Sleep slow wave and spindle wave enhancement method and system based on sound wave stimulation

Abstract

The invention discloses a sleep slow wave and spindle wave enhancement method and system based on sound wave stimulation, and relates to the technical field of sleep quality improvement. The sleep slow wave and spindle wave enhancement method based on sound wave stimulation comprises the following steps: acquiring brain wave data and preprocessing the brain wave data; feature extraction is carried out on the preprocessed brain wave data, sleep stage classification is carried out based on a deep neural network, and a slow wave rising phase is identified; if the current sleep feature is in the slow wave rising phase, determining a sound wave stimulation scheme based on the current sleep feature and the personalized parameters of the user so as to enhance the sleep slow wave and spindle wave, and solving the problems that the existing method is not targeted enough, does not fully consider individual differences and cannot better meet the requirements of different users.

Description

Technical Field

[0001] The present invention relates to the technical field of improving sleep quality, and in particular to a method and system for enhancing sleep slow waves and spindle waves based on sound wave stimulation. Background Art

[0002] Existing technical means to enhance sleep slow waves and spindles mainly include transcranial magnetic stimulation technology, which regulates neuronal activity by acting on the cerebral cortex through magnetic fields, thereby enhancing sleep slow waves and spindles; there are also drug-assisted means, such as the use of benzodiazepines under the guidance of a doctor. Sedative hypnotic drugs such as sedatives can help regulate sleep to increase slow waves and spindle waves. In addition, there are also methods such as music therapy, which can relax the brain nerves by playing music of specific frequencies and rhythms, creating a brain state that is conducive to the appearance of slow waves and spindle waves, thereby enhancing the effect to a certain extent.

[0003] However, the existing enhancement methods are not targeted enough, do not fully consider individual differences, and cannot better meet the needs of different users. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a method and system for enhancing sleep slow waves and spindle waves based on sound wave stimulation, which solves the problems that the existing methods are not targeted enough, do not fully consider individual differences, and cannot better meet the needs of different users.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for enhancing sleep slow waves and spindles based on sound wave stimulation, comprising the following steps: collecting and preprocessing brain wave data; extracting features from the preprocessed brain wave data, and classifying sleep stages based on a deep neural network, and identifying the rising phase of slow waves; if the sleep state is in the rising phase of slow waves, determining a sound wave stimulation scheme based on current sleep characteristics and user personalized parameters to enhance sleep slow waves and spindles.

[0006] Furthermore, the EEG data collection and preprocessing includes the following steps: collecting the initial EEG signal based on the electrode sheet, amplifying the initial EEG signal through the amplifier and outputting the amplified EEG signal V1; filtering the amplified EEG signal V1: performing low-pass filtering on the slow wave to filter out the signal with a frequency higher than the set cutoff frequency f c1 The interference signal is obtained to obtain the slow wave related signal; the spindle wave is subjected to bandpass filtering and the set frequency range [f c2 ,f c3 ], where f c2 and f c3They are the minimum frequency and the maximum frequency of the set frequency range respectively; superpose the slow-wave related signal and the spindle-wave related signal to obtain the filtered electroencephalogram signal; the analog-to-digital converter samples the filtered electroencephalogram signal based on the set sampling rate, converts the analog filtered electroencephalogram signal into a digital filtered electroencephalogram signal, and obtains a digital electroencephalogram data sequence, which is the preprocessed electroencephalogram data.

[0007] Further, feature extraction is performed on the preprocessed electroencephalogram data, including the following steps: decompose the preprocessed electroencephalogram data based on the empirical mode decomposition algorithm to obtain multiple intrinsic mode function components IMF i (t), where i is the number of the intrinsic mode function component, i = 1, 2, 3, …, m, and m is the total number of the intrinsic mode function components; perform wavelet transform time-frequency analysis on the decomposed intrinsic mode function components to obtain the time-frequency coefficient matrix W i (a, b).

[0008] Further, sleep stage classification is performed based on a deep neural network, including the following steps: construct a multi-layer perceptron deep learning neural network model, including an input layer, multiple hidden layers and an output layer, where the number of nodes in the hidden layer is N1, and the number of nodes in the output layer is N2, and N2 corresponds to the number of different sleep stage categories; train based on the backpropagation algorithm, the activation function is the ReLU function, and the loss function is the cross-entropy loss function; input the time-frequency coefficient matrix W i (a, b) into the trained multi-layer perceptron deep learning neural network model to obtain the probabilities P of each sleep stage category, P = [p1, p2, p3,..., p N2 , ], p1, p2, p3, p N2 are the probabilities of the first sleep stage category, the second sleep stage category, the third sleep stage category and the N2th sleep stage category respectively, determine the maximum probability value p max , and the sleep stage category corresponding to this maximum probability value p max is the predicted sleep stage category.

[0009] Further, the identification of the slow-wave rising phase includes the following steps: when the sleep stage category is judged to be the deep sleep stage and the corresponding probability p ds in the deep sleep stage is greater than the set probability threshold, perform the identification of the slow-wave rising phase; obtain the preprocessed electroencephalogram data and extract the slow-wave related signal; calculate the slow-wave signal amplitude A(t) based on the slow-wave related signal; within the set time window, calculate the slow-wave amplitude change rate r(t) based on the slow-wave signal amplitude within the window, Δt is the time interval; when the slow-wave amplitude change rate r(t) is greater than 0 and the duration exceeds the set duration threshold, it is judged to be in the slow-wave rising phase.

[0010] Further, a sound wave stimulation scheme is determined based on the current sleep characteristics and the user's personalized parameters, including the following steps: When it is judged to be in the slow wave rising phase, obtain the slow wave real-time frequency f SW and the spindle wave density D sp ; Denote the slow wave real-time frequency f SW , the spindle wave density D sp and the slow wave signal amplitude A(t) as the current sleep characteristics; Integrate the current sleep characteristics and the user's personalized parameters to construct a feature vector F, where the user's personalized parameters include the response data to different sound wave stimulations during sleep monitoring, the user's age, gender, and sleep habit SlHa; Perform particle optimization based on the particle swarm optimization algorithm to determine the optimal sound wave stimulation parameters, where the optimal sound wave stimulation parameters include the sound wave signal frequency f st and the sound wave signal intensity I st ; Generate sound wave signals corresponding to each sound wave signal type based on the optimal sound wave stimulation parameters; Evaluate the influence of the generated sound wave signals on sleep slow waves and spindle waves in a simulation environment; Determine the sound wave signal corresponding to the optimal sound wave signal type.

[0011] Further, the sleep habit SlHa is obtained based on the sleep latency parameter, the body activity parameter during sleep, and the body temperature parameter during sleep, and the process is as follows: Obtain the sleep latency parameter, where the sleep latency parameter includes the average sleep latency qSt and the standard deviation of sleep latency σ qSt ; Obtain the body activity parameter during sleep, where the body activity parameter during sleep includes the total number of body activities, the number of micro-movements, and the number of turns during each sleep; Obtain the number of body activities per hour SZt, the average number of micro-movements per hour SWt, and the average number of turns per hour SFt based on the body activity parameter during sleep; Obtain the body temperature parameter during sleep, where the body temperature parameter during sleep includes the starting body temperature, the ending body temperature, as well as the highest body temperature, the lowest body temperature, and the body temperature change range during sleep; Perform mean processing on the body temperature parameter during sleep to obtain the average starting body temperature Swq, the average highest body temperature Swg, the average lowest body temperature Swd, and the average body temperature change range Swf; Based on the sleep habit processing model, obtain the sleep habit SlHa:

[0012] SlHa = f1 + f2 + f3;

[0013]

[0014] where f1, f2, and f3 are all transfer functions, cSt is the specified average sleep latency, σ cStLet Std be the standard deviation of the sleep latency, CZt be the number of body movements per hour, CWt be the number of micro-movements per hour on average, CFt be the number of turns per hour on average, Cwq be the average starting body temperature, Cwg be the average highest body temperature, Cwd be the average lowest body temperature, and Cwf be the average amplitude change of body temperature.

[0015] Further, to evaluate the influence of the generated acoustic wave signals on sleep slow waves and spindle waves in a simulated environment, the following steps are included: Input the current sleep characteristics and user personalized parameters into the sleep simulation model, and initialize the running state of the sleep simulation model; In the simulated environment, based on the determined time combination, input each type of acoustic wave signal generated based on the optimal acoustic wave stimulation parameters into the sleep simulation model after initializing the running state, and obtain the simulated electroencephalogram data. The time combination includes the signal application start time and the signal duration; Obtain the average instantaneous amplitude mA of the simulated slow waves SW and the density mD of the simulated spindle waves SP ; Based on the average instantaneous amplitude mA of the simulated slow waves SW and the average amplitude sA of the slow wave signals of the patient S W obtain the relative change rate RA of the slow wave amplitude SW , Based on the density mD of the simulated spindle waves SP and the density sD of the spindle waves of the patient SP obtain the relative change rate RD of the spindle wave density SP , Obtain the evaluation coefficient Px based on the relative change rate of the slow wave amplitude and the relative change rate of the spindle wave density; Determine the smallest evaluation coefficient, obtain the acoustic wave signal implementation plan corresponding to the smallest evaluation coefficient. The acoustic wave signal implementation plan includes the application start time and the duration, and each type of acoustic wave signal generated based on the optimal acoustic wave stimulation parameters.

[0016] Further, the method for obtaining the evaluation coefficient Px is as follows: Px = α1 * RA SW + α2 * RD SP ; where α1 is the weight factor of RA SW , and α2 is the weight factor of RD SP .

[0017] A sleep slow wave and spindle wave enhancement system based on acoustic wave stimulation, which is used for the above-mentioned sleep slow wave and spindle wave enhancement method based on acoustic wave stimulation, includes a data preprocessing module, a slow wave rising phase recognition module, and an acoustic wave stimulation scheme determination module, where: the data preprocessing module is used to collect electroencephalogram data and perform preprocessing; the slow wave rising phase recognition module is used to extract features from the preprocessed electroencephalogram data, classify sleep stages based on a deep neural network, and recognize the slow wave rising phase; the acoustic wave stimulation scheme determination module is used to, when in the slow wave rising phase, determine an acoustic wave stimulation scheme based on the current sleep characteristics and the user's personalized parameters to enhance sleep slow waves and spindle waves.

[0018] The present invention has the following beneficial effects:

[0019] The sleep slow wave and spindle wave enhancement method and system based on acoustic wave stimulation accurately obtain the basic information reflecting the sleep state by collecting and preprocessing electroencephalogram data, providing a reliable basis for subsequent analysis. Through feature extraction and sleep stage classification, especially the recognition of the slow wave rising phase, the key stages during sleep can be accurately grasped. Based on this, an acoustic wave stimulation scheme is determined by combining the current sleep characteristics and the user's personalized parameters, realizing personalized and precise intervention.

[0020] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages simultaneously. Description of the Drawings

[0021] Figure 1 It is a flowchart of the sleep slow wave and spindle wave enhancement method based on acoustic wave stimulation of the present invention.

[0022] Figure 2 It is a block diagram of the sleep slow wave and spindle wave enhancement system based on acoustic wave stimulation of the present invention. Detailed Embodiments

[0023] Please refer to Figure 1 , the embodiment of the present invention provides a technical solution: a sleep slow wave and spindle wave enhancement method based on acoustic wave stimulation, including the following steps: collecting electroencephalogram data and performing preprocessing:

[0024] Collect the initial electroencephalogram signal based on electrode patches, and wear a well-equipped electrode patch system. The electrode patches are in contact with the scalp, and the weak electrical signals generated by the activities of brain neurons are used as input signals, and their voltage amplitudes are usually between 1 - 100 μV. The electrode patches collect the electroencephalogram signal and transmit the signal to the acquisition device through connecting wires. The preamplifier inside the acquisition device starts to work and preliminarily amplifies the input signal. The amplification factor of the preamplifier is set to A1, and generally takes a value between 100 - 1000. According to the amplification principle of the amplifier, the calculation formula for the amplified electroencephalogram signal V1 is: V1 = Vin *A1, V in is the voltage amplitude of the initial electroencephalogram signal collected by the electrode patch.

[0025] The initial electroencephalogram signal is amplified by an amplifier and then the amplified electroencephalogram signal V1 is output, with the amplitude being increased, which is more convenient for subsequent processing. The signal is transmitted to the filter. Through the amplification process, the amplitude of the electroencephalogram signal is significantly increased, enabling subsequent operations such as filtering and analysis to capture and process signal features more clearly, avoiding information loss due to too weak signals, and providing a more reliable data basis for accurately judging the sleep state. The amplified electroencephalogram signal V1 is filtered: a bandpass filter bank is used for filtering, which is divided into a low-pass filter for slow waves (0.5 - 4 Hz) and a bandpass filter for spindle waves (11 - 16 Hz). Both the low-pass filter and the bandpass filter are designed using Butterworth low-pass filters.

[0026] Low-pass filtering is performed on the slow waves to filter out interference signals above the set cut-off frequency f c1 to obtain slow-wave related signals; the low-pass filtering process can accurately filter out interference signals above the set cut-off frequency and only retain the signal components related to slow waves. This can highlight the slow-wave characteristics, reduce the interference of other frequency signals, improve the accuracy of subsequent slow-wave feature analysis, and help to more accurately judge the activity of slow waves during sleep. Bandpass filtering is performed on the spindle waves to retain spindle-wave related signals within the set frequency range [f c2 , f c3 , where f c2 and f c3 are respectively the minimum and maximum frequencies of the set frequency range; the slow-wave related signals and the spindle-wave related signals are superimposed to obtain the filtered electroencephalogram signal; in this way, the spindle-wave signals can be clearly separated, providing a pure signal source for analyzing parameters such as the characteristics and density of spindle waves, making the subsequent research and analysis of spindle waves more accurate and reliable, and thus providing strong support for judging the sleep state and formulating intervention measures.

[0027] The analog-to-digital converter samples the filtered electroencephalogram signal based on the set sampling rate, converts the analog filtered electroencephalogram signal into a digital filtered electroencephalogram signal, obtaining a digital electroencephalogram data sequence, which is the preprocessed electroencephalogram data. The preprocessed electroencephalogram data is stored in a local storage device in digital form and is transmitted to the background data processing terminal in real time through a wireless transmission module. Each data point is attached with accurate timestamp information, providing basic data for subsequent analysis.

[0028] Weak initial electroencephalogram (EEG) signals are collected through electrode patches. After amplification, filtering, and analog-to-digital conversion, preprocessed EEG data is obtained. This step can extract effective signals containing slow waves and spindles from complex brain electrical activities, remove interference, and provide a reliable data basis for subsequent accurate analysis of sleep states. If the EEG data is inaccurate or interfered, subsequent judgments on sleep stages and slow wave up-phases, as well as the determination of acoustic stimulation schemes, will deviate.

[0029] Feature extraction is performed on the preprocessed EEG data, and sleep stage classification is carried out based on a deep neural network to identify slow wave up-phases. The empirical mode decomposition algorithm and wavelet transform time-frequency analysis are used to extract features, and then the powerful pattern recognition ability of the deep neural network is utilized for sleep stage classification, and further to identify slow wave up-phases. This helps to accurately locate the key sleep stages where the brain generates slow waves and spindles, provides an accurate basis for implementing acoustic stimulation at the most effective time, improves the pertinence of intervention, and avoids ineffective stimulation in non-critical stages.

[0030] Based on the empirical mode decomposition algorithm, the preprocessed EEG data is decomposed to obtain multiple intrinsic mode function components IMF i (t), where i is the number of the intrinsic mode function component, i = 1, 2, 3, …, m, and m is the total number of intrinsic mode function components; the decomposition process is achieved through a screening process, and each screening separates an IMF component from the original signal. EEG data is complex non-linear signals containing various components with different frequencies and amplitudes. The empirical mode decomposition algorithm can adaptively decompose it into multiple intrinsic mode function (IMF) components. Each IMF component represents the fluctuation characteristics of the signal at different time scales and reflects the inherent patterns of the EEG signal. This decomposition process can separate signals with different characteristics in the EEG, such as extracting signals related to slow waves and spindles from the complex EEG background, making subsequent analysis more targeted, improving the recognition of sleep-related EEG characteristics, and helping to more accurately analyze sleep states.

[0031] Wavelet transform time-frequency analysis is performed on the decomposed intrinsic mode function components. The Morlet wavelet is selected as the wavelet basis function to obtain the time-frequency coefficient matrix W corresponding to each intrinsic mode function component i(a, b). Preprocess these time-frequency coefficient matrices and convert them into a format suitable for the input of a deep neural network (DNN). Wavelet transform time-frequency analysis can analyze signals in both the time and frequency dimensions. Performing this analysis on the IMF components can obtain the frequency distribution of each component at different time points, resulting in a time-frequency coefficient matrix. This is crucial for studying sleep brainwaves because the frequencies and occurrence times of slow waves and spindle waves during sleep are dynamically changing. Through the time-frequency coefficient matrix, it is possible to clearly observe the frequency change patterns, durations, and their evolution over time of slow waves and spindle waves in different sleep stages, providing rich and accurate feature information for sleep stage classification and slow wave up-phase recognition based on a deep neural network, improving the accuracy of classification and recognition, and thus providing strong support for determining the optimal acoustic stimulation timing and parameters.

[0032] Construct a multi-layer perceptron deep learning neural network model, including an input layer, multiple hidden layers, and an output layer. The number of nodes in the hidden layer is N1, and the number of nodes in the output layer is N2. N2 corresponds to the number of different sleep stage categories, such as wakefulness, light sleep, deep sleep, rapid eye movement sleep (REM), etc.

[0033] Train based on the backpropagation algorithm, with the activation function being the ReLU function and the loss function being the cross-entropy loss function; during the training process, use a large amount of brainwave data with labeled sleep stages as the training set, and adjust the weights and biases of the network by minimizing the loss function. The backpropagation algorithm is an effective method for training neural networks. By calculating the error between the predicted result and the true label and propagating the error from the output layer back to the input layer, the weights and biases of each layer in the neural network are adjusted, enabling the predicted result of the model to continuously approach the true value. The activation function ReLU (Rectified Linear Unit) can solve the gradient vanishing problem, accelerate the convergence speed of the model, and enable the model to learn effective feature representations faster. The cross-entropy loss function has good performance for classification problems, can measure the difference between the model's predicted result and the true label, and by minimizing this loss function, the performance of the model in the sleep stage classification task can be optimized, improving the classification accuracy.

[0034] Input the time-frequency coefficient matrix W i (a, b) into the trained multi-layer perceptron deep learning neural network model to obtain the probabilities P of each sleep stage category, P = [p1, p2, p3,..., p N2 ,], where p1, p2, p3, p N2 are the probabilities of the 1st sleep stage category, the 2nd sleep stage category, the 3rd sleep stage category, and the N2th sleep stage category respectively, and determine the maximum probability value p max , this maximum probability value pmax The corresponding sleep stage category is the predicted sleep stage category. By selecting the category with the highest probability as the predicted sleep stage, the learning results of the model can be fully utilized to achieve accurate judgment of the sleep stage. This probability-based classification method can not only clarify the current sleep stage but also reflect the confidence of the model in the classification result, providing a reliable basis for subsequent judgment of the slow wave rising phase. If the sleep stage is inaccurately judged, subsequent identification of the slow wave rising phase and formulation of the acoustic stimulation plan may deviate. Accurate sleep stage classification can effectively avoid these problems and improve the accuracy and effectiveness of the entire sleep intervention system.

[0035] When the sleep stage category is judged to be the deep sleep stage and the corresponding probability p of the deep sleep stage ds is greater than the set probability threshold, the slow wave rising phase is identified; the preprocessed electroencephalogram data is obtained, and the slow wave-related signals are extracted; the slow wave signal amplitude A(t) is calculated based on the slow wave-related signals; within the set time window, the slow wave amplitude change rate r(t) is calculated based on the slow wave signal amplitude within the window, where Δt is the time interval; when the slow wave amplitude change rate r(t) is greater than 0 and the duration exceeds the set duration threshold, it is judged to be in the slow wave rising phase. The sleep process includes multiple stages, and slow waves mainly appear in the deep sleep stage. By setting this prerequisite, the range for searching the slow wave rising phase can be narrowed, focusing on the sleep period most likely to have the slow wave rising phase. Only when in the deep sleep stage and the judgment probability of this stage is greater than the set threshold, subsequent identification is carried out, which increases the accuracy and reliability of the identification, avoids ineffective searches in other sleep stages, reduces waste of computing resources, and improves the identification efficiency.

[0036] Combining the two conditions that the slow wave amplitude change rate is greater than 0 and the duration exceeds the set duration threshold can more accurately determine the slow wave rising phase. The slow wave amplitude change rate being greater than 0 indicates that the amplitude is in an upward state, and the duration exceeding the set duration threshold excludes the interference of short-term fluctuations, ensuring that the judged slow wave rising phase has a certain persistence and stability. This dual-condition judgment method greatly improves the accuracy of slow wave rising phase identification, provides a reliable time judgment basis for subsequent implementation of targeted acoustic stimulation during the slow wave rising phase, enables the acoustic stimulation to be carried out at the most effective time, and thus more effectively enhances sleep slow waves and spindles.

[0037] If in the slow wave rising phase, determine the acoustic wave stimulation scheme based on the current sleep characteristics and the user's personalized parameters to enhance sleep slow waves and spindle waves. Obtain the current sleep characteristics such as the real-time frequency of slow waves and the density of spindle waves, and construct a feature vector by combining the personalized parameters such as the user's age, gender, sleep habits, and past response data to acoustic wave stimulation. Use the particle swarm optimization algorithm to determine the optimal acoustic wave stimulation parameters, generate and evaluate different types of acoustic signals, and finally determine the optimal scheme. In this way, the most suitable acoustic wave stimulation scheme can be customized according to each user's unique sleep condition and individual characteristics, realizing personalized treatment, maximizing the enhancement of sleep slow waves and spindle waves, effectively improving sleep quality, and meeting the different individual's differentiated needs.

[0038] When it is judged to be in the slow wave rising phase, obtain the real-time frequency f of slow waves SW and the density D of spindle waves sp ; record the real-time frequency f of slow waves SW , the density D of spindle waves sp and the amplitude A(t) of slow wave signals as the current sleep characteristics; integrate the current sleep characteristics and the user's personalized parameters to construct a feature vector F, and the user's personalized parameters include the response data to different acoustic wave stimulations during sleep monitoring, the user's age, gender, and sleep habits SlHa; integrating these key sleep characteristics forms a comprehensive set of current sleep characteristics. This can comprehensively reflect the important information related to slow waves and spindle waves in the current sleep stage, providing a comprehensive data basis for formulating the acoustic wave stimulation scheme in combination with personalized parameters. These characteristics are interrelated and jointly affect sleep quality and the response to acoustic wave stimulation. Considering them comprehensively can more accurately grasp the sleep state and needs.

[0039] Based on the particle swarm optimization algorithm (PSO), perform particle optimization to determine the optimal acoustic wave stimulation parameters, and the optimal acoustic wave stimulation parameters include the frequency f of the acoustic wave signal st and the intensity I of the acoustic wave signal st ; set the relevant parameters of the PSO algorithm, the particle swarm size, learning factor, and inertia weight. Define the fitness function J to evaluate the pros and cons of different combinations of acoustic wave stimulation parameters. Aiming to increase the amplitude of slow waves and the density of spindle waves, the fitness function J can be expressed as J = β1 * ΔA SW + β2 * ΔD SW , where ΔA SW is the expected increment of slow wave amplitude, ΔD SW is the expected increment of spindle wave density, β1 is the weight factor of ΔA SW , and β2 is the weight factor of ΔD SWThe weight factor. When determining the acoustic stimulation parameters (acoustic signal frequency and acoustic signal intensity), using this algorithm can quickly search for the optimal solution among numerous possible parameter combinations. With the goal of increasing slow-wave amplitude and spindle density, through continuous iterative optimization, the acoustic stimulation parameters most suitable for the current user and the current sleep state are found. Compared with randomly selecting or simply setting parameters based on experience, this optimization algorithm can significantly improve the effect of acoustic stimulation, ensuring that the applied acoustic stimulation can maximize the enhancement of sleep slow waves and spindle waves.

[0040] Generate acoustic signals corresponding to each acoustic signal type based on the optimal acoustic stimulation parameters. The acoustic signal types include pink noise or sine modulation signals, etc.; Evaluate the impact of the generated acoustic signals on sleep slow waves and spindle waves in a simulated environment; Determine the acoustic signals corresponding to the optimal acoustic signal type. The diverse acoustic signal types provide more choices for subsequent screening of the most effective signals. Different types of acoustic signals have different effects on sleep brainwaves. By generating multiple signals, comparative evaluations can be carried out in a simulated environment to find the acoustic signals most effective for a specific user and sleep state, further improving the accuracy and effectiveness of the intervention.

[0041] The sleep habit SlHa is obtained based on the sleep latency parameter, the body activity parameter during sleep, and the body temperature parameter during sleep. The process is as follows: Obtain the sleep latency parameter, where the sleep latency parameter includes the average sleep latency qSt and the standard deviation of sleep latency σ qSt ; Obtain the body activity parameter during sleep, where the body activity parameter during sleep includes the total number of body activities, the number of micro-movements, and the number of turns during each sleep; Based on the body activity parameter during sleep, obtain the number of body activities per hour SZt, the average number of micro-movements per hour SWt, and the average number of turns per hour SFt; Obtain the body temperature parameter during sleep, where the body temperature parameter during sleep includes the starting body temperature, the ending body temperature, as well as the highest body temperature, the lowest body temperature, and the body temperature change range during the sleep process; Perform mean processing on the body temperature parameter during sleep to obtain the average starting body temperature Swq, the average highest body temperature Swg, the average lowest body temperature Swd, and the average body temperature change range Swf; Based on the sleep habit processing model, obtain the sleep habit SlHa:

[0042] SlHa = f1 + f2 + f3;

[0043]

[0044] Among them, f1, f2, and f3 are all transfer functions, cSt is the specified average sleep latency, σ cStLet the standard deviation of the sleep latency be SDt, the number of body movements per hour be CZt, the number of micro-movements per hour on average be CWt, the number of turns per hour on average be CFt, the average starting body temperature be Cwq, the average highest body temperature be Cwg, the average lowest body temperature be Cwd, and the average body temperature change range be Cwf.

[0045] Using the sleep habit processing model, combining various parameters obtained above (average sleep latency, standard deviation of sleep latency, number of body movements per hour, number of micro-movements per hour on average, number of turns per hour on average, average starting body temperature, average highest body temperature, average lowest body temperature, and average body temperature change range) and the determined standard parameters (determined average sleep latency, determined standard deviation of sleep latency, etc.), the sleep habit is calculated through a transfer function. This method can comprehensively integrate sleep-related information from multiple dimensions and comprehensively and systematically describe the user's sleep habit. By comparing the individual sleep parameters with the determined standards, the characteristics of the individual sleep habit and the degree of deviation from the normal range can be clearly seen. And accurately understanding the sleep habit is crucial for determining the acoustic stimulation plan because different sleep habits respond differently to acoustic stimulation. Based on this, a personalized acoustic stimulation plan can better adapt to individual differences and more effectively enhance sleep slow waves and spindle waves, improving sleep quality.

[0046] Input the current sleep characteristics and the user's personalized parameters into the sleep simulation model, and initialize the running state of the sleep simulation model; in the simulation environment, based on the determined time combination, each type of acoustic signal generated based on the optimal acoustic stimulation parameters is respectively input into the sleep simulation model after the initialization of the running state to obtain simulated electroencephalogram data. The time combination includes the starting time of signal application and the signal duration; different starting times and durations will produce different effects. By systematically testing various combinations, the most suitable acoustic stimulation time arrangement can be found. Obtaining the simulated electroencephalogram data provides a direct data source for subsequent analysis. These data reflect the effects of acoustic signals on sleep slow waves and spindle waves under simulated conditions and are the key basis for evaluating the effects of acoustic signals.

[0047] Based on the simulated electroencephalogram data, obtain the average instantaneous amplitude mA of the simulated slow waves SW and the density mD of the simulated spindle waves SP ; based on the average instantaneous amplitude mA of the simulated slow waves SW and the average amplitude sA of the slow wave signal of the patient SW obtain the relative change rate RA of the slow wave amplitude SW , Based on the density mD of the simulated spindle waves SP and the density sD of the spindle waves of the patient SP obtain the relative change rate RD of the spindle wave density SP , An evaluation coefficient Px is obtained based on the relative change rate of slow wave amplitude and the relative change rate of spindle wave density; the minimum evaluation coefficient is determined, and the implementation scheme of the acoustic wave signal corresponding to the minimum evaluation coefficient is obtained. The implementation scheme of the acoustic wave signal includes the application start time and duration, and various types of acoustic wave signals generated based on the optimal acoustic wave stimulation parameters.

[0048] The method for obtaining the evaluation coefficient Px is as follows: Px = α1 * RA SW + α2 * RD SP ; where α1 is the weight factor of RA SW and α2 is the weight factor of RD SP Combining the relative change rate of slow wave amplitude and the relative change rate of spindle wave density to obtain the evaluation coefficient comprehensively considers the dual effects of acoustic wave signals on sleep slow waves and spindle waves. Since both slow waves and spindle waves play important roles during sleep, a comprehensive evaluation coefficient can more comprehensively measure the overall effect of acoustic wave signals. This avoids the drawback of only focusing on a single index and ignoring other factors, making the evaluation result more comprehensive and scientific, and providing a more comprehensive basis for selecting the optimal acoustic wave signal.

[0049] A sleep slow wave and spindle wave enhancement system based on acoustic wave stimulation is used for the above-mentioned method for enhancing sleep slow waves and spindle waves based on acoustic wave stimulation, as Figure 2 shown, including a data preprocessing module, a slow wave rising phase recognition module, and an acoustic wave stimulation scheme determination module, where: the data preprocessing module is used to collect electroencephalogram data and perform preprocessing; the slow wave rising phase recognition module is used to extract features from the preprocessed electroencephalogram data, classify the sleep stages based on a deep neural network, and identify the slow wave rising phase; the acoustic wave stimulation scheme determination module is used to, when in the slow wave rising phase, determine the acoustic wave stimulation scheme based on the current sleep characteristics and the user's personalized parameters to enhance sleep slow waves and spindle waves.

[0050] An electronic device includes: a processor; and a memory in which computer program instructions are stored, and when the computer program instructions are run by the processor, the processor executes the above-mentioned method for enhancing sleep slow waves and spindle waves based on acoustic wave stimulation.

[0051] A computer-readable storage medium is used to store a program, and when the program is executed by a processor, it implements the above-mentioned method for enhancing sleep slow waves and spindle waves based on acoustic wave stimulation.

[0052] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0053] The present invention is described with reference to the flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0054] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that realize the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0055] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, such that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable devices provide steps for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0056] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they know the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0057] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A method for enhancing sleep slow waves and spindles based on acoustic stimulation, characterized in that, It includes the following steps: Collect electroencephalogram (EEG) data and perform preprocessing; Extract features from the preprocessed EEG data, classify sleep stages based on a deep neural network, and identify the slow wave rising phase; If in the slow wave rising phase, determine an acoustic stimulation scheme based on the current sleep characteristics and user personalization parameters to enhance sleep slow waves and spindles.

2. The method for enhancing sleep slow waves and spindle waves based on acoustic stimulation according to claim 1, wherein EEG data collection and preprocessing include the following steps: Collect the initial EEG signal based on electrode patches, and the initial EEG signal is amplified by an amplifier and then outputs the amplified EEG signal V1; Perform filtering on the amplified EEG signal V1: Perform low-pass filtering on the slow wave to filter out interference signals higher than the set cut-off frequency f c1 to obtain slow wave-related signals; Perform band-pass filtering on spindle waves, retaining spindle wave-related signals within the set frequency range [f c2 , f c3 , where f c2 and f c3 are respectively the minimum and maximum frequencies of the set frequency range; Superimpose the slow wave related signal and the spindle wave related signal to obtain the filtered EEG signal; The analog-to-digital converter samples the filtered EEG signal based on a set sampling rate, converts the analog filtered EEG signal into a digital filtered EEG signal, and obtains a digital EEG data sequence, which is the preprocessed EEG data.

3. A method for enhancing sleep slow waves and spindles based on acoustic stimulation according to claim 1, characterized in that Feature extraction from the preprocessed EEG data includes the following steps: Decompose the preprocessed electroencephalogram data based on the empirical mode decomposition algorithm to obtain multiple intrinsic mode function components IMF i (t), where i is the number of the intrinsic mode function component, i = 1, 2, 3, …, m, and m is the total number of the intrinsic mode function components; Perform wavelet transform time-frequency analysis on the decomposed intrinsic mode function components to obtain the time-frequency coefficient matrix \(W\) corresponding to each intrinsic mode function component i (a,b).

4. A method for enhancing sleep slow waves and spindles based on acoustic stimulation according to claim 3, characterized in that Sleep stage classification based on a deep neural network includes the following steps: Construct a multi-layer perceptron deep learning neural network model, including an input layer, multiple hidden layers, and an output layer, where the number of nodes in the hidden layer is N1, and the number of nodes in the output layer is N2, and N2 corresponds to the number of different sleep stage categories; Train based on the backpropagation algorithm, with the activation function being the ReLU function and the loss function being the cross-entropy loss function; Input the time-frequency coefficient matrix W i (a, b) into the trained multi-layer perceptron deep learning neural network model to obtain the probabilities P of each sleep stage category, where P = [p1, p2, p3,..., p N2 ,], and p1, p2, p3, p N2 are the probabilities of the first sleep stage category, the second sleep stage category, the third sleep stage category, and the N2th sleep stage category respectively. Determine the maximum probability value p max , and the sleep stage category corresponding to this maximum probability value p max is the predicted sleep stage category.

5. A method for enhancing sleep slow waves and spindles based on acoustic stimulation according to claim 4, characterized in that Identification of the slow wave rising phase includes the following steps: When the sleep stage category is judged to be the deep sleep stage and the corresponding probability p ds of the deep sleep stage is greater than the set probability threshold, slow wave rising phase recognition is performed; Obtain the preprocessed EEG data and extract the slow wave related signal; Calculate the slow wave signal amplitude A(t) based on the slow wave related signal; Within a set time window, calculate the slow-wave amplitude change rate r(t) based on the slow-wave signal amplitude within the window, Δt is the time interval; When the slow wave amplitude change rate r(t) is greater than 0 and the duration exceeds the set duration threshold, it is judged to be in the slow wave rising phase.

6. A method for enhancing sleep slow waves and spindles based on acoustic stimulation according to claim 1, characterized in that Determining the acoustic stimulation scheme based on the current sleep characteristics and user personalization parameters includes the following steps: When it is determined that it is in the slow-wave rising phase, obtain the slow-wave real-time frequency f SW and the spindle density D sp ; Record the real-time frequency f of slow waves SW , the spindle density D sp and the amplitude A(t) of the slow wave signal as the current sleep characteristics; Integrate the current sleep characteristics and user personalization parameters to construct a feature vector F, and the user personalization parameters include response data to different acoustic stimulations during sleep monitoring, the user's age, gender, and sleep habit SlHa; Particle optimization is carried out based on the particle swarm optimization algorithm to determine the optimal acoustic wave stimulation parameters, where the optimal acoustic wave stimulation parameters include the acoustic wave signal frequency f st and the acoustic wave signal intensity I st ; Generate acoustic signals corresponding to each acoustic signal type based on the optimal acoustic stimulation parameters; Evaluate the impact of the generated acoustic signals on sleep slow waves and spindles in a simulated environment; Determine the acoustic signal corresponding to the optimal acoustic signal type.

7. A method for enhancing sleep slow waves and spindles based on acoustic stimulation according to claim 6, characterized in that, The sleep habit SlHa is obtained based on the sleep latency parameter, the body activity parameter during sleep, and the body temperature parameter during sleep, and the process is as follows: Obtain the sleep latency parameters, where the sleep latency parameters include the average sleep latency qSt and the standard deviation of sleep latency σ qSt ; Obtain the body activity parameter during sleep, and the body activity parameter during sleep includes the total number of body activities, the number of micro-movements, and the number of turns during each sleep; Based on the body activity parameter during sleep, obtain the number of body activities per hour SZt, the average number of micro-movements per hour SWt, and the average number of turns per hour SFt; Obtain the body temperature parameter during sleep, and the body temperature parameter during sleep includes the starting body temperature, the ending body temperature of each sleep, as well as the highest body temperature, the lowest body temperature, and the body temperature change range during sleep; Perform mean processing on the body temperature parameters during sleep to obtain the average starting body temperature Swq, the average highest body temperature Swg, the average lowest body temperature Swd, and the average body temperature change amplitude Swf; Based on the sleep habit processing model, obtain the sleep habit SlHa: SlHa = f1 + f2 + f3; Among them, f1, f2, and f3 are all transfer functions, cSt is the specified average sleep latency, and σ cSt is the standard deviation of the specified sleep latency, CZt is the specified number of body movements per hour, CWt is the specified average number of fine movements per hour, CFt is the specified average number of turns per hour, Cwq is the specified average starting body temperature, Cwg is the specified average maximum body temperature, Cwd is the specified average minimum body temperature, and Cwf is the specified average body temperature change range.

8. A method for enhancing sleep slow waves and spindles based on acoustic stimulation according to claim 7, characterized in that, In a simulation environment, evaluate the impact of the generated acoustic signal on sleep slow waves and spindle waves, including the following steps: Input the current sleep characteristics and the user's personalized parameters into the sleep simulation model to initialize the running state of the sleep simulation model; In the simulation environment, based on the determined time combination, input each type of acoustic signal generated based on the optimal acoustic stimulation parameters into the sleep simulation model after initializing the running state to obtain simulated electroencephalogram data. The time combination includes the signal application start time and the signal duration; Obtaining the mean value mA of the simulated slow wave instantaneous amplitude based on simulated brain wave data SW and the simulated spindle wave density mD SP ; Based on the mean value mA of the instantaneous amplitude of the simulated slow waves SW and the mean value sA of the amplitudes of the slow wave signals of the patient SW the relative change rate RA of the slow wave amplitude is obtained SW , Based on the simulated spindle density mD SP and the patient's spindle density sD SP the relative change rate RD of the spindle density is obtained SP , Obtain the evaluation coefficient Px based on the relative change rate of slow wave amplitude and the relative change rate of spindle wave density; Determine the minimum evaluation coefficient, obtain the acoustic signal implementation plan corresponding to the minimum evaluation coefficient. The acoustic signal implementation plan includes the application start time and the duration, and each type of acoustic signal generated based on the optimal acoustic stimulation parameters.

9. A method for enhancing sleep slow waves and spindles based on acoustic stimulation according to claim 8, characterized in that, The method for obtaining the evaluation coefficient Px is as follows: Px = α1 * RA SW + α2 * RD SP ; Among them, α1 is the weighting factor of RA SW and α2 is the weighting factor of RD SP .

10. A sleep slow wave and spindle wave enhancement system based on acoustic wave stimulation, for use in the method for enhancing sleep slow waves and spindle waves based on acoustic wave stimulation according to any one of claims 1-9, characterized in that, It includes a data preprocessing module, a slow wave rising phase recognition module, and an acoustic stimulation scheme determination module, where: The data preprocessing module is used to collect electroencephalogram data and perform preprocessing; The slow wave rising phase recognition module is used to extract features from the preprocessed electroencephalogram data, classify the sleep stages based on a deep neural network, and identify the slow wave rising phase; The acoustic stimulation scheme determination module is used to, when in the slow wave rising phase, determine the acoustic stimulation scheme based on the current sleep characteristics and the user's personalized parameters to enhance sleep slow waves and spindle waves.

Images