipitotemporal area.
  • the amplifier can be built with separate components or implemented with an existing integrated chip, such as the AD-620 chip of Analog Device.
  • the present invention can use the signal amplifying circuit shown in FIG. 4.
  • the circuit shown in Figure 4 consists of a front and back two-stage amplifier and a band-pass filter.
  • the front-stage instrumentation amplifier includes a differential amplifier circuit composed of IC1 and IC2, and a subtractor composed of IC3.
  • Vin is connected to the signal measuring electrode
  • Vref is connected to the reference electrode
  • the amplification factor of the preamplifier circuit of the present invention is 5-20, and 10 is recommended.
  • the bandpass filter can be a Butterworth or Chebyshev filter, and the recommended passband range is 0.05Hz ⁇ 40Hz.
  • the magnification of the post-amplifier (IC4) is from 50 to 200, and 100 is recommended.
  • the present invention needs to collect signals synchronously for multiple EEG electrode channels. Therefore, according to the number of electrodes used, the EEG acquisition module should include a corresponding number of multiple single-channel amplifiers.
  • the amplified output signal Vo is converted into a digital signal by an A/D conversion module.
  • the A/D conversion module converts the collected analog signal into a digital signal, which can be realized by the commonly used A/D chip in the market.
  • the present invention recommends an A/D conversion accuracy of at least 16 bits, and the sampling rate is recommended to be above 128 Hz.
  • the sleep monitoring module estimates the current sleep period of the subject according to the recorded EEG signal.
  • the present invention adopts the short-time Fourier transform (short time Fourier transform, STFT) method to obtain the real-time time-frequency diagram of the EEG signal s(t), and respectively calculates the energy of the shuttle wave, theta wave, the high delta wave, and the low delta wave relative to The relative value of the energy of ⁇ wave and ⁇ wave.
  • STFT short time Fourier transform
  • the spindle wave is dominant, it is considered to be in stage 1 of non-REM sleep; theta wave is dominant, and it is considered to be in stage 2 of non-REM sleep; if high delta waves are dominant, it is considered to be in stage 3 of non-REM sleep; low delta waves are dominant, and it is considered In the 4th stage of non-REM sleep; alpha wave and beta wave are dominant, it is considered to be in the REM sleep/wake period.
  • the sound stimulation selection module selects a specified number of optimal sound stimuli marked as the same sleep period from the sound library according to the current sleep period given by the sleep monitoring module. If you are currently in non-REM sleep stage 1, select the sound marked as non-REM sleep stage 1 or 2; if you are currently in non-REM sleep stage 2, select the sound marked as non-REM sleep stage 2 or 3; if currently If you are in stage 3 of non-REM sleep, select the sound marked as stage 3 or stage 4 of non-REM sleep; if you are currently in stage 4 of non-REM sleep, select the sound marked as stage 4 of non-REM sleep.
  • the system of the present invention plays sound stimulation to the subject and receives the EEG signal fed back by the subject. Therefore, the sound stimulation can be adjusted according to the intensity of the sleep wave in the feedback EEG signal, thereby forming a "sound stimulation-real-time EEG-sound stimulation" Closed loop to effectively optimize sound stimulation and obtain optimal sleep regulation for individual subjects.
  • An implementation method of the closed-loop optimization module of the system of the present invention is to select one from a set of optimal sound stimuli given by the sound stimulus selection module for playback. While playing the sound, obtain the current sleep state corresponding to the EEG frequency band from the sleep monitoring module.
  • Real-time energy relative value and use this relative value as the evaluation value of sleep quality in the current sleep period, and record it; then, switch to the next sound stimulus in this group and record the corresponding sleep quality evaluation value until all the most Excellent sound stimulation; finally, select the first several specified number of sound stimuli that dominate the sleep quality evaluation value as the optimal sound stimulus personalized for the subject; once the optimal sound stimulus personalized for the subject has been determined
  • the system of the present invention can skip the sound stimulation selection module and the closed-loop optimization module, and directly select the sound stimulus corresponding to the sleep period from the subject’s personalized optimal sound stimulus for playback until the detected sleep quality evaluation value drops significantly (For example, the decrease rate exceeds the original 30%), the sound stimulation selection and closed-loop optimization can be performed again.
  • the switch/volume control determines whether to play sound stimulation and adjusts the volume of the sound stimulation according to the current sleep period given by the sleep monitoring module. From stage 1 to stage 4 of non-REM sleep, the volume of sound stimulation can be gradually reduced from the initial volume (such as 60% of the maximum volume) to a preset volume (such as 10% of the maximum volume), and at the end of non-REM sleep Then (in REM sleep or wakefulness) turn off the sound stimulation.
  • the initial volume such as 60% of the maximum volume
  • a preset volume such as 10% of the maximum volume
  • the sound playback module includes an audio amplifier, which drives the speaker to play the sound stimulation according to the selected sound stimulus and the set volume/switch state.
  • the system and method of the present invention are not limited to brainwave EEG signals, but are also applicable to other physiological signals that can reflect sleep states, such as magnetoencephalography (MEG), near infrared spectroscopy (NIRS), functional magnetic resonance (fMRI), etc.
  • MEG magnetoencephalography
  • NIRS near infrared spectroscopy
  • fMRI functional magnetic resonance
  • the sleep monitoring method adopted by the system of the present invention is not limited to the above-mentioned method, and can also be replaced by a classifier-based monitoring method such as neural network, support vector machine, and other sleep state judgment methods based on power spectrum.
  • the brain wave acquisition module, sound stimulation selection module, switch/volume control module, sound playback module, loudspeaker, etc. in the system of the present invention can be replaced with other modules with equivalent or similar functions.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Veterinary Medicine (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Psychiatry (AREA)
  • Artificial Intelligence (AREA)
  • Psychology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physiology (AREA)
  • Signal Processing (AREA)
  • Anesthesiology (AREA)
  • Mathematical Physics (AREA)
  • Fuzzy Systems (AREA)
  • Evolutionary Computation (AREA)
  • Pain & Pain Management (AREA)
  • Acoustics & Sound (AREA)
  • Hematology (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

一种利用深度神经网络来选择和优化声音刺激，以调节与改善人体睡眠质量的系统和方法。深度神经网络具有表征人脑皮层神经元对外部刺激(图像、声音等)信息进行加工的能力。通过将海量声音刺激输入深度神经网络，寻找使模型估计的睡眠脑电波达到最优的声音模式，并将此声音模式应用于真实人体，经过闭环优化，增强人体在不同睡眠阶段对应睡眠波的强度，达到调节睡眠的目的。主要解决在利用声音刺激手段(音乐、语音、自然声音、白/有色噪声等)辅助人体睡眠时，如何通过深度神经网络选择和优化所用声音刺激以最大程度地改善睡眠质量的技术问题。

Description

一种用于睡眠调节的深度声音刺激系统和方法 技术领域
本发明涉及一种用于睡眠调节的深度声音刺激系统和方法，针对慢波睡眠进行调节，以增强慢波活动，促进受试者尽快进入深睡期。
背景技术
神经科学与临床医学研究结果表明，睡眠对于维持人体正常生理机能至关重要。通过睡眠调节来提高睡眠质量，对于促进人体身心健康具有重要价值，而对于罹患睡眠障碍的人群而言尤为重要。目前睡眠调节的途径包括：(1)精神类药物调节，通过使用精神类药物，抑制中枢神经系统的相关兴奋性活动，从而促进睡眠，缺点是易引起药物依赖性，并伴随较大的副作用；(2)心理疗法，通过心理健康咨询缓解因心理问题引起的睡眠障碍，缺点是治疗周期长，且较难准确对症治疗；(3)神经环路疗法，通过脑部手术修复引起睡眠障碍的神经环路，但相关治疗方案尚不成熟，目前仅尝试性地用于抑郁症等的治疗；(4)声、光刺激法，通过声音(音乐、语音、自然声音、白/有色噪声等)、光线等外部刺激诱导大脑皮层产生有助于睡眠的神经活动，是目前行之有效且风险最小的睡眠调节方法。本发明即属于通过声音刺激调节睡眠方法的范畴。
睡眠调节通常作用于一个睡眠周期的不同时段。一个睡眠周期可分为快速眼动(rapid eye movement sleep，REM)睡眠和非快速眼动(非REM)睡眠两个阶段，并从非REM睡眠开始，到REM睡眠，再到下一个周期的非REM睡眠，通常持续时间约90分钟。非REM睡眠即慢波睡眠，以头皮电极记录到的特定慢节律睡眠脑电波(electroencephalograph，EEG)为特征，分为4期：第1期为浅睡期，持续数分钟，主要表现为在EEG中出现 的4～7Hz的θ波；第2期睡眠稍深，持续5～15分钟，主要表现为在EEG中出现的偶发性的8～14Hz梭形波与K复合波；第3期和第4期为深度睡眠，主要表现为EEG中出现的4Hz以下高幅度的δ波，其中第4期是最深的睡眠阶段，可持续20～40分钟，其EEG节律表现为2Hz以下的高幅δ波。REM睡眠则以频繁的眼动及EEG中14Hz以上的快速β波为特征。尽管目前关于非REM睡眠对机体修复的机制尚不完全知晓，但最近研究结果表明，非REM睡眠与某些神经毒素(如β-淀粉样蛋白)的清除有关，而非REM睡眠慢波活动的减少则可能与衰老及脑萎缩有关；REM睡眠则对于记忆的强化或有重要作用。与非REM睡眠相对单调的慢波相比，REM睡眠的EEG节律则更接近觉醒状态，具有更加丰富的成分，而调节也更为复杂，因此目前根据EEG节律进行睡眠调节的技术主要集中在非REM睡眠阶段。本发明即针对慢波睡眠进行调节，以增强慢波活动，促进受试者尽快进入深睡期。
现有基于声音刺激调节睡眠的技术，主要利用上述睡眠EEG与声音刺激的频谱特征、声学特征等之间的关联关系选择声音刺激。例如，根据EEG节律变化检测当前所处的睡眠时段，并生成与当前EEG节律频谱特性相似的催眠声音(α波、θ波、高δ波、低δ波等声音)(专利公布号：CN107715276A)；或从音乐库中选择声学特征(节奏、音高、音调等)与当前睡眠时段所记录的睡眠EEG相关性最大的音乐(专利公布号：CN105451801A)；或直接通过声音拾取装置记录到的自然界中频谱较宽的白噪声(风声、雨声、流水声等)，播放给受试者(专利公布号：CN101773696B)。
现有技术从睡眠EEG与声音刺激二者在信号特征上表现出的外在相似之处出发，选择声音刺激，而不是从声音刺激与睡眠相关神经活动的内在关联性出发，选择声音刺激来强化相关神经活动，因此对声音刺激的筛选较为粗糙，同时缺少相关优化。根据睡眠EEG频谱选择同频带声音进行刺激的方法仅考虑睡眠EEG和所用声音刺激在频谱上的相似性，而各睡眠 EEG节律(θ波、δ波等)即使属于低频慢波，仍具有一定的频带宽度，同时利用这种同频段声音刺激能否达到睡眠相关神经活动的最佳强化仍不可知，而直接利用频谱范围更宽的自然白噪声进行刺激的方法则更加缺乏对声音的筛选。相似地，根据音乐的声学特征与睡眠EEG信号的相关性从音乐库中选择声音刺激的方法，也无法确知所选声音刺激对睡眠相关神经活动的激励作用，同时这一方法依赖于受试者对实际音乐的反应，因此如果从海量音乐中选择最适合受试者的音乐，将因面临巨大的工作量而难以实现，因此较难对声音刺激进行较好的优化。
针对现有技术中的上述缺点，本发明从深度神经网络的类脑特性出发，寻找可使估计所得睡眠EEG反应最大的声音刺激，将所选声音刺激播放给受试者，并经过闭环优化，使睡眠相关神经活动得到最大程度的强化；同时，本发明主要在训练好的深度神经网络上进行声音的选择，无需受试者对海量声音进行测试，因此大大降低依靠受试者自身进行声音选择的负担。
发明内容
与现有技术相比，本发明主要包括两方面优点：(1)现有技术仅从睡眠EEG与声音刺激二者在信号特征上表现出的外在相似之处出发，选择声音刺激，同时音源较少，因此对声音刺激的筛选较为粗糙，缺少相关优化；本发明从声音刺激与睡眠相关神经活动的内在关联性出发，依靠深度神经网络，从海量自然声音与合成声音中选择声音刺激来强化相关神经活动，可获得对提高睡眠质量更有效果的声音刺激。(2)现有技术根据受试者EEG与声音刺激声学特征的相关性来选择声音刺激，效率较低，且较难进行大规模测试；本发明首先利用深度神经网络进行声音刺激的选择，而后通过闭环优化选择个性化的最优声音刺激，因此无需记录受试者对海量声音的反应，可高效地开展大规模测试，进而从海量声音中选择对睡眠调节最有益的声音刺激。
附图说明
图1用于睡眠调节的深度声音刺激系统框图。描述本发明系统的各组成模块。
图2基于深度神经网络预测EEG活动流程图。描述本发明基于深度神经网络对EEG活动预测的方法。
图3声音刺激选择与优化流程图。描述本发明基于深度神经网络对自然声音和合成声音进行选择和优化的方法。
图4单通道脑电波放大电路。描述本发明所用脑电波采集模块中信号放大器的一种实现电路。
具体实施方式
下面结合实施例和附图对本发明的系统和方法做出详细说明。
(1)深度优化声音库
本发明基于深度神经网络从自然声音和合成声音中选择最优的声音刺激构成声音库，包括两个步骤：首先建立深度神经网络预测EEG活动的映射模型，而后根据所建立的映射模型选择最优声音刺激。
如图2所示，将相同的音乐旋律输入深度语音识别网络，及播放给受试者(用户)，同步记录神经网络中模型各层神经元输出的变化与受试者真实EEG信号的变化，确定模型神经元到真实EEG信号的最优映射关系，建立神经网络对EEG活动预测的映射模型。EEG信号可以自行采集，也可使用公开的人在音乐刺激下的on-going EEG数据，这时，输入深度语音识别神经网络中的音乐旋律应和产生on-going EEG所采用的音乐相同。使用的音乐旋律可采用舒缓的钢琴曲、流行音乐等。
深度语音识别网络可使用公开的已在大规模语音库上训练好的声学模型，也可以自行建立模型(如HMM-CNN语音识别模型)，并在大规模语音库(如LibriSpeech、VoxCeleb等大型人声/语料库)上训练得到。深 度网络由输入层、多个中间层、输出层构成。本发明重点关注深度网络中间层中靠前的神经元层。图1中所使用音乐旋律进行的刺激及所建议记录的睡眠EEG，可减少人脑后期的高级认知加工对早期EEG信号造成的干扰。
深度神经网络和EEG活动映射关系可通过但不局限于如下方法建立：将深度神经网络第i层神经元的输出向量记作x i(t)，对应的EEG映射权重记作ω i，预测EEG信号记作y i(t)，当前记录到的实际EEG信号记作s(t)，则有y i(t)＝<ω i,x i(t)>，其中<·,·>表示向量内积。采用典型相关分析法(canonical correlation analysis，CCA)，评估各y i(t)与s(t)之间的相关性，选取与实际EEG信号相关性最大的模型神经元层及对应的映射权重，作为最优映射，用于深度神经网络对EEG活动的预测。
图3所示为基于所建立的深度网络对EEG活动的映射模型，从海量自然声音与合成声音中选择最优声音刺激的流程图。自然声音包括可广泛获取的各类音乐、语音、自然白噪声等，人工合成声音可通过深度音乐生成网络得到。深度音乐生成网络可使用公开的已训练好的声学模型，也可自行建立模型(如LSTM模型)训练得到。深度EEG预测网络即为图2所示深度语音识别网络对EEG活动的最优映射。声音刺激输入深度EEG预测网络，得到当前声音刺激下深度网络对EEG信号的估计，称作EEG估计信号。对EEG估计信号中表征睡眠不同时段的梭形波、θ波、高δ波、低δ波所占的频带能量比例进行计算(可通过快速傅里叶变换(FFT)法估计频带能量)，并依据这些频带能量比例对输入的自然声音从高到低进行排序，选取排序靠前的十分位数以内的声音刺激，作为不同睡眠时段对应的最优声音刺激，标记后存储到深度最优声音库当中。
(2)脑电波采集模块
脑电波采集模块由干电极、放大器、A/D转换模块组成。干电极采用Ag/AgCl电极，安置在发卡式头戴装置或由高弹性纤维编织的电极帽等固定装置上。后者在使用时佩戴在用户的头部，以保证电极和头皮的有效接触。与传统Ag/AgCl湿电极相比，干电极无需涂抹导电膏，更有利于用户 随时佩戴与使用。为更好的记录睡眠波，本发明建议采用按照10/20国际EEG系统标准排布的F3、F4两个电极；也可扩展使用更多电极，如增加枕颞区TP7、TP8等电极。放大器可采用分离元器件搭建或使用现有的集成芯片实现，如Analog Device公司的AD-620芯片。作为示例，本发明可采用图4所示的信号放大电路。图4所示电路由前、后两级放大器及一个带通滤波器组成。前级仪表放大器包括IC1、IC2构成的差分放大电路，与IC3构成的减法器。Vin接信号测量电极，Vref接参考电极，模拟地接GND电极。由于运算放大器的“虚短”效应，流经Rg的电流为(Vin-Vref)/Rg；同时，由于“虚断”效应，流经Rg的电流全部通过负反馈电路到达IC1与IC2的输出端，从而得到(V1-V2)＝(Vin-Vref)(1+2R/Rg)。同理，可知IC3构成减法器的输出电压Vp＝(V2-V1)。因此，前级放大器的总体放大倍数为A＝(1+2R/Rg)。为避免运算放大器电流饱和，本发明前级放大电路的放大倍数取5～20，建议取10。带通滤波器可采用巴特沃兹或切比雪夫滤波器，通带范围建议为0.05Hz～40Hz。后级放大器(IC4)放大倍数取50～200，建议取100。本发明需要对多个EEG电极通道同步采集信号，因此根据所用的电极数量，EEG采集模块应包含相应数目的多个单通道放大器。放大后的输出信号Vo通过A/D转换模块转换为数字信号。A/D转换模块将采集的模拟信号转换为数字信号，可采用市场上常用的A/D芯片实现。本发明建议至少16bit的A/D转换精度，采样率建议在128Hz以上。
(3)睡眠监测模块
睡眠监测模块根据记录到的EEG信号，估计受试者当前所处的睡眠时段。本发明采用短时傅里叶变换(short time Fourier transform,STFT)法得到EEG信号s(t)的实时时频图，分别计算梭形波、θ波、高δ波、低δ波能量相对于α波与β波能量的相对值。如果梭形波占优，则认为处于非REM睡眠1期；θ波占优，认为处于非REM睡眠2期；高δ波占优，认为处于非REM睡眠3期；低δ波占优，认为处于非REM睡眠4 期；α波与β波占优，则认为处于REM睡眠/觉醒期。
(4)声音刺激选择模块
声音刺激选择模块根据睡眠监测模块给出的当前的睡眠时段，从声音库中选择一组指定数量的标记为相同睡眠时段的最优声音刺激。如果当前处于非REM睡眠1期，则选择标记为非REM睡眠1期或2期的声音；如果当前处于非REM睡眠2期，则选择标记为非REM睡眠2期或3期的声音；如果当前处于非REM睡眠3期，则选择标记为非REM睡眠3期或4期的声音；如果当前处于非REM睡眠4期，则选择标记为非REM睡眠4期的声音。
(5)闭环优化模块
本发明系统对受试者播放声音刺激，同时接收受试者反馈的EEG信号，因此可根据反馈的EEG信号中睡眠波的强度调整声音刺激，从而形成“声音刺激-实时EEG-声音刺激”的闭环，以有效优化声音刺激，获得针对受试者个体的最优睡眠调节。本发明系统闭环优化模块的一种实现方法为：从声音刺激选择模块给出的一组最优声音刺激选择一个进行播放，在播放声音的同时，从睡眠监测模块获得当前睡眠状态对应EEG频段的实时能量相对值，并以此相对值作为当前睡眠时段下睡眠质量的评估值，记录下来；而后，切换到这一组中下一个声音刺激，记录对应的睡眠质量评估值，直到遍历所有的最优声音刺激；最后，从中选择睡眠质量评估值占优的前若干个指定数量的声音刺激，作为对应受试者个性化的最优声音刺激；一旦已经确定受试者个性化的最优声音刺激，本发明系统可跳过声音刺激选择模块与闭环优化模块，而直接从受试者个性化的最优声音刺激中选择对应睡眠时段的声音刺激进行播放，直到检测到的睡眠质量评估值显著下降(例如下降幅度超过原来的30％)，可重新进行声音刺激选择与闭环优化。
(6)开关/音量控制模块
开关/音量控制根据睡眠监测模块给出的当前的睡眠时段，决定是否播放声音刺激及调节声音刺激的音量。从非REM睡眠1期到4期，播放声 音刺激的音量可逐渐从初始音量(如最大音量的60％)减少到某一预设音量(如最大音量的10％)，并在非REM睡眠结束后(处于REM睡眠或觉醒状态)关闭声音刺激。
(7)声音播放模块
声音播放模块包含一个音频放大器，根据所选择的声音刺激与所设定的音量/开关状态，驱动扬声器播放声音刺激。
本发明的关键点和欲保护点如下：
(1)基于深度神经网络预测EEG活动的系统和方法；
(2)基于深度神经网络的声音刺激选择系统和方法；
(3)基于短时傅里叶变换进行睡眠监测的系统和方法；
(4)基于“声音刺激-实时EEG-声音刺激”的闭环对声音刺激进行优化的系统和方法；
(5)根据睡眠监测状态选择声音刺激的系统和方法；
(6)根据睡眠监测状态控制声音刺激播放音量/开关的系统和方法。
本发明系统和方法不仅限于脑电波EEG信号，同样适用于其他可反映睡眠状态的生理信号，如脑磁图(MEG)，近红外光谱(NIRS)，功能性磁共振(fMRI)等。
本发明系统所采用睡眠监测方法不限于所述的方法，同样可替代为神经网络、支持向量机等基于分类器的监测方法及其他基于功率谱等的睡眠状态判断方法。
本发明系统中的脑电波采集模块、声音刺激选择模块、开关/音量控制模块、声音播放模块、扬声器等均可替换为其他具备同等或类似功能的模块。

Claims (11)

  1. 一种用于睡眠调节的深度声音刺激系统，所述系统至少包括：深度优化声音库模块，所述深度优化声音库模块由自然声音和合成声音中选择最优的声音刺激构成；脑电波采集模块，所述脑电波采集模块用于采集受试者的脑电信号；睡眠监测模块，所述睡眠监测模块根据所述脑电波采集模块记录到的脑电信号，估计受试者当前所处的睡眠时段；声音刺激选择模块，所述声音刺激选择模块根据所述睡眠监测模块给出的所述当前的睡眠时段，从所述深度优化声音库中选择一组指定数量的标记为相同睡眠时段的最优声音刺激；闭环优化模块，所述闭环优化模块根据所述脑电波采集模块采集到的脑电信号中睡眠波的强度调整声音刺激，形成“声音刺激-实时EEG-声音刺激”的闭环，优化声音刺激，以获得针对受试者个体的最优睡眠调节；播放模块，所述播放模块用于将所述个体的最优睡眠调节声音播放给受试者。
  2. 如权利要求1所述的深度声音刺激系统，其特征在于，所述闭环优化模块的一种实现方法为：从声音刺激选择模块给出的一组最优声音刺激选择一个进行播放，在播放声音的同时，从睡眠监测模块获得当前睡眠状态对应EEG频段的实时能量相对值，并以此相对值作为当前睡眠时段下睡眠质量的评估值，记录下来；而后，切换到这一组中下一个声音刺激，记录对应的睡眠质量评估值，直到遍历所有的最优声音刺激；最后，从中选择睡眠质量评估值占优的前若干个指定数量的声音刺激，作为对应受试者个性化的最优声音刺激；一旦已经确定受试者个性化的最优声音刺激，跳过声音刺激选择模块与闭环优化模块，而直接从受试者个性化的最优声音刺激中选择对应睡眠时段的声音刺激进行播放，直到检测到的睡眠质量评估值显著下降(例如下降幅度超过原来的30％)，可重新进行声音刺激选择与闭环优化。
  3. 如权利要求1所述的深度声音刺激系统，其特征在于，深度音乐生成网络使用公开的已训练好的声学模型，或自行建立模型训练得到，深度 EEG预测网络即为深度语音识别网络对EEG活动的最优映射模型。
  4. 如权利要求1所述的深度声音刺激系统，其特征在于，所述深度优化声音库模块，将相同的音乐旋律输入深度语音识别网络，及播放给受试者(用户)，同步记录神经网络中模型各层神经元输出的变化与受试者真实EEG信号的变化，确定模型神经元到真实EEG信号的最优映射关系，建立神经网络对EEG活动预测的最优映射模型。
  5. 如权利要求1所述的深度声音刺激系统，其特征在于，在所述深度优化声音库模块中，将声音刺激输入深度EEG预测网络，得到当前声音刺激下深度网络对EEG信号的估计，称作EEG估计信号，对所述EEG估计信号中表征睡眠不同时段的梭形波、θ波、高δ波、低δ波所占的频带能量比例进行计算，并依据这些频带能量比例对输入的自然声音由高到低进行排序，选取排序靠前的十分位数以内的声音刺激，作为不同睡眠时段对应的最优声音刺激，标记后存储到深度最优声音库当中。
  6. 一种用于睡眠调节的深度声音刺激方法，所述方法至少包括如下步骤：1)采集受试者的脑电信号；2)根据所述脑电波采集模块记录到的脑电信号，估计受试者当前所处的睡眠时段；3)根据所述睡眠监测模块给出的所述当前的睡眠时段，从所述深度优化声音库中选择一组指定数量的标记为相同睡眠时段的最优声音刺激，4)根据所述脑电波采集模块采集到的脑电信号中睡眠波的强度调整声音刺激，形成“声音刺激-实时EEG-声音刺激”的闭环，优化声音刺激，5)将优化后的声音刺激播放给受试者，以获得针对受试者个体的最优睡眠调节。
  7. 如权利要求1所述的深度声音刺激方法，其特征在于，步骤4)的实现方法为：从声音刺激选择模块给出的一组最优声音刺激选择一个进行播放，在播放声音的同时，从睡眠监测模块获得当前睡眠状态对应EEG频段的实时能量相对值，并以此相对值作为当前睡眠时段下睡眠质量的评估值，记录下来；而后，切换到这一组中下一个声音刺激，记录对应的睡眠质量评估值，直到遍历所有的最优声音刺激；最后，从中选择睡眠质量评 估值占优的前若干个指定数量的声音刺激，作为对应受试者个性化的最优声音刺激；一旦已经确定受试者个性化的最优声音刺激，则跳过声音刺激选择模块与闭环优化模块，而直接从受试者个性化的最优声音刺激中选择对应睡眠时段的声音刺激进行播放，直到检测到的睡眠质量评估值显著下降(例如下降幅度超过原来的30％)，可重新进行声音刺激选择与闭环优化。
  8. 如权利要求1所述的深度声音刺激方法，其特征在于，所述深度优化声音库的构成步骤为：首先建立深度神经网络预测EEG活动的映射模型，而后根据所建立的映射模型选择最优声音刺激。
  9. 如权利要求8所述的深度声音刺激方法，其特征在于，建立深度神经网络预测EEG活动的映射模型的步骤为：将相同的音乐旋律输入深度语音识别网络，及播放给受试者(用户)，同步记录神经网络中模型各层神经元输出的变化与受试者真实EEG信号的变化，确定模型神经元到真实EEG信号的最优映射关系，建立神经网络对EEG活动预测的映射模型。
  10. 如权利要求9所述的深度声音刺激方法，其特征在于：所述深度神经网络和所述EEG活动映射关系通过如下方法建立：将深度神经网络第i层神经元的输出向量记作x i(t)，对应的EEG映射权重记作ω i，预测EEG信号记作y i(t)，当前记录到的实际EEG信号记作s(t)，则y i(t)＝<ω i,x i(t)>，其中<·,·>表示向量内积，采用典型相关分析法(canonical correlation analysis，CCA)，评估各y i(t)与s(t)之间的相关性，选取与实际EEG信号相关性最大的模型神经元层及对应的映射权重，作为最优映射，用于深度神经网络对EEG活动的预测。
  11. 一种声音闭环优化模块实现方法，其包括如下步骤：从声音刺激选择模块给出的一组最优声音刺激选择一个进行播放，在播放声音的同时，从睡眠监测模块获得受试者当前睡眠状态对应EEG频段的实时能量相对值，并以此相对值作为当前睡眠时段下睡眠质量的评估值，记录下来；而后，切换到这一组中下一个声音刺激，记录对应的睡眠质量评估值，直到 遍历所有的最优声音刺激；最后，从中选择睡眠质量评估值占优的前若干个指定数量的声音刺激，作为对应受试者个性化的最优声音刺激；一旦已经确定受试者个性化的最优声音刺激，跳过声音刺激选择模块与闭环优化模块，而直接从受试者个性化的最优声音刺激中选择对应睡眠时段的声音刺激进行播放，直到检测到的睡眠质量评估值显著下降(例如下降幅度超过原来的30％)，重新进行声音刺激选择与闭环优化。
PCT/CN2019/123156 2019-11-28 2019-12-05 一种用于睡眠调节的深度声音刺激系统和方法 WO2021103084A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/729,871 US20220249017A1 (en) 2019-11-28 2022-04-26 Deep sound stimulation system and method for sleep regulation

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911191990.0A CN110841169B (zh) 2019-11-28 2019-11-28 一种用于睡眠调节的深度学习声音刺激系统和方法
CN201911191990.0 2019-11-28

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/729,871 Continuation US20220249017A1 (en) 2019-11-28 2022-04-26 Deep sound stimulation system and method for sleep regulation

Publications (1)

Publication Number Publication Date
WO2021103084A1 true WO2021103084A1 (zh) 2021-06-03

Family

ID=69606563

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/123156 WO2021103084A1 (zh) 2019-11-28 2019-12-05 一种用于睡眠调节的深度声音刺激系统和方法

Country Status (3)

Country Link
US (1) US20220249017A1 (zh)
CN (1) CN110841169B (zh)
WO (1) WO2021103084A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023056568A1 (en) * 2021-10-08 2023-04-13 Interaxon Inc. Systems and methods to induce sleep and other changes in user states

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021258245A1 (zh) * 2020-06-22 2021-12-30 华为技术有限公司 更新助眠音频信号的方法及装置
CN111790038B (zh) * 2020-06-23 2024-03-29 深圳市联奕实业有限公司 音乐辅助睡眠的方法、装置、计算机设备及存储介质
CN112121283A (zh) * 2020-09-08 2020-12-25 杭州趣安科技有限公司 一种自适应放松辅助系统、方法、设备以及存储介质
CN113018635B (zh) * 2021-03-08 2023-07-14 恒大新能源汽车投资控股集团有限公司 车辆用户睡眠智能唤醒方法及装置
CN113368365A (zh) * 2021-05-21 2021-09-10 苏州声动医疗科技有限公司 脑功能监测的声音振动调控设备、方法、头枕及头戴设备
CN114366983A (zh) * 2021-12-02 2022-04-19 范建萍 一种改善睡眠质量的方法、系统、装置、电子设备及介质
CN114425119A (zh) * 2021-12-17 2022-05-03 松研科技（杭州）有限公司 一种近红外光谱调节睡眠功能的自动仪
CN114366038B (zh) * 2022-02-17 2024-01-23 重庆邮电大学 基于改进的深度学习算法模型的睡眠信号自动分期方法
CN114652938B (zh) * 2022-02-18 2023-12-26 南京安睡科技有限公司 一种促进睡眠的智能闭环调控刺激系统及使用方法
CN116671875B (zh) * 2023-08-04 2024-04-30 安徽星辰智跃科技有限责任公司 基于小波变换的睡眠可持续性检测调节方法、系统和装置
CN116705247B (zh) * 2023-08-07 2024-04-02 安徽星辰智跃科技有限责任公司 基于局部分解的睡眠可持续性检测调节方法、系统和装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140350706A1 (en) * 2013-05-23 2014-11-27 Yamaha Corporation Sound Generator Device and Sound Generation Method
CN107066801A (zh) * 2011-06-10 2017-08-18 X-系统有限公司 用于分析声音的方法和系统
CN107106063A (zh) * 2014-11-02 2017-08-29 恩戈格勒公司 智能音频头戴式耳机系统
CN107715276A (zh) * 2017-11-24 2018-02-23 陕西科技大学 闭环路径中睡眠状态反馈的声音睡眠控制系统及其方法
CN107998499A (zh) * 2017-11-28 2018-05-08 广州视源电子科技股份有限公司 睡眠辅助内容的处理方法和系统、睡眠辅助服务器系统
US20190246936A1 (en) * 2014-04-22 2019-08-15 Interaxon Inc System and method for associating music with brain-state data

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
SG11201510213UA (en) * 2013-06-11 2016-01-28 Agency Science Tech & Res Sound-induced sleep method and a system therefor
CN103412646B (zh) * 2013-08-07 2016-03-30 南京师范大学 基于脑机交互的音乐情绪化推荐方法
US9924271B2 (en) * 2013-10-02 2018-03-20 Acousticsheep Llc Functional headwear
CN104133879B (zh) * 2014-07-25 2017-04-19 金纽合（北京）科技有限公司 脑电信号与音乐进行匹配的方法及其系统
CN104732983B (zh) * 2015-03-11 2018-03-16 浙江大学 一种交互式音乐可视化方法和装置
CN107126615A (zh) * 2017-04-20 2017-09-05 重庆邮电大学 基于脑电信号的音乐诱导睡眠方法及系统
CN107961429A (zh) * 2017-11-28 2018-04-27 广州视源电子科技股份有限公司 睡眠辅助方法和系统、睡眠辅助装置
US11123009B2 (en) * 2017-12-21 2021-09-21 Koninklijke Philips N.V. Sleep stage prediction and intervention preparation based thereon
CN109107016B (zh) * 2018-08-17 2021-05-25 贵州优品睡眠健康产业有限公司 体感振动音乐助眠系统
CN109523993B (zh) * 2018-11-02 2022-02-08 深圳市网联安瑞网络科技有限公司 一种基于cnn与gru融合深度神经网络的语音语种分类方法
CN109316170B (zh) * 2018-11-16 2020-12-22 武汉理工大学 基于深度学习的脑电波辅助睡眠及唤醒系统
CN109731204A (zh) * 2019-02-13 2019-05-10 深兰科技（上海）有限公司 一种睡眠刺激方法及装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107066801A (zh) * 2011-06-10 2017-08-18 X-系统有限公司 用于分析声音的方法和系统
US20140350706A1 (en) * 2013-05-23 2014-11-27 Yamaha Corporation Sound Generator Device and Sound Generation Method
US20190246936A1 (en) * 2014-04-22 2019-08-15 Interaxon Inc System and method for associating music with brain-state data
CN107106063A (zh) * 2014-11-02 2017-08-29 恩戈格勒公司 智能音频头戴式耳机系统
CN107715276A (zh) * 2017-11-24 2018-02-23 陕西科技大学 闭环路径中睡眠状态反馈的声音睡眠控制系统及其方法
CN107998499A (zh) * 2017-11-28 2018-05-08 广州视源电子科技股份有限公司 睡眠辅助内容的处理方法和系统、睡眠辅助服务器系统

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023056568A1 (en) * 2021-10-08 2023-04-13 Interaxon Inc. Systems and methods to induce sleep and other changes in user states

Also Published As

Publication number Publication date
CN110841169B (zh) 2020-09-25
CN110841169A (zh) 2020-02-28
US20220249017A1 (en) 2022-08-11

Similar Documents

Publication Publication Date Title
WO2021103084A1 (zh) 一种用于睡眠调节的深度声音刺激系统和方法
CN111477299B (zh) 结合脑电检测分析控制的声电刺激神经调控方法及装置
CN105451801B (zh) 声音诱导的睡眠方法和用于该方法的系统
WO2021103829A1 (zh) 基于脑波音乐的个性化精神状态调节系统及调节方法
CN106725462B (zh) 基于脑电信号的声光睡眠干预系统和方法
Wang et al. Representation of a species-specific vocalization in the primary auditory cortex of the common marmoset: temporal and spectral characteristics
CN102973277B (zh) 一种频率跟随响应信号测试系统
CN113303814B (zh) 基于深度迁移学习的单通道耳脑电自动睡眠分期方法
CN110947076B (zh) 一种可进行精神状态调节的智能脑波音乐可穿戴设备
Tseng et al. Emotion recognition of EEG underlying favourite music by support vector machine
CN107802938B (zh) 一种用于镇痛的音乐电刺激的产生方法
CN108543193A (zh) 一种用户状态干预方法及装置
Ogg et al. Psychophysiological indices of music-evoked emotions in musicians
CN116327195A (zh) 一种基于EEG和fNIRS的可穿戴设备及推荐系统
Ramdinmawii et al. The effect of music on the human mind: A study using brainwaves and binaural beats
CN110302462A (zh) 一种基于脑电信号的神经反馈训练系统
CN113171534B (zh) 基于音乐和能量波函数的叠加增强神经调制方法及装置
CN116098619B (zh) 一种基于高频听觉稳态响应的抑郁症自动检测系统
CN113509183A (zh) 一种基于ar人工智能分析情感焦虑抑郁紧张的方法
CN112618911A (zh) 基于信号处理的音乐反馈调节系统
CN115486844A (zh) 用于精神状态监测与干预的音频智能干预系统及其方法
CN114053550B (zh) 一种基于高频心电的耳机式情绪压力调节装置
CN109589112B (zh) 基于iab在线调控睁、闭眼状态的静息脑电采集方法
KR20140104662A (ko) 게임을 이용한 자기 조절능력 측정 방법
Liang et al. Classification of EEG signals from musicians and non-musicians by neural networks

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19954293

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19954293

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 11.01.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 19954293

Country of ref document: EP

Kind code of ref document: A1