An intelligent sleep regulation method and system based on closed-loop acousto-optic brain entrainment

Abstract

The present application relates to a kind of intelligent sleep regulation method and system based on closed-loop acoustooptic brain wave entrainment, real-time acquisition and preprocessing electroencephalogram (EEG) by wireless dry electrode, high-precision sleep staging is realized using deep learning model.The system uses fuzzy PID controller to dynamically adjust acoustooptic synergistic stimulation: acoustic module generates binaural beat signal, optical module outputs blue light (470nm) and amber light (590nm) pulse, through phase synchronization (phase difference ≤10°) and golden section frequency coupling (f_L=1.618f_A) to enhance brain wave entrainment efficiency.Closed-loop feedback updates parameters every 30 seconds, adjusts sound pressure level (30-50dB) and light intensity (10-100lux) according to Weber-Fechner law, response delay <200ms.Three-level safety mechanism monitors gamma wave anomaly, epileptiform discharge and impedance overrun in real time, triggers graded protection (alarm, cut off light stimulation, shutdown).Clinical verification shows that the success rate of delta wave and theta wave entrainment is 71% and 68% respectively, and the sleep improvement effect is good, providing a safe and efficient intervention scheme for sleep disorders.

Description

Technical Field

[0001] This patent is based on the interdisciplinary integration of neuroscience, biomedical engineering and electronic technology, and constructs a closed-loop sleep regulation system with brainwave entrainment as the core and multimodal synergistic stimulation as the means. Background Technology

[0002] The neural modulation mechanism of brainwave entrainment involves guiding the oscillation frequency of the cerebral cortex to synchronize with the target frequency band (delta waves: 0.5-4Hz, theta waves: 4-8Hz) through combined acoustic and optical stimulation. The acoustic module employs binaural beats, inputting pure tones with a frequency difference Δf to each ear (e.g., 200Hz in the left ear and 204Hz in the right ear can induce 4Hz theta waves). Through auditory nerve integration in the superior olivary nucleus, an EEG rhythm consistent with Δf is generated in the thalamus. The optical module, via the retina-hypothalamus pathway, uses periodic pulse stimulation of specific wavelengths (470nm blue light inhibits melatonin, 590nm amber light promotes pineal gland regulation) to directly act on the circadian clock neurons in the suprachiasmatic nucleus. The synergistic effect of both significantly improves brainwave entrainment efficiency; experimental data show that its synchronization success rate is 42% higher than that of the single modality (p<0.05).

[0003] This patent utilizes a multimodal spatiotemporal collaborative optimization model to overcome the limitations of traditional devices that separate sound and light stimuli, establishing a collaborative control model based on phase-frequency coupling. The control model includes time and spectral dimensions. In the time dimension, phase synchronization technology ensures that the phase difference between the sound pulse and the light flicker is ≤10°, avoiding multisensory conflict. In the spectral dimension, the light stimulation frequency f_L and the sound stimulation frequency f_A are set to satisfy the golden ratio (f_L=1.618f_A), utilizing the resonance characteristics of the auditory and visual pathways to enhance the entrainment effect. Based on the Weber-Fechner law, the light intensity (10-100 lux) and sound pressure level (30-50 dB) are dynamically adjusted to maintain a logarithmic linear relationship between stimulation intensity and neural response. Simultaneously, a closed-loop feedback control system is established, using a prefrontal cortex dry electrode array to collect EEG signals in real time and combining them with an adaptive algorithm to achieve dynamic adjustment. Features include: sleep stage identification: based on the improved AASM standard, millisecond-level stage classification is achieved through delta wave power ratio (>20% to determine deep sleep) and beta wave burst inhibition detection (to identify the wakefulness state); dynamic parameter adjustment: using a fuzzy PID controller, stimulation parameters are adjusted in real time according to the current brainwave frequency error (±0.3Hz accuracy) and phase coherence coefficient (0-1 scale), with a response delay of <200ms; and a safety protection mechanism that automatically performs safety processing when abnormal high-frequency activity is detected to avoid the risk of inducing epilepsy.

[0004] This patent addresses the problems of traditional open-loop preset parameters that ignore individual physiological differences, asynchronous acoustic and optical signals, brainwave entrainment efficiency of only 19-32%, reliance on offline analysis, and delays exceeding 5 minutes. This patent uses a closed-loop feedback system that updates stimulation parameters every 30 seconds to achieve individualized solutions; a spatiotemporal coupling algorithm enables multimodal phase locking, increasing the entrainment success rate to an average of 67%; and an edge computing architecture achieves 200ms-level real-time response, ensuring real-time intervention.

[0005] This patent is the first to introduce closed-loop control theory into the field of multimodal brainwave entrainment, by establishing a real-time control loop of "perception-decision-execution" (such as...). Figure 1 As shown in the figure, this system solves three major problems in traditional techniques: rigid stimulation parameters, insufficient modal coordination, and delayed intervention. Clinically validated, this system can shorten the sleep latency of insomnia patients from 34.2±12.1 minutes to 21.5±8.7 minutes (p=0.008) and improve sleep efficiency by 19.3%, providing a revolutionary technical approach for non-pharmacological intervention of sleep disorders. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes an intelligent sleep regulation method and system based on closed-loop audio-visual brainwave entrainment. This method identifies sleep stages (Wake / N1 / N2 / N3 / REM stages) using EEG (Electroencephalogram) sleep staging and induces sleep for different stages. To ensure safety, a real-time monitoring bypass system is introduced to monitor brainwave status in real time. When real-time monitoring of energy changes in the gamma band (30-100Hz) is conducted, a primary alarm is triggered if the energy exceeds the baseline value by 3 standard deviations for 5 seconds. When epileptiform discharges (sharp wave frequency > 3Hz and amplitude > 200μV) are detected, a secondary intervention is initiated, cutting off the light stimulation output. When impedance abnormalities (> 50kΩ) persist for 10 seconds, a tertiary protection level is activated, and the system automatically shuts down. The control component uses a fuzzy PID controller for photoelectric frequency output control, and the phase synchronization subsystem and frequency coupling subsystem perform brainwave entrainment neuromodulation based on the input.

[0007] The technical solution provided by this invention includes the following:

[0008] This application provides an intelligent sleep regulation method based on closed-loop audio-visual brainwave entrainment, comprising: extracting human brainwaves; determining sleep stages based on the acquired brainwave signals; determining in real time whether a monitoring target has appeared based on the acquired brainwave signals; determining a fuzzy PID (Proportional-Integral-Derivative) controller based on the determined sleep stages to perform audio-visual frequency control; and determining phase synchronization and frequency coupling based on the determined fuzzy PID to complete brainwave entrainment.

[0009] In some implementation examples, brainwaves are extracted from the human brain; the raw brainwave signal is determined based on an 8-channel wireless dry electrode array; the preprocessed brainwave signal is determined based on the raw signal; and the brainwave frequency band decomposition and the brainwave energy spectrum of each frequency band are determined based on the preprocessed brainwave signal.

[0010] In some implementation examples, the acquired brainwave signals are segmented into sleep stages; sleep stages are also segmented according to brainwave frequency bands.

[0011] In some implementation examples, the acquired brainwave signals are used to determine in real time whether the monitored target has appeared; to determine the energy changes in the gamma band (30-100Hz) based on the acquired brainwave signals; to determine epileptiform discharges based on the acquired brainwave signals; and to determine impedance abnormalities based on the acquired brainwave signals.

[0012] In some implementation examples, a fuzzy PID controller is used to control the sound and light frequencies based on the defined sleep stages; the sound frequency adjustment amount Δf is determined based on a fuzzy rule base. A (Universe of discourse: ±0.5Hz); Determine the light intensity correction coefficient K based on the fuzzy rule base. L (Domain of discourse: 0.5-1.5);

[0013] In some implementation examples, phase synchronization and frequency coupling are determined based on a defined fuzzy PID to complete brainwave entrainment; based on the fuzzy PID...

[0014] A phase synchronization algorithm ensures that the phase difference between the audio-visual stimulus and brainwaves is ≤10°, enhancing the entrainment effect; frequency coupling is determined based on the golden ratio method.

[0015] Intensity co-optimization is performed based on the determined phase synchronization and frequency coupling; the acousto-optic output is determined based on the determined intensity co-optimization results.

[0016] The beneficial effects of this invention are as follows: This application, through a closed-loop sound-optic brainwave entrainment method, can effectively improve the success rate of brainwave entrainment. Experiments show that the success rate of brainwave entrainment is significantly higher than that of pure sound stimulation and pure light stimulation; specific comparative data are shown in Table 1.

[0017]

[0018] Table 1

[0019] Meanwhile, by monitoring brainwave activity in real time, it is possible to provide personalized and targeted brainwave-assisted neuromodulation solutions, solving the problem that traditional solutions ignore individual physiological differences by using open-loop preset parameters. Attached Figure Description

[0020] The scope of this disclosure can be better understood by reading the following detailed description of exemplary embodiments in conjunction with the accompanying drawings. The accompanying drawings are:

[0021] Figure 1 A flowchart illustrating an intelligent sleep regulation method based on closed-loop acoustic-optical brainwave entrainment, provided in an embodiment of the present invention.

[0022] Figure 2 A flowchart of an EEG signal processing method provided in an embodiment of the present invention;

[0023] Figure 3 EEG acquisition site diagram provided for embodiments of the present invention (international 10-20 system standard);

[0024] Figure 4 A flowchart of real-time EEG monitoring provided in an embodiment of the present invention;

[0025] Figure 5 A flowchart of a fuzzy PID control of acoustic-optical frequency is provided as an embodiment of the present invention;

[0026] Figure 6 A flowchart of a brainwave-entrained neural modulation method using phase synchronization and frequency coupling is provided as an embodiment of the present invention. Detailed implementation method:

[0027] In practical applications, traditional open-loop preset parameter schemes ignore individual physiological differences, resulting in a low success rate of brainwave-enhanced neuromodulation. To improve the success rate of brainwave-enhanced neuromodulation, this invention designs a closed-loop brainwave-enhanced neuromodulation method. This method can monitor brainwave changes in real time and adaptively adjust the scheme parameters, thereby improving the success rate of brainwave-enhanced mental modulation. Simultaneously, to ensure safety, a safety alarm bypass system is designed to promptly alert and automatically handle potential safety issues. This scheme improves the success rate of brainwave-enhanced mental modulation while ensuring safety, and can be deployed in embedded computers and dedicated neural network accelerators.

[0028] The brainwave-assisted mental modulation method includes:

[0029] Brainwaves are extracted from the human brain; sleep stages are determined based on the acquired brainwave signals; the presence of a monitoring target is determined based on the acquired brainwave signals; a fuzzy PID controller is used for sound and light frequency control based on the determined sleep stage; and phase synchronization and frequency coupling are determined based on the determined fuzzy PID controller to complete brainwave entrainment.

[0030] Step S1:

[0031] In some implementation examples, brainwaves are extracted from the human brain; the raw brainwave signal is determined based on an 8-channel wireless dry electrode array; a preprocessed brainwave signal is determined based on the raw signal; and brainwave frequency band decomposition and the energy spectrum of each frequency band are determined based on the preprocessed brainwave signal. The brainwave processing flow is described below. Figure 2 .

[0032] Step S11: Extract EEG data using dry electrodes at Fp1, Fp2, F3, F4, C3, C4, P3, and P4. Electrode locations are shown in [reference needed]. Figure 3 .

[0033] Step S12: Determine the preprocessed EEG signal based on the original signal.

[0034] Step S121: Filter using a Butterworth filter. The transfer function H(z) can be transformed using a bilinear transform. Step S122: Use a notch filter to eliminate power frequency interference (50Hz or 60Hz) and its harmonics. Use a narrowband stop filter based on IIR (Infinite Impulse Response). Its parameters include a center frequency f. notch =50Hz, control bandwidth narrowness (quality factor Q=30, bandwidth) The transfer function (second-order IIR notch filter) formula is as follows: (f s (where r is the sampling rate), and r is the pole radius (the stopband is narrower when it is close to 1, such as r = 0.99).

[0035] Step S123: To eliminate electromyography artifacts or motion interference noise, it is also necessary to dynamically track and eliminate the noise. The LMS algorithm formula is: error signal e[n] = d[n] - w T [n]x[n], d[n] is the noisy signal, x[n] is the reference noise signal, and w[n] is the adaptive weight vector. The weight update formula is w[n+1]=w[n]+μ·e[n]·x[n], where μ is the step size factor (which needs to be adjusted to avoid divergence).

[0036] Step S13: Determine the brainwave frequency band decomposition and the brainwave energy spectrum of each frequency band based on the preprocessed EEG signal. Real-time feature extraction is performed using Short Time Fourier Transform (STFT) with a 2-second window (Hamming window) and a 50% overlap rate. Key parameters are extracted, including delta waves (0.5-4Hz) and their power percentage. θ / β wave power ratio Extracting the instantaneous phase using Hilbert transform The formula for calculating phase coherence is: Step S2:

[0037] In some implementation examples, the acquired brainwave signals are segmented into sleep stages; sleep stages are also segmented according to brainwave frequency bands.

[0038] Step S21: Segment sleep based on brainwave frequency bands. Use deep learning for sleep staging, employing a hybrid model (CNN+LSTM) that combines spatiotemporal features for training and staging. Example of CNN layer formula. x t+k Given the input signal (time step), y t The output feature map value is given by , where 'b' is the bias term, which is the data offset of the decision surface relative to the origin. Classification accuracy: Wake / N1 / N2 / N3 / REM. Experimental results show that the five-stage classification accuracy reaches 87.3%.

[0039] Step S3:

[0040] In some implementation examples, the presence of the monitored target is determined in real time based on the acquired brainwave signals; the process is described below. Figure 4 Determine energy changes in the gamma band (30-100Hz) based on the acquired brainwave signals; determine epileptiform discharges based on the acquired brainwave signals; determine impedance abnormalities based on the acquired brainwave signals;

[0041] Step S31: Monitor the energy change in the γ band (30-100Hz) in real time, and trigger a primary alarm when the energy exceeds the baseline value by 3 standard deviations for 5 consecutive seconds.

[0042] Step S32: When epileptiform discharge (sharp frequency > 3 Hz and amplitude > 200 μV) is detected, a secondary intervention is initiated to cut off the light stimulation output.

[0043] Step S33: When the impedance abnormality (>50kΩ) lasts for 10 seconds, the third-level protection is activated and the brainwave-entrained neural modulation is turned off.

[0044] Step S4:

[0045] In some implementation examples, a fuzzy PID controller is used for sound and light frequency control based on the determined sleep stages. The process is as follows: Figure 5The audio frequency adjustment amount Δf is determined based on the fuzzy rule base. A (Universe of discourse: ±0.5Hz); Determine the light intensity correction coefficient K based on the fuzzy rule base. L (Domain of discourse: 0.5-1.5);

[0046] Step S41: Based on the determined sleep stage segmentation, determine the fuzzy PID controller for sound and light frequency control, establish a fuzzy rule base, and calculate the matching degree (activation strength) of its preconditions. Assume that the preconditions are connected by logical "AND", and the activation strength is the minimum value or product of the membership degree of the input variables.

[0047] Example:

[0048] Rule 1: IFe_fis Positive(μ1) AND Cis Low(μ2) activation intensity w1=min(μ1,μ2)

[0049] Rule 2: If IFe_f is Zero (μ3) AND C is High (μ4), then the live intensity w2 = min (μ3, μ4).

[0050] Rule S42: Determine the audio adjustment amount Δf based on the fuzzy rule base. A (Universe of discourse: ±0.5Hz), for the output variable (Δf) A The activation strength of each rule is multiplied by its corresponding output value to obtain the weighted output Δf. A weighted sum Sum of activation strengths of all rules ∑w i =w1+w2; The final output value is calculated by weighted average.

[0051] Rule S43: Determine the light intensity correction coefficient K based on the fuzzy rule base. L (Universe of discourse: 0.5-1.5), for the output variable (K) L The activation strength of each rule is multiplied by its corresponding output value to obtain the weighted output Δf. A weighted sum Sum of activation strengths of all rules ∑w i =w1+w2; The final output value is calculated by weighted average.

[0052] Step S5:

[0053] In some implementation examples, phase synchronization and frequency coupling are determined based on a defined fuzzy PID controller to complete brainwave-entrained neural modulation; the process is as follows. Figure 6The phase synchronization algorithm is determined based on fuzzy PID to ensure that the phase difference between the audio-visual stimulus and brain waves is ≤10°, thereby enhancing the entrainment effect. The frequency coupling is determined based on the golden ratio method. The intensity synergy optimization is performed based on the determined phase synchronization and frequency coupling. The audio-visual output is determined based on the determined intensity synergy optimization results.

[0054] S51: The phase synchronization algorithm is determined based on fuzzy PID to ensure that the phase difference between the sound and light stimulation and brain waves is ≤10°, thereby avoiding sensory conflict and enhancing the brain wave entrainment effect.

[0055] S52: Frequency coupling is determined according to the golden ratio method. The optical frequency and the audio frequency are set according to the golden ratio (f_L=1.618f_A) to enhance the resonance effect.

[0056] S53: Intensity synergy optimization is performed based on the determined phase synchronization and frequency coupling. The sound pressure level formula is applied according to the Weber-Fechner law.

[0057] SPL (dB) = 20log10(P / P0) + k × ln(EEG_amplitude), and the light intensity formula I (lux) = I0 × e^(α × δ_power) (where α is the attenuation coefficient and δ_power is the percentage of delta wave power) is used for dynamic adjustment. The adjustment strategy is to calculate the current energy distribution of the EEG band every 30 seconds; if the percentage of delta waves is >20%, the light intensity is reduced exponentially (to avoid excessive suppression); if the coherence of theta waves is <0.6, the sound pressure level is increased linearly (maximum not exceeding 50dB).

[0058] S54: Determines the audio-visual output based on the determined intensity co-optimization results. Acoustic module: Binaural beat synthesis, real-time generation of sine waves, frequency resolution 0.1Hz, dynamic range control (DRC): automatic gain adjustment based on ambient noise (microphone feedback); Optical module: LED driver constant current source (0-50mA), PWM dimming frequency 1kHz, pulse waveform rise time <1ms, duty cycle adjustable from 10% to 30%.

Claims

1. An intelligent sleep regulation system based on closed-loop audio-visual brainwave entrainment, characterized in that, Includes the following modules: The EEG signal acquisition module is configured to acquire EEG signals in real time via an 8-channel wireless dry electrode array, including Fp1, Fp2, F3, F4, C3, C4, P3 and P4. The signal processing module is configured to preprocess the raw EEG signal, including Butterworth bandpass filtering, notch filtering and LMS adaptive filtering to eliminate EMG artifacts, wherein the Butterworth bandpass filtering has a frequency range of 0.5-30Hz, and the notch filtering has a center frequency of 50Hz and a quality factor Q of 30. Feature extraction module: configured to decompose the delta wave, theta wave, beta wave and gamma wave brainwave frequency bands through short-time Fourier transform, and calculate the power ratio and phase coherence coefficient of each frequency band; Sleep staging module: configured to classify brainwave signals into sleep stages based on a hybrid deep learning model, wherein the sleep stages include Wake, N1, N2, N3 and REM sleep; Dynamic control module: configured to adjust the audio-visual stimulation parameters based on sleep stages and real-time brain wave characteristics using a fuzzy PID controller, including: Acoustic module: Generates binaural beat signals, where the frequency difference Δf is the target frequency band, and the frequency resolution is 0.1Hz; Optical module: Outputs periodic light pulses, the wavelengths of which include 470nm blue light and 590nm amber light, the light intensity range is 10 to 100 lux, and the duty cycle range is 10% to 30%; Phase synchronization unit: ensures that the phase difference between the audio-visual stimulation and brain waves is less than or equal to 10°; Frequency coupling unit: The optical frequency and the audio frequency are set according to the golden ratio of f_L = 1.618 × f_A; Safety protection module: Configured to monitor gamma wave energy, epileptiform discharges, and impedance anomalies in real time, and trigger a three-level protection mechanism: Primary alert: Gamma wave energy exceeds the baseline value by 3 standard deviations for 5 seconds; Secondary intervention: When a spike frequency greater than 3 Hz and an amplitude greater than 200 μV are detected, the light stimulation is cut off; Level 3 protection: Impedance abnormality, i.e., the system shuts down when the impedance is greater than 50 kΩ for 10 seconds.

2. The intelligent sleep regulation system according to claim 1, characterized in that: The transfer function of the notch filter is: , r=0.99, where The signal sampling rate, =50Hz.

3. The intelligent sleep regulation system according to claim 1, characterized in that: The input variables of the fuzzy PID controller include brainwave frequency error. The domain of the brainwave frequency error is ±0.5Hz, the phase coherence coefficient C has a domain of 0-1, and the output variable is the audio frequency adjustment amount. and light intensity correction coefficient The control rule base calculates the final output using a weighted average method.

4. The intelligent sleep regulation system according to claim 1, characterized in that: The dynamic control module dynamically adjusts the sound pressure level and light intensity according to the Weber-Fechner law, wherein the sound pressure level is adjustable from 30 dB to 50 dB, and the light intensity is adjusted from 10 lux to 100 lux, as specified in the formula: The sound pressure level formula is SPL(dB) = 20log10(P / P0)+k×ln(EEG_amplitude); Light intensity formula I(lux) = I0×e^(α× δ _power).

Images