Deep sound stimulation system and method for sleep regulation

Patent metadata

Field	Value
Publication number	US-12569188-B2
Application number	US-202217729871-A
Country	US
Kind code	B2
Filing date	Apr 26, 2022
Priority date	Nov 28, 2019
Publication date	Mar 10, 2026
Grant date	Mar 10, 2026

Abstract

The present invention describes a system and method for selecting and optimizing a sound stimulus using a deep neural network to regulate and improve human sleep quality. The deep neural network has the capability of characterizing processing of human brain cortical neurons for external stimulus (images, sounds, etc.) information. By inputting massive sound stimuli into the deep neural network, a sound mode which causes model-estimated sleep electroencephalograph to be optimal can be found, the sound mode is applied to a real human body, and the intensity of corresponding sleep waves of the human body in different sleep stages is enhanced through closed-loop optimization so as to realize the purpose of regulating sleep. The present invention mainly aims at solving the technical problem of how to select and optimize, when a sound stimulus means (music, speech, natural sounds, white/colored noise, etc.) is used to assist in human sleep.

First claim

What is claimed is:

1 . A deep sound stimulation system for sleep regulation, wherein the system at least comprises: a deep optimization sound library module, the deep optimization sound library module consisting of optimal sound stimuli selected from natural sounds and synthetic sounds; an Electroencephalograph (EEG) collection module, the EEG collection module being configured to collect an EEG signal of a subject; a sleep monitoring module, the sleep monitoring module estimating, according to the EEG signal recorded by the EEG collection module, a current sleep stage of the subject; a sound stimulus selection module, the sound stimulus selection module selecting, from the deep optimization sound library and according to the current sleep stage given by the sleep monitoring module, a group of designated number of optimal sound stimuli marked as the same sleep stage; a closed-loop optimization module, the closed-loop optimization module adjusting the sound stimuli according to the intensity of sleep waves in the EEG signal collected by the EEG collection module to form a closed-loop of sound stimulus-real-time EEG-sound stimulus, and optimizing the sound stimuli to obtain an optimal sleep regulation sound for the subject individual; and a playing module, the playing module being configured to play the optimal sleep regulation sound corresponding to the subject to the subject.

2 . The deep sound stimulation system according to claim 1 , wherein the closed-loop optimization module is configured to: select one of the group of optimal sound stimuli given by the sound stimulus selection module for playing, obtain a real-time energy relative value of an EEG frequency band corresponding to a current sleep state from the sleep monitoring module while playing the sound, take the relative value as an evaluation value of the sleep quality in the current sleep stage and record same; then, switch to a next sound stimulus in the group, and record a corresponding sleep quality evaluation value until all optimal sound stimuli are traversed; finally, select, from all the optimal sound stimuli, a designated number of sound stimuli having the highest sleep quality evaluation values as personalized optimal sound stimuli of the corresponding subject; and once the personalized optimal sound stimuli for the subject have been determined, skip the sound stimulus selection module and the closed-loop optimization module, and directly select, from the personalized optimal sound stimuli for the subject, a sound stimulus corresponding to the sleep stage for playing, until the measured sleep quality evaluation value reduces by more than 30% of the original sleep quality evaluation value, so that sound stimulus selection and closed-loop optimization may be re-performed.

3 . The deep sound stimulation system according to claim 1 , wherein the deep optimization sound library module comprises a deep speech recognition network, inputs a same music melody into the deep speech recognition network, plays same to the subject (user), synchronously records a change outputted by each layer of neurons in a neural network model and a change of a real EEG signal of the subject, determines an optimal mapping relationship between the neurons of the model and the real EEG signal, and establishes an optimal mapping model of the neural network for EEG activity prediction.

4 . The deep sound stimulation system according to claim 1 , wherein the deep speech recognition network is trained by a disclosed trained acoustic model or an established model, the optimal mapping model of the neural network for EEG activity prediction forms a deep EEG prediction network.

5 . The deep sound stimulation system according to claim 4 , wherein in the deep optimization sound library module, the sound stimulus is inputted into the deep EEG prediction network to obtain an estimation of the deep network for the EEG signal under the current sound stimulus, which is referred to as an EEG estimation signal; frequency band energy ratios occupied by spindles, θ waves, high-δ waves, and low-δ waves representing different sleep stages in the EEG estimation signal are calculated, and the inputted natural sounds are sorted in descending order according to the frequency band energy ratios; and top-ranking sound stimuli within a decile are selected as optimal sound stimuli corresponding to different sleep stages, and marked and stored into the deep optimization sound library.

6 . A deep sound stimulation method for sleep regulation, wherein the method at least comprises the following steps: 1) collecting an Electroencephalograph (EEG) signal of a subject; 2) estimating, according to the EEG signal recorded by an EEG collection module, a current sleep stage of the subject; 3) selecting, from a deep optimization sound library and according to the current sleep stage given by a sleep monitoring module, a group of designated number of optimal sound stimuli marked as the same sleep stage; 4) adjusting the sound stimuli according to the intensity of sleep waves in the EEG signal collected by the EEG collection module to form a closed-loop of sound stimulus-real-time EEG-sound stimulus, and optimizing the sound stimuli; and 5) playing the optimized sound stimuli to the subject to obtain an optimal sleep regulation sound for the subject individual.

7 . The deep sound stimulation method according to claim 6 , wherein a method for implementing step 4) comprises: selecting one of the group of optimal sound stimuli given by a sound stimulus selection module for playing, obtaining a real-time energy relative value of an EEG frequency band corresponding to a current sleep state from the sleep monitoring module while playing the sound, taking the relative value as an evaluation value of the sleep quality in the current sleep stage and recording same; then, switching to a next sound stimulus in the group, and recording a corresponding sleep quality evaluation value until all optimal sound stimuli are traversed; finally, selecting, from all the optimal sound stimuli, a designated number of sound stimuli having the highest sleep quality evaluation values as personalized optimal sound stimuli of the corresponding subject; and once the personalized optimal sound stimuli for the subject have been determined, skipping the sound stimulus selection module and a closed-loop optimization module, and directly selecting, from the personalized optimal sound stimuli for the subject, a sound stimulus corresponding to the sleep stage for playing, until the measured sleep quality evaluation value reduces by more than 30% of the original sleep quality evaluation value, so that sound stimulus selection and closed-loop optimization may be re-performed.

8 . The deep sound stimulation method according to claim 6 , wherein steps of constructing the deep optimization sound library are: firstly, establishing a mapping model of a deep neural network for EEG activity prediction, and then selecting an optimal sound stimulus according to the established mapping model.

9 . The deep sound stimulation method according to claim 8 , wherein the steps of establishing the mapping model of the deep neural network for EEG activity prediction are: inputting a same music melody into a deep speech recognition network, playing same to the subject, synchronously recording a change outputted by each layer of neurons in the neural network model and a change of a real EEG signal of the subject, determining an optimal mapping relationship between the neurons of the model and the real EEG signal, and establishing the mapping model of the neural network for EEG activity prediction.

10 . The deep sound stimulation method according to claim 9 , wherein the mapping relationship

Assignees

• Shenzhen Inst Of Adv Tech Cas

Inventors

• Zhou Huihui
• Ma Zheng
• Xie Jin

Classifications

• A61B5/374

  Detecting the frequency distribution of signals, e.g. detecting delta, theta, alpha, beta or gamma waves · CPC title

• A61B5/291

  for electroencephalography [EEG] · CPC title

• A61B5/055

  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging · CPC title

• A61B5/0075

  by spectroscopy, i.e. measuring spectra, e.g. Raman spectroscopy, infrared absorption spectroscopy (A61B5/0071 takes precedence) · CPC title

• A61B5/7415

  Sound rendering of measured values, e.g. by pitch or volume variation (A61B5/741 takes precedence) · CPC title