Sleep quality analysis system based on deep learning

Abstract

The invention provides a sleep quality analysis system based on deep learning, and relates to the technical field of biomedical engineering and artificial intelligence, and the sleep quality analysis system comprises a multi-modal signal collection module which is used for collecting a physiological signal and an environment signal of a target user; the physiological signals comprise brain wave signals, electromyogram signals and skin resistance signals; the environment signal comprises a temperature and humidity signal and an air quality signal; the de-noising module is used for de-noising the physiological signal and the environment signal to obtain a de-noised signal; the feature extraction module is used for performing feature extraction on the de-noised signal to obtain a signal feature; the feature processing module is used for performing data cleaning, data filling and normalization processing on the signal features to obtain normalized features; the evaluation module is used for inputting the normalized features into a trained sleep quality evaluation model to obtain an evaluation result; the sleep quality evaluation model is constructed through a deep learning network. According to the invention, the sleep quality can be evaluated more accurately.

Description

Technical Field

[0001] The present invention relates to the fields of biomedical engineering and artificial intelligence technology, and in particular to a sleep quality analysis system based on deep learning. Background Art

[0002] As the pace of life accelerates, sleep quality issues are receiving increasing attention. Traditional sleep monitoring methods usually rely on polysomnography (PSG), but its equipment is complex and inconvenient for daily use. In recent years, sleep analysis methods based on data collection from wearable devices combined with deep learning technology have gradually become a research hotspot.

[0003] The invention patent with application number 202311397993.6 discloses a user sleep quality management system based on detection data analysis, and the system includes: a sleep state monitoring module, including an environment monitoring unit, a physiological monitoring unit and a sleeping position monitoring unit, which are used to monitor the user's sleeping environment state, physiological parameters and sleeping position state respectively to obtain monitoring data; an analysis unit, which is used to evaluate the user's sleep state according to the monitoring data and the user's historical sleep data to obtain an evaluation result; a tracing module, which is used to trace the factors affecting the user's sleep according to the evaluation result and issue corresponding prompts and suggestions; the sleeping position monitoring unit includes an induction pillow and an induction mattress, the induction pillow includes a pillow body and several groups of pressure sensors; the induction mattress includes a mattress body and several groups of pressure sensors; the process of evaluating the user's sleep state includes: S1, monitoring the user's sleeping position change state according to the parameters of the pressure sensor in the induction pillow induction mattress, and obtaining sleeping position state monitoring data; S2, obtaining the values ​​of various physiological parameters of the user's sleep process through the physiological monitoring unit; S3, evaluating the user's sleep state according to the sleeping position state monitoring data and the values ​​of various physiological parameters. However, this technology only relies on pressure sensors, environment and basic physiological parameters. However, sleep quality is affected by more factors, such as brain wave activity, skin electrical response, etc., which are not covered, resulting in poor sleep quality monitoring effect. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, an object of the present invention is to provide a sleep quality analysis system based on deep learning, which can more accurately evaluate sleep quality.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A sleep quality analysis system based on deep learning, comprising:

[0007] A multimodal signal acquisition module, used to collect physiological signals and environmental signals of the target user; the physiological signals include brain wave signals, electromyography signals and skin resistance signals; the environmental signals include temperature and humidity signals and air quality signals;

[0008] A denoising module, used for performing denoising processing on the physiological signal and the environmental signal to obtain a denoised signal;

[0009] A feature extraction module, used to extract features from the denoised signal to obtain signal features;

[0010] A feature processing module, used for performing data cleaning, data filling and normalization processing on the signal features to obtain normalized features;

[0011] An evaluation module is used to input the normalized features into a trained sleep quality evaluation model to obtain an evaluation result; the sleep quality evaluation model is constructed through a deep learning network.

[0012] Preferably, the denoising module comprises:

[0013] A signal decomposition submodule, used to decompose the physiological signal and the environmental signal into sub-signals of different frequencies using basis functions respectively;

[0014] The coefficient acquisition submodule is used to obtain the transform coefficients corresponding to the sub-signals of different frequencies;

[0015] The variance calculation submodule is used to calculate the variance of the transformation coefficients at each decomposition scale;

[0016] A model building submodule, used to build a signal denoising model based on the variance of the transform coefficients;

[0017] The preprocessing submodule is used to preprocess the physiological signal and the environmental signal using the signal denoising model to obtain the denoised signal.

[0018] Preferably, the model building submodule includes:

[0019] A standard deviation estimation unit, used for estimating the standard deviation of abnormal noise according to the variance of the transform coefficients;

[0020] A threshold construction unit is used to construct a coefficient removal threshold according to the standard deviation of the noise; wherein the coefficient removal threshold is: Among them, t represents the coefficient removal threshold, σ 0 represents the variance of the transform coefficients, σ p represents the standard deviation of abnormal noise, median(d p ) represents the median value of the transform coefficients at the pth decomposition scale;

[0021] The denoising model building unit is used to build a signal denoising model by using the coefficient removal threshold.

[0022] Preferably, the signal denoising model is:

[0023]

[0024] Among them, α represents the transformation coefficient, w j,k represents the transformation coefficient at the jth decomposition scale, m represents the convergence coefficient, sgn represents the sign function, and w j,k Represents the transformation coefficients for removing outliers.

[0025] Preferably, the preprocessing submodule includes:

[0026] A coefficient setting unit, used to set initial transformation coefficients and convergence coefficients for the signal denoising model;

[0027] A processing unit, used to process the transformation coefficients using the set signal denoising model to obtain the transformation coefficients with outliers removed;

[0028] A reconstruction unit, used for reconstructing the transformation coefficients after removing the abnormal values ​​to obtain a reconstructed signal;

[0029] A signal-to-noise ratio calculation unit, used to calculate the signal-to-noise ratio between the reconstructed signal and the original physiological signal and the environmental signal;

[0030] The iteration unit is used to reset the transformation coefficient and the convergence coefficient when the signal-to-noise ratio is not within the preset range until the signal-to-noise ratio remains within the preset range.

[0031] Preferably, the signal characteristics include physiological signal sub-features and environmental signal sub-features; the physiological signal sub-features include: frequency band energy, power spectral density, entropy and complexity of brain wave signals, average amplitude, frequency and zero crossing rate of electromyography signals, conductivity change and skin electrical response frequency of skin resistance signals; the environmental signal sub-features include: average temperature and humidity and change rate of temperature and humidity signals, carbon dioxide concentration and volatile organic compound level of air quality signals.

[0032] Preferably, the feature processing module includes:

[0033] The first traversal submodule is used to traverse all signal features, detect duplicate and invalid data, and delete the detected duplicate and invalid data;

[0034] The second traversal submodule is used to traverse all signal features and detect missing features in the data, wherein features with missing values ​​exceeding 80% are deleted, and features with missing values ​​less than 80% are filled based on the KNN algorithm;

[0035] The imbalance submodule is used to process the data imbalance of the feature data after data deletion and data filling based on the SMOTE algorithm;

[0036] The normalization submodule is used to normalize the data after data imbalance processing based on the Z-score normalization method to obtain the normalized features.

[0037] Preferably, the sleep quality assessment model includes a multimodal input layer, a single signal convolution layer, a shared feature fusion layer, a timing analysis layer, a feature attention mechanism layer and an output evaluation layer; the multimodal input layer is provided with an input channel corresponding to each signal type; the multimodal input layer is connected to the single signal convolution layer; the shared feature fusion layer is connected to the single signal convolution layer; the shared feature fusion layer is used to aggregate features from different single signal convolution layers, and integrate the features through a fusion algorithm to obtain fused features; the timing analysis layer is connected to the shared feature fusion layer, and the timing analysis layer is used to perform time series analysis on the fused features using a long short-term memory network; the feature attention mechanism layer is connected to the shared feature fusion layer, and the feature attention mechanism layer is deployed with an attention mechanism to dynamically adjust the feature weights of different signals; the output evaluation layer is connected to the timing analysis layer, and the output evaluation layer is used to perform a comprehensive calculation of the evaluation results based on the results of the time series analysis; the evaluation results include a sleep quality score.

[0038] Preferably, the calculation formula of the sleep quality score is:

[0039]

[0040] Among them, SQ is the sleep quality score, N is the number of signal features, and a i is the attention weight of the i-th signal feature, f i is the value of the i-th signal feature, LSTM output is the output result of the LSTM layer, M is the number of environmental signal features, and w j is the contribution coefficient of the jth environmental feature, p j is the value of the jth environmental signal feature, such as average temperature, humidity change, and air quality index. α controls the impact of the summary of single signal features on the total score. β controls the impact of the dynamic characteristics of the time series. γ adjusts the proportion of environmental features in the total score. δ is a bias term used to adjust the benchmark of the final score.

[0041] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0042] The present invention provides a sleep quality analysis system based on deep learning, including: a multimodal signal acquisition module for collecting physiological signals and environmental signals of a target user; the physiological signals include brain wave signals, electromyogram signals and skin resistance signals; the environmental signals include temperature and humidity signals and air quality signals; a denoising module for denoising the physiological signals and the environmental signals to obtain denoised signals; a feature extraction module for extracting features from the denoised signals to obtain signal features; a feature processing module for performing data cleaning, data filling and normalization on the signal features to obtain normalized features; an evaluation module for inputting the normalized features into a trained sleep quality evaluation model to obtain evaluation results; the sleep quality evaluation model is constructed through a deep learning network. The present invention simultaneously collects physiological signals (such as brain waves, electromyograms, and skin resistance) and environmental signals (such as temperature and humidity, air quality), and multiple data sources provide a more comprehensive understanding of sleep quality. And different signals can complement each other, physiological signals reflect physical conditions, and environmental signals can optimize conditions, thereby more accurately evaluating sleep quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0044] Figure 1 A schematic diagram of the system structure provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] The purpose of the present invention is to provide a sleep quality analysis system based on deep learning, which can accurately evaluate sleep quality.

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] Figure 1 A schematic diagram of the system structure provided by an embodiment of the present invention, such as Figure 1As shown, the present invention provides a sleep quality analysis system based on deep learning, comprising:

[0049] A multimodal signal acquisition module, used to collect physiological signals and environmental signals of the target user; the physiological signals include brain wave signals, electromyography signals and skin resistance signals; the environmental signals include temperature and humidity signals and air quality signals;

[0050] A denoising module, used for performing denoising processing on the physiological signal and the environmental signal to obtain a denoised signal;

[0051] A feature extraction module, used to extract features from the denoised signal to obtain signal features;

[0052] A feature processing module, used for performing data cleaning, data filling and normalization processing on the signal features to obtain normalized features;

[0053] An evaluation module is used to input the normalized features into a trained sleep quality evaluation model to obtain an evaluation result; the sleep quality evaluation model is constructed through a deep learning network.

[0054] Specifically, the physiological signal acquisition device of this embodiment includes:

[0055] Electroencephalogram (EEG) sensor, select a high-precision EEG headset, which usually includes multiple electrodes to collect electrical signals of brain activity.

[0056] Electromyography (EMG) sensors use patch-type or wearable EMG sensors attached to the surface of muscles to detect electrical signals of muscle activity.

[0057] Skin resistance signal (GSR) sensor, a GSR sensor worn on the finger or wrist, is used to measure changes in skin conductivity, reflecting the activity of the autonomic nervous system.

[0058] Furthermore, the environmental signal acquisition device of this embodiment includes:

[0059] Temperature and humidity sensor, using integrated temperature and humidity sensor module, can monitor ambient temperature and humidity in real time.

[0060] Air quality sensors, choose sensor modules that can detect carbon dioxide concentration and volatile organic compounds (VOCs) to assess air cleanliness.

[0061] Furthermore, the present embodiment correctly installs the physiological signal sensor at a designated location on the user's body to ensure the accuracy of signal acquisition. The environmental sensor is placed in the user's sleeping environment. Physiological and environmental signals are acquired in real time through the built-in acquisition module of the sensor. Usually these devices collect data at a certain sampling rate (such as hundreds of times per second). Use Bluetooth, Wi-Fi or other wireless communication technologies to transmit the collected signal data to a central processing unit (such as a computer or mobile device). Exemplarily, the present embodiment ensures that the timestamps of all signal data are synchronized for subsequent multimodal data fusion and analysis.

[0062] Furthermore, the present embodiment stores the data transmitted to the central processing unit and saves it in a time series format. After the data is stored, a preliminary denoising process is first performed, and a filter (such as a low-pass filter, a high-pass filter) is used to remove noise and interference in the signal. Then, key features are extracted from the denoised signal, such as the frequency band energy of EEG, the average amplitude of EMG, the conductivity change of GSR, etc. Missing data and outliers are then processed to ensure the integrity and consistency of the data, and the feature data is normalized for subsequent model input and analysis. Finally, the present embodiment inputs the preprocessed data into the trained deep learning model to evaluate sleep quality. Based on the analysis results, real-time feedback and suggestions are provided to help users improve their sleeping environment and habits.

[0063] Through the above steps, the system can effectively collect and process the user's physiological and environmental signals, providing a reliable data basis for sleep quality assessment.

[0064] Preferably, the denoising module comprises:

[0065] A signal decomposition submodule, used to decompose the physiological signal and the environmental signal into sub-signals of different frequencies using basis functions respectively;

[0066] The coefficient acquisition submodule is used to obtain the transform coefficients corresponding to the sub-signals of different frequencies;

[0067] The variance calculation submodule is used to calculate the variance of the transformation coefficients at each decomposition scale;

[0068] A model building submodule, used to build a signal denoising model based on the variance of the transform coefficients;

[0069] The preprocessing submodule is used to preprocess the physiological signal and the environmental signal using the signal denoising model to obtain the denoised signal.

[0070] Preferably, the model building submodule includes:

[0071] A standard deviation estimation unit, used for estimating the standard deviation of abnormal noise according to the variance of the transform coefficients;

[0072] A threshold construction unit is used to construct a coefficient removal threshold according to the standard deviation of the noise; wherein the coefficient removal threshold is: Among them, t represents the coefficient removal threshold, σ 0 represents the variance of the transform coefficients, σ p represents the standard deviation of abnormal noise, median(d p ) represents the median value of the transform coefficients at the pth decomposition scale;

[0073] The denoising model building unit is used to build a signal denoising model by using the coefficient removal threshold.

[0074] Preferably, the signal denoising model is:

[0075]

[0076] Among them, α represents the transformation coefficient, w j,k represents the transformation coefficient at the jth decomposition scale, m represents the convergence coefficient, sgn represents the sign function, and wj ,k Represents the transformation coefficients for removing outliers.

[0077] The present invention transforms the original signal into sub-signals of different frequencies according to the principle of wavelet transform. The above signal denoising model is continuous at t, the function value is (1-α)t, and the first order is differentiable, and the derivative value at t is 1-α. The convergence speed can be adjusted by m, and the larger the m, the faster the convergence. The shrinkage degree of the wavelet coefficient within ±t can be adjusted by α, so that the above function can be transformed between the conventional hard threshold function and the soft threshold function, which can better adapt to the characteristics of the sequence signal, thereby perfectly removing the interference of noise.

[0078] Preferably, the preprocessing submodule includes:

[0079] A coefficient setting unit, used to set initial transformation coefficients and convergence coefficients for the signal denoising model;

[0080] A processing unit, used to process the transformation coefficients using the set signal denoising model to obtain the transformation coefficients with outliers removed;

[0081] A reconstruction unit, used for reconstructing the transformation coefficients after removing the abnormal values ​​to obtain a reconstructed signal;

[0082] A signal-to-noise ratio calculation unit, used to calculate the signal-to-noise ratio between the reconstructed signal and the original physiological signal and the environmental signal;

[0083] The iteration unit is used to reset the transformation coefficient and the convergence coefficient when the signal-to-noise ratio is not within the preset range until the signal-to-noise ratio remains within the preset range.

[0084] Preferably, the signal characteristics include physiological signal sub-features and environmental signal sub-features; the physiological signal sub-features include: frequency band energy, power spectral density, entropy and complexity of brain wave signals, average amplitude, frequency and zero crossing rate of electromyography signals, conductivity change and skin electrical response frequency of skin resistance signals; the environmental signal sub-features include: average temperature and humidity and change rate of temperature and humidity signals, carbon dioxide concentration and volatile organic compound level of air quality signals.

[0085] Specifically, the physiological signal sub-features of this embodiment are as follows:

[0086] Brain wave signal (EEG), frequency band energy, extracts the energy of α, β, θ, and δ bands to identify sleep stages (such as REM, deep sleep).

[0087] Power spectral density (PSD), which analyzes the power distribution of different frequency components to help identify sleep depth and quality.

[0088] Entropy and complexity, such as sample entropy and approximate entropy, are used to assess the complexity and variability of brain waves.

[0089] Electromyography (EMG):

[0090] Average amplitude and frequency, which are used to detect muscle activity levels and help identify movement events during sleep (such as leg restlessness).

[0091] Zero-crossing rate, used to analyze the frequency characteristics of muscle activity.

[0092] Skin resistance signal (GSR):

[0093] Changes in conductivity reflect autonomic nervous system activity and help identify stress responses and arousal states.

[0094] Galvanic skin response frequency, used to assess mood and stress levels.

[0095] Furthermore, the environmental signal sub-features of this embodiment are as follows:

[0096] Temperature and humidity signal:

[0097] Average temperature and humidity, used to assess the comfort of the sleeping environment.

[0098] Change rate, analyze fluctuations in temperature and humidity, and identify environmental changes that may affect sleep.

[0099] Air quality signal:

[0100] Carbon dioxide concentration, high concentration may lead to poor sleep quality.

[0101] Volatile organic compound (VOC) levels, used to assess air cleanliness.

[0102] Furthermore, this embodiment analyzes EEG band energy and PSD to identify different sleep stages (such as light sleep, deep sleep, REM), thereby evaluating sleep structure. EMG features are used to identify movement events and awakening times during sleep and evaluate sleep continuity. GSR features are used to analyze autonomic nervous system activity and evaluate stress responses during sleep. Combined with temperature, humidity and air quality features, the impact of the comfort of the sleeping environment on sleep quality is evaluated.

[0103] The above-mentioned features of this embodiment are comprehensively analyzed through a deep learning model, which can provide a comprehensive assessment of sleep quality, help identify key factors affecting sleep, and provide personalized improvement suggestions.

[0104] Preferably, the feature processing module includes:

[0105] The first traversal submodule is used to traverse all signal features, detect duplicate and invalid data, and delete the detected duplicate and invalid data;

[0106] The second traversal submodule is used to traverse all signal features and detect missing features in the data, wherein features with missing values ​​exceeding 80% are deleted, and features with missing values ​​less than 80% are filled based on the KNN algorithm;

[0107] The imbalance submodule is used to process the data imbalance of the feature data after data deletion and data filling based on the SMOTE algorithm;

[0108] The normalization submodule is used to normalize the data after data imbalance processing based on the Z-score normalization method to obtain the normalized features.

[0109] Preferably, the sleep quality assessment model includes a multimodal input layer, a single signal convolution layer, a shared feature fusion layer, a timing analysis layer, a feature attention mechanism layer and an output evaluation layer; the multimodal input layer is provided with an input channel corresponding to each signal type; the multimodal input layer is connected to the single signal convolution layer; the shared feature fusion layer is connected to the single signal convolution layer; the shared feature fusion layer is used to aggregate features from different single signal convolution layers, and integrate the features through a fusion algorithm to obtain fused features; the timing analysis layer is connected to the shared feature fusion layer, and the timing analysis layer is used to perform time series analysis on the fused features using a long short-term memory network; the feature attention mechanism layer is connected to the shared feature fusion layer, and the feature attention mechanism layer is deployed with an attention mechanism to dynamically adjust the feature weights of different signals; the output evaluation layer is connected to the timing analysis layer, and the output evaluation layer is used to perform a comprehensive calculation of the evaluation results based on the results of the time series analysis; the evaluation results include a sleep quality score.

[0110] Specifically, the model structure of this embodiment includes:

[0111] The multimodal input layer designs a separate input channel for each signal type (EEG, EMG, GSR, temperature and humidity, air quality) so that the characteristics of each signal can enter the model independently.

[0112] Single signal convolution layer, uses convolutional neural network (CNN) for each input channel to extract local features, which can capture the data patterns of the signal, such as time characteristics, frequency domain changes, etc.

[0113] The shared feature fusion layer aggregates features from different signal convolution layers and integrates the features through a fully connected network (FCN) or a fusion algorithm (such as the attention mechanism) to capture the relationship between signals.

[0114] The time series analysis layer uses a long short-term memory network (LSTM) or a gated recurrent unit (GRU) to perform time series analysis on the fused features, which can capture the temporal dynamic changes and trends of the signal.

[0115] Feature attention mechanism layer, deploys attention mechanism to dynamically adjust the feature weights of different signals, highlighting and identifying the signal features that have the greatest impact on sleep quality.

[0116] The output evaluation layer uses a multi-layer perceptron (MLP) to comprehensively calculate the evaluation results and output a sleep quality score or classification (such as good, average, or poor).

[0117] Further, each signal feature of this embodiment enters through its corresponding input layer respectively. The design of different input channels allows the model to optimize processing for different data structures (such as one-dimensional time series and two-dimensional spectrograms). Each single signal convolution layer extracts local features of each signal. This causes each convolution layer to output a feature atlas, reflecting the local characteristics of each signal. The features output by multiple convolution layers are integrated in the shared feature fusion layer. This fusion helps to reveal the potential combination effect between signals, such as EEG and EMG combined to reveal the relationship between sleep and movement. The fused feature sequence is input to the LSTM layer, which is suitable for mining information that changes over time, and is important for understanding how movements, wakefulness cycles, and environmental changes affect sleep. By calculating the attention weights (dynamic weight allocation) of different features, it is identified that certain signals in specific situations may be more important to the final sleep quality assessment, thereby improving the accuracy of the assessment results. The final comprehensive feature obtains a sleep quality score through MLP, providing users with a quantitative assessment of their overall sleep status.

[0118] Furthermore, in the process of model training, in order to realize personalized improvement solutions, it is necessary to first collect a large amount of historical sleep data of individual users, including physiological signals, environmental signals, and users' daily routines and living habits. These multimodal data are processed and analyzed by deep learning models to capture the sleep performance characteristics of individual users under different environmental conditions and different work and rest patterns. The model uses task-specific loss functions to optimize the accuracy of personalized suggestions. At the same time, by realizing the continuous learning ability of the model, the user's personal profile can be updated after the new data provided by the user is input, and sleep problems under specific conditions can be identified. Then, combined with classification analysis and regression analysis, targeted suggestions are generated to truly match the user's personal needs. For example, when the user is often affected by environmental noise or changes in air quality, the system may recommend installing air purification equipment or using a white noise machine. At the same time, the system can also recommend adjustments to work and rest based on long-term sleep habit data, such as gradually changing bedtime to better adapt to the user's physiological rhythm. This makes the improvement plan not only scientific and individually adaptable, but also dynamically adjusted and continuously optimized.

[0119] Preferably, the calculation formula of the sleep quality score is:

[0120]

[0121] Among them, SQ is the sleep quality score, N is the number of signal features, and a i is the attention weight of the i-th signal feature, f i is the value of the i-th signal feature, LSTM output is the output result of the LSTM layer, M is the number of environmental signal features, and w jis the contribution coefficient of the jth environmental feature, p j is the value of the jth environmental signal feature, such as average temperature, humidity change, and air quality index. α controls the impact of the summary of single signal features on the total score. β controls the impact of the dynamic characteristics of the time series. γ adjusts the proportion of environmental features in the total score. δ is a bias term used to adjust the benchmark of the final score.

[0122] Specifically, the parameters of the above formula in this embodiment are explained as follows:

[0123] SQ (Sleep Quality): Sleep quality score, the final output result, usually a score from 0 to 100, indicating the overall sleep quality, from poor to excellent.

[0124] N: The number of signal features, usually the total number of different features extracted from a multimodal signal.

[0125] a i : The attention weight of the i-th signal feature is dynamically adjusted by the feature attention mechanism layer according to the contribution of different features to sleep quality.

[0126] f i : The value of the i-th signal feature, which is extracted from the shared feature fusion layer and may include features such as EEG band energy, EMG amplitude, GSR conductivity change, etc.

[0127] LSTM output : The output of the LSTM layer captures the temporal dynamic change information of the fused features and reflects the impact of temporal features on sleep quality.

[0128] M: The number of environmental signal features, such as the number of related features of temperature, humidity and air quality.

[0129] w j : The contribution coefficient of the jth environmental feature, which evaluates the impact of this environmental feature on the overall sleep quality, usually based on prior knowledge or data-driven adjustments.

[0130] p j : The value of the jth environmental signal feature, such as average temperature, humidity change, and air quality index.

[0131] α, β, γ, and δ are all hyperparameters of the model, which are used to control the impact of each part on the final sleep quality score. The specific values ​​can be adjusted through model training and validation sets, specifically:

[0132] α: controls the impact of summarizing single signal features on the total score.

[0133] β: Controls the influence of the dynamic characteristics of the time series.

[0134] γ: adjusts the weight of environmental characteristics in the total score.

[0135] δ: Bias term used to adjust the benchmark of the final score.

[0136] Optionally, in addition to the sleep quality score, the evaluation results also include sleep stage analysis, number of awakenings, sleep efficiency, sleep latency, deep sleep ratio, environmental influencing factor analysis, and personalized improvement suggestions. Sleep stage analysis provides the user's time distribution in different sleep stages (such as light sleep, deep sleep, REM); the number of awakenings and sleep efficiency reflect the continuity and quality of sleep; sleep latency refers to the time required to fall asleep; deep sleep ratio is a key indicator for measuring restorative sleep. Analysis of environmental influencing factors reveals the impact of temperature, humidity and air quality on sleep. Personalized improvement suggestions are based on the evaluation results and provide users with specific measures to optimize their sleep environment and habits. Through this comprehensive information, users can fully understand their own sleep conditions and take effective measures to improve them.

[0137] The beneficial effects of the present invention are as follows:

[0138] (1) The present invention collects physiological signals (such as brain waves, electromyography, skin resistance) and environmental signals (such as temperature, humidity, and air quality) at the same time. Multiple data sources provide a more comprehensive understanding of sleep quality. Different signals can complement each other. Physiological signals reflect physical conditions, while environmental signals can optimize conditions, thereby more accurately evaluating sleep quality.

[0139] (1) The denoising module of the present invention effectively reduces the false signals introduced by the environment and equipment, ensuring the accuracy of the data. The feature extraction module can extract key features from complex signals, actively capture information in the time domain and frequency domain, and improve the analysis effect.

[0140] (3) The deep learning network of the present invention can adaptively learn feature relationships and capture potential patterns based on the complexity of multimodal input data, which is difficult to achieve in traditional models. Deep learning has powerful nonlinear mapping capabilities and can handle complex data associations, improving the flexibility and accuracy of the model.

[0141] (4) The present invention combines deep learning technology, and the system can dynamically process data and provide sleep quality scores and suggestions in real time, thereby guiding users to improve their sleep. During the model training process, personalized improvement plans can be generated based on the user's specific data, such as adjusting the sleep environment, optimizing work and rest, etc.

[0142] (6) The model of the present invention is trained with a large amount of historical data, which reduces the dependence of human experience on the results and improves the reliability and accuracy of the results. By continuously collecting user feedback and monitoring the results, the system can be continuously iterated and optimized.

[0143] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0144] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A sleep quality analysis system based on deep learning, characterized in that: include: A multimodal signal acquisition module, used to collect physiological signals and environmental signals of the target user; the physiological signals include brain wave signals, electromyography signals and skin resistance signals; the environmental signals include temperature and humidity signals and air quality signals; A denoising module, used for performing denoising processing on the physiological signal and the environmental signal to obtain a denoised signal; A feature extraction module, used to extract features from the denoised signal to obtain signal features; A feature processing module, used for performing data cleaning, data filling and normalization processing on the signal features to obtain normalized features; An evaluation module is used to input the normalized features into a trained sleep quality evaluation model to obtain an evaluation result; the sleep quality evaluation model is constructed through a deep learning network.

2. The sleep quality analysis system based on deep learning according to claim 1, characterized in that: The denoising module comprises: A signal decomposition submodule, used to decompose the physiological signal and the environmental signal into sub-signals of different frequencies using basis functions respectively; The coefficient acquisition submodule is used to obtain the transform coefficients corresponding to the sub-signals of different frequencies; The variance calculation submodule is used to calculate the variance of the transformation coefficients at each decomposition scale; A model building submodule, used to build a signal denoising model based on the variance of the transform coefficients; The preprocessing submodule is used to preprocess the physiological signal and the environmental signal using the signal denoising model to obtain the denoised signal.

3. The sleep quality analysis system based on deep learning according to claim 2, characterized in that: The model building submodule includes: A standard deviation estimation unit, used for estimating the standard deviation of abnormal noise according to the variance of the transform coefficients; A threshold construction unit is used to construct a coefficient removal threshold according to the standard deviation of the noise; wherein the coefficient removal threshold is: Among them, t represents the coefficient removal threshold, σ0 represents the variance of the transformation coefficient, σ p represents the standard deviation of abnormal noise, median(d p ) represents the median value of the transform coefficients at the pth decomposition scale; The denoising model building unit is used to build a signal denoising model by using the coefficient removal threshold.

4. The sleep quality analysis system based on deep learning according to claim 3, characterized in that: The signal denoising model is: Among them, α represents the transformation coefficient, w j,k represents the transformation coefficient at the jth decomposition scale, m represents the convergence coefficient, sgn represents the sign function, and w j,k Represents the transformation coefficients for removing outliers.

5. The deep learning-based sleep quality analysis system according to claim 4, characterized in that: The preprocessing submodule includes: A coefficient setting unit, used to set initial transformation coefficients and convergence coefficients for the signal denoising model; A processing unit, used to process the transformation coefficients using the set signal denoising model to obtain the transformation coefficients with outliers removed; A reconstruction unit, used for reconstructing the transformation coefficients after removing the abnormal values ​​to obtain a reconstructed signal; A signal-to-noise ratio calculation unit, used to calculate the signal-to-noise ratio between the reconstructed signal and the original physiological signal and the environmental signal; The iteration unit is used to reset the transformation coefficient and the convergence coefficient when the signal-to-noise ratio is not within the preset range until the signal-to-noise ratio remains within the preset range.

6. The sleep quality analysis system based on deep learning according to claim 1, characterized in that: The signal characteristics include physiological signal sub-features and environmental signal sub-features; the physiological signal sub-features include: frequency band energy, power spectrum density, entropy and complexity of brain wave signals, average amplitude, frequency and zero crossing rate of electromyography signals, conductivity change and skin electrical response frequency of skin resistance signals; the environmental signal sub-features include: average temperature and humidity and change rate of temperature and humidity signals, carbon dioxide concentration and volatile organic compound level of air quality signals.

7. The sleep quality analysis system based on deep learning according to claim 1, characterized in that: The feature processing module comprises: The first traversal submodule is used to traverse all signal features, detect duplicate and invalid data, and delete the detected duplicate and invalid data; The second traversal submodule is used to traverse all signal features and detect missing features in the data, wherein features with missing values ​​exceeding 80% are deleted, and features with missing values ​​less than 80% are filled based on the KNN algorithm; The imbalance submodule is used to process the data imbalance of the feature data after data deletion and data filling based on the SMOTE algorithm; The normalization submodule is used to normalize the data after data imbalance processing based on the Z-score normalization method to obtain the normalized features.

8. The sleep quality analysis system based on deep learning according to claim 1, characterized in that: The sleep quality assessment model includes a multimodal input layer, a single signal convolution layer, a shared feature fusion layer, a timing analysis layer, a feature attention mechanism layer and an output assessment layer; the multimodal input layer is provided with an input channel corresponding to each signal type; the multimodal input layer is connected to the single signal convolution layer; the shared feature fusion layer is connected to the single signal convolution layer; the shared feature fusion layer is used to aggregate features from different single signal convolution layers, and integrate the features through a fusion algorithm to obtain fused features; The timing analysis layer is connected to the shared feature fusion layer, and the timing analysis layer is used to perform time series analysis on the fused features using a long short-term memory network; the feature attention mechanism layer is connected to the shared feature fusion layer, and the feature attention mechanism layer is deployed with an attention mechanism to dynamically adjust the feature weights of different signals; the output evaluation layer is connected to the timing analysis layer, and the output evaluation layer is used to perform a comprehensive calculation of the evaluation results based on the results of the time series analysis; the evaluation results include a sleep quality score.

9. The deep learning-based sleep quality analysis system according to claim 8, characterized in that: The calculation formula of the sleep quality score is: Among them, SQ is the sleep quality score, N is the number of signal features, a i is the attention weight of the i-th signal feature, f i is the value of the i-th signal feature, LSTM output is the output result of the LSTM layer, M is the number of environmental signal features, and w j is the contribution coefficient of the jth environmental feature, p j is the value of the jth environmental signal feature, such as average temperature, humidity change, and air quality index. α controls the impact of the summary of single signal features on the total score. β controls the impact of the dynamic characteristics of the time series. γ adjusts the proportion of environmental features in the total score. δ is a bias term used to adjust the benchmark of the final score.

Images