CN119950941A - A sleeping cabin for emotional memory regulation - Google Patents
A sleeping cabin for emotional memory regulation Download PDFInfo
- Publication number
- CN119950941A CN119950941A CN202510037148.0A CN202510037148A CN119950941A CN 119950941 A CN119950941 A CN 119950941A CN 202510037148 A CN202510037148 A CN 202510037148A CN 119950941 A CN119950941 A CN 119950941A
- Authority
- CN
- China
- Prior art keywords
- sleep
- layer
- feature
- subject
- stimulation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention belongs to the technical field of medical equipment, and particularly relates to a sleep cabin for emotion memory regulation and control. The sleep cabin comprises a sleep bed for a subject to sleep, a sleep emotion multi-module monitoring device for monitoring sleep sign data of the subject in real time, a data discrimination and decision judgment device, a multi-sensory stimulation intervention strengthening device for stimulating the subject according to the stimulation mode and parameters and realizing regulation of the sleep, wherein the data discrimination and decision judgment device is integrated with a decision operation unit for obtaining a stimulation mode and parameters for regulating and controlling the sleep of the subject according to the sleep sign data. The invention can regulate and control emotion memory in sleep of patients with depression (also can be used for common people and people with sleep disorder), eliminate negative memory, strengthen positive memory, improve sleep, adjust emotion, relieve depression symptoms and improve life quality.
Description
Technical Field
The invention belongs to the technical field of medical equipment, and particularly relates to a sleep cabin for emotion memory regulation and control.
Background
Depression is one of the most common mental problems worldwide, being a ubiquitous psychological disease. Patients often exhibit reduced mood, reduced interest, and insufficient energy.
The severity of depression prompted extensive theoretical and practical studies of its pathogenesis by numerous researchers, revealing the pathological features of depressed patients, including ① sustained and long-term mood-down or imbalance in mood memory regulation, enhancement of ② negative memory and attenuation of positive memory, ③ sleep disorder.
Methods of treating depression include drug therapy, psychotherapy, and physical therapy, but these methods vary widely in therapeutic effect among individuals. This discrepancy makes the contradiction between the increasing number of patients and the perfected treatment regimen increasingly prominent, and there is an urgent need to develop specific, targeted, widely applicable, and operationally robust devices and methods to achieve personalized, precise treatment of depressive patients.
Clinically, about 70% of depressed patients have sleep problems, and the population with sleep problems is also at significantly higher risk of suffering from depression or anxiety than the population with normal sleep. Thus, treatment of depression by intervention in sleep of depressed patients is a new therapeutic approach (furthermore, intervention in sleep of normal persons may also have a preventive effect on depression). In the prior art, although research has been conducted to investigate the relationship between depression and sleep, how to determine specific means and parameters for intervention in sleep of a patient suffering from depression is still a problem to be solved.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a sleep cabin for emotion memory regulation and control.
A sleep compartment for mood memory regulation comprising:
A sleep bed for the subject to sleep;
the sleep emotion multi-module monitoring device is used for monitoring sleep sign data of a subject in real time;
The data discrimination and decision judgment equipment is integrated with a decision operation unit and is used for obtaining a stimulation mode and parameters for regulating and controlling the sleep of the subject according to the sleep sign data;
The multisensory stimulation intervention strengthening device is used for stimulating the subject according to the stimulation mode and the parameters so as to realize the regulation and control of sleep;
Wherein the decision operation unit comprises:
an information receiving port configured to receive sleep sign data of a subject in real time;
the central computer processing system is configured to input the sleep sign data into a machine learning model, obtain a judging result of the sleep emotion of the subject through the machine learning model, and regulate and control a stimulus mode and parameters of the sleep of the subject, wherein the machine learning model is a mixed depth neural network model;
and the information sending port is configured to output the judging result of the sleep emotion of the subject, the stimulation mode and the parameters.
Preferably, the sleep emotion multi-module monitoring device comprises at least one of a contact polysomnography, an optical fiber sensing sleep monitoring device and an intelligent glove integrated with a sensor.
Preferably, the multi-sensory stimulation intervention strengthening device comprises at least one of a VR visual stimulation device, an auditory stimulation device, a tactile stimulation device, an olfactory stimulation device, a temperature regulation system and an air humidity regulation device.
Preferably, the multi-sensory stimulation intervention enhancement device is configured to stimulate the subject when the subject enters a rapid eye movement sleep stage.
Preferably, the sleep sign data comprises at least one of an electroencephalogram, an electrocardiogram, a heart rate, a blood pressure, a FVC, a FEV1/FVC, an electromyogram, a body temperature, wherein each FVC, a FEV1/FVC means Forced Vital Capacity (FVC), a first second forced breathing volume (FEV 1), a one second rate (FEV 1/FVC).
And/or, in the central computer processing system, inputting basic information of the patient into the machine learning model, the basic information including at least one of age, sex, height, weight, waist circumference, smoking history, drinking history, medical history/hospitalization history/accident, operation history, transfusion history, allergy history, long-term medication condition, family genetic disease;
And/or the stimulation mode comprises at least one of a touch stimulation, a temperature sense stimulation, an auditory stimulation, an olfactory stimulation, an air humidity, a transcranial direct current stimulation.
Preferably, the hybrid deep neural network model includes:
The original data extraction layer is used for decoding, decompiling and disassembling the input original data and converting the original data into data types which are more readable and easier to understand;
The feature extraction layer is used for extracting features from the processing results of the original data extraction layer;
the feature vector generation layer is used for generating a plurality of feature vectors by respectively using different algorithms aiming at the existing type features and the discrete type features, wherein part of the features are spliced by a fuzzy algorithm to jointly generate the feature vectors;
The multi-mode sleep emotion detection layer is used for obtaining a judging result of the sleep emotion of the subject according to the feature vector, and a stimulation mode and parameters for regulating and controlling the sleep of the subject;
The multi-mode sleep emotion detection layer comprises an initial network layer and a final network layer, wherein the initial network layer comprises a plurality of mutually unconnected DNN networks, the feature vectors are mutually and independently sent into the initial network layer, the final network layer comprises a DNN network structure, the multi-mode sleep emotion detection layer adopts a pre-fusion mode to fuse deep features, and the DNN network structure of the final network layer outputs a judging result of sleep emotion of a subject and a stimulation mode and parameters for regulating and controlling the sleep of the subject.
Preferably, in the initial network layer, each independent DNN network is composed of an input layer and a plurality of hidden layers, in the DNN network, neurons in the middle of each layer are not connected with each other, each layer of neurons is only connected with the adjacent front and back layers, and the layers are in a fully connected structure;
In the final network layer, the DNN network structure consists of a fusion layer, a plurality of hidden layers and an output layer, wherein the output layer has only one neuron and is responsible for outputting an analysis result.
Preferably, in the feature vector generation layer, the existing feature vector generation method includes:
Sequentially inquiring whether the corresponding feature of each sample appears in the feature database, if so, the feature is assigned as 1, otherwise, the feature is assigned as 0;
The discrete feature vector generation method comprises the following steps:
The method comprises the steps of normalizing feature data, carrying out K-means clustering on the normalized data to obtain a group of centroids of each feature data, respectively calculating distances between the feature data extracted from a sample and different centroids of corresponding types, selecting the centroid with the shortest distance, wherein the length of a vector, which points to a current feature point, of the centroid projected on different dimensions represents similarity of the feature of the data and the centroid on the different dimensions, taking the projection length of 0.4 as a threshold value, and setting elements in feature vectors corresponding to features with the projection length exceeding 0.4 as 0, otherwise setting the elements as 1.
Preferably, the feature vector generation layer extracts a feature vector for each discrete feature, and splices all existing features to generate a feature vector.
Preferably, in the feature vector generation layer, for extracting incomplete features, 0 vector is used for complement.
The invention provides a sleep cabin and a control system thereof, wherein the system can realize real-time sleep regulation of 'monitoring-processing-feedback-re-monitoring-processing-better feedback' by deep learning, and can carry out individual, adjustable and precise stimulation effect regulation on the sleep of a subject.
According to the purpose, the invention provides an optimized mixed depth neural network model, compared with the existing similar model, the mixed depth neural network model is optimized for uncertainty, accuracy and individuation, and is a multi-mode fusion emotion recognition model based on Choquet fuzzy integration, which is the most robust, most accurate and complete data learning and prediction method reported so far.
Hybrid deep neural network models have been proposed in the prior art for drug-drug interaction (DDI) prediction. The invention is applied to emotion memory regulation for the first time, and the model is optimized aiming at the purpose, so that the accuracy and individuation are further improved.
The advantages are specifically expressed in that:
1. The mixed depth neural network model can provide more comprehensive data analysis by integrating information (such as respiration, electrocardio, body temperature and the like) from different data sources. Fuzzy integration may help to handle uncertainties and ambiguities in clinical data so that models can learn and make predictions from such data more robustly.
2. The mixed depth neural network model can diagnose diseases or forecast treatment results more accurately by combining with fuzzy integration, especially under the condition of incomplete data or variability.
3. Personalized treatment planning fuzzy integration can help the hybrid deep neural network model of the present invention better understand and predict individual patient responses to treatment, thereby helping the sleep compartment to formulate a more personalized treatment plan.
4. Optimal decision support in clinical decision support systems, fuzzy integration can help weigh the importance of different factors and formulate a more comprehensive and balanced treatment scheme for the sleep compartment.
5. Support of interpretable artificial intelligence (XAI) fuzzy integration provides a natural way to explain how a model makes decisions on combining different factors, which can enhance the interpretability of the model, which is particularly important for medical applications, as it can provide a more transparent decision process, which is critical to improving the trust and acceptance of AI systems by doctors. Depression patients are characterized by continuous and long-term mood swings, enhanced negative and reduced positive memory, and sleep disorders, whose regulation is central to the radical management of depression. The sleep cabin system can regulate and control emotion memory sleep in sleep of patients with depression (also can be used for common people and people with sleep disorder), eliminate negative memory, enhance positive memory, improve sleep, adjust emotion, relieve depression symptoms and improve life quality. Therefore, the invention has good application prospect.
It should be apparent that, in light of the foregoing, various modifications, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
The above-described aspects of the present invention will be described in further detail below with reference to specific embodiments in the form of examples. It should not be understood that the scope of the above subject matter of the present invention is limited to the following examples only. All techniques implemented based on the above description of the invention are within the scope of the invention.
Drawings
FIG. 1 is a schematic diagram of the components of a sleep module system of the present invention;
FIG. 2 is a schematic view of a preferred construction of the sleeping cabin of the present invention;
FIG. 3 is a schematic flow chart of a hybrid deep neural network model according to the present invention;
FIG. 4 is a schematic diagram of a network structure of a multi-modal sleep emotion detection layer according to the present invention;
FIG. 5 is a schematic diagram of fuzzy integral fusion of multiple classifiers.
Detailed Description
It should be noted that, in the embodiments, algorithms of steps such as data acquisition, transmission, storage, and processing, which are not specifically described, and hardware structures, circuit connections, and the like, which are not specifically described may be implemented through the disclosure of the prior art.
Example 1 sleep cabin for mood memory modulation
The sleep cabin provided by the embodiment can detect the sleep sign data of the subject in real time, judge the sleep emotion of the subject based on the sleep sign data, provide proper stimulus in real time and regulate and control the sleep of the subject.
Specifically, as shown in fig. 1-2, the system of the present embodiment includes:
A sleep bed for the subject to sleep;
the sleep emotion multi-module monitoring device is used for monitoring sleep sign data of a subject in real time;
The data discrimination and decision judgment equipment is integrated with a decision operation unit and is used for obtaining a stimulation mode and parameters for regulating and controlling the sleep of the subject according to the sleep sign data;
and the multisensory stimulation intervention strengthening equipment is used for stimulating the subject according to the stimulation mode and the parameters so as to realize the regulation and control of sleep.
The specific structure of each part is as follows:
1. Sleep emotion multi-module monitoring equipment
The sleep emotion multi-module monitoring device may be implemented by existing techniques. As a preferred manner, the sleep emotion multi-module monitoring device of the present embodiment includes the following components:
1. device for connecting head end of sleeping bed
The contact type polysomnography is particularly adopted and is used for recording electroencephalogram, electrocardiogram and electromyogram, and recording electroencephalogram, electrocardiogram and electromyogram information of a patient in the sleeping process;
The principle is that electrodes attached to the skin transmit bioelectric information to the sensor, and the different parts transmit information together. The sensor cable writes the signal to the head box device, and then the analog signal is converted into a digital signal after the process of amplifying, filtering and digitizing by the head box, and is transmitted to the fixed station. The signal can also be transmitted to the stationary station via the auxiliary input device, the signal being amplified and adjusted directly in the auxiliary device.
2. Sleep monitoring equipment based on low-cost optical fiber sensing
A low cost fiber optic sensing device embedded in a mattress that can be used to measure respiratory rate and heart rate for extended periods of time, non-invasively and anti-electromagnetic interference. Four behavior states (not in bed, lying, moving and leaving bed) can be distinguished, breathing and heart rate under different positions and postures can be measured, the method is suitable for long-term sleep monitoring, and acquired data can be used for analyzing diseases such as sleep disorder and the like.
The system comprises a mattress, a plurality of temperature-pressure integrated fiber bragg grating sensors, a demodulator and an upper computer, wherein the plurality of physical sign fiber bragg grating sensors are fixedly arranged in a gap between the mattress and a bed net and keep the depth on the same horizontal plane, and the plurality of physical sign fiber bragg grating sensors are connected with the demodulator and transmit data to the upper computer. The system identifies the pressure born by the sensor by detecting the wavelength of the peak splitting point of the fiber bragg grating spectrum signal, identifies the ambient temperature of the sensor by detecting the wavelength drift amount of the center wavelength of the fiber bragg grating spectrum signal, monitors the multi-sign information of a user when the user is in a bed, comprises the body temperature, the respiratory rate, the heart rate, the falling-off state and the like, and has the advantages of simple structure, small volume, electromagnetic interference resistance, low cost and the like.
3. Left side of the sleeping bed is connected with the front end of Dormio system
The time and state of sleep onset are monitored. Three physiological indicators of sleep onset are used.
① The subject lies down while sleeping and gently closes their hand, allowing a bending sensor to monitor the gradual loss of muscle tone by stretching the hand, and the sensors around the index finger can track the muscle tone (Kelly JM,Strecker RE,Bianchi MT.Recent developments in home sleep-monitoring devices.ISRN Neurol.2012;2012:768794.doi:10.5402/2012/768794.Epub 2012Oct 14.PMID:23097718;PMCID:PMC3477711.). because the loss of muscle tone is temporally related to the appearance of sleep images.
② Heart rate decline- (Herlan A,Ottenbacher J,Schneider J,Riemann D,Feige B.Electrodermal activity patterns in sleep stages and their utility for sleep versus wake classification.J Sleep Res.2019Apr;28(2):e12694.doi:10.1111/jsr.12694.Epub 2018May 2.PMID:29722079.), in line with loss of muscle tone-the heart rate can be monitored in the middle finger of the subject
③ Skin electrical activity (EDA) is consistent with loss of muscle tone (Herlan A,Ottenbacher J,Schneider J,Riemann D,Feige B.Electrodermal activity patterns in sleep stages and their utility for sleep versus wake classification.J Sleep Res.2019
Apr, 28 (2): e12694.Doi:10.1111/jsr.12694.Epub 2018May 2.PMID:29722079.) EDA can be measured by two electrodes at the bottom of the wrist.
4. Integrated equipment for sleep monitoring
The integrated device is an intelligent glove and comprises a blood oxygen detection module, a humidity detection module, a temperature detection module, a blood pressure detection module and/or a pulse detection module. The blood oxygen detection module is used for detecting blood oxygen parameters and respiratory parameters of a subject, further recording the change conditions of blood oxygen and respiratory functions when emotion changes, the humidity detection module is used for detecting the secretion condition of palm body diaphoresis fluid of the subject, the temperature detection module is used for detecting body temperature data and the change condition of the subject, the blood pressure detection module is used for detecting blood pressure parameters of the subject, and the pulse detection module is used for detecting pulse parameters of the subject. In order to improve the accuracy of the physiological characteristic parameters detected by each detection module, the blood oxygen detection modules are arranged at fingertip positions inside the intelligent glove, the blood oxygen detection modules can be multiple and respectively and correspondingly arranged at each fingertip, the humidity detection modules are arranged at palm positions inside the intelligent glove, the temperature detection modules are arranged at outer wrist positions inside the intelligent glove, and the blood pressure detection modules and/or the pulse detection modules are arranged at inner wrist positions inside the intelligent glove. Preferably, the smart glove may be a smart glove made of an elastic breathable material. The elastic ventilation material can improve the comfort level of a subject, and the ventilation material can improve the accuracy of physiological characteristic parameter detection. The various detection modules described above can be implemented according to the prior art.
An advantageous embodiment of the integrated sleep monitoring device may be arranged according to the chinese patent application "2020212854437 mental health detection system".
5. Recording storage device
Comprises a memory chip for recording various vital sign data such as specific change conditions of blood oxygen, respiratory function and the like during emotion change,
6. Communication apparatus
The sleep bed is connected with a Bluetooth data processing element and is a core element of the sleep emotion multi-module monitoring equipment and used for summarizing and transmitting data. The modules adopt wireless modes such as Bluetooth and WiFi or wired modes such as USB and HDMI to transmit and exchange data and signals.
2. Data discrimination and decision-making apparatus
The decision operation unit is used for selecting a stimulation mode and parameters suitable for the subject according to the sleep sign data of the subject. The system comprises:
an information receiving port configured to receive sleep sign data of a subject in real time;
the central computer processing system is configured to input basic information of a subject and the sleep sign data into a machine learning model, and obtain a stimulation mode and parameters for regulating and controlling the sleep of the subject through the machine learning model;
and the information sending port is configured to output the judging result, the stimulation mode and the parameters of the sleep emotion of the subject.
Wherein,
The sleep sign data comprises the following indexes of electroencephalogram, heart function (electrocardiogram, heart rate and blood pressure), respiratory function (FVC, FEV1 and FEV 1/FVC), electromyogram and body temperature.
The basic information includes the following indexes of age, sex, height, weight, waistline, smoking history, drinking history, disease history, operation history, blood transfusion history, allergy history, long-term medication condition and familial genetic disease.
The machine learning model is a hybrid deep neural network model, and the hybrid deep neural network model is shown in fig. 3, and has the specific structure that:
The original data extraction layer is used for decoding, decompiling and disassembling the input original data and converting the original data into data types which are more readable and easier to understand;
The feature extraction layer is used for extracting features from the processing results of the original data extraction layer;
the feature vector generation layer is used for generating a plurality of feature vectors by respectively using different algorithms aiming at the existing type features and the discrete type features, wherein part of the features are spliced to generate the feature vectors together;
The multi-mode sleep emotion detection layer is used for obtaining a judging result of the sleep emotion of the subject according to the feature vector, and a stimulation mode and parameters for regulating and controlling the sleep of the subject;
the multi-mode sleep emotion detection layer comprises an initial network layer and a final network layer, wherein the initial network layer comprises a plurality of mutually unconnected DNN networks, the feature vectors are mutually and independently sent into the initial network layer, the final network layer comprises a DNN network structure, the final layer of each DNN network in the initial network layer is fully connected with the first layer of the DNN network in the final network layer, the multi-mode sleep emotion detection layer adopts a pre-fusion mode to fuse deep features, and the DNN network structure of the final network layer outputs a judging result of sleep emotion of a subject and a stimulus mode and parameters for regulating and controlling the sleep of the subject.
In the initial network layer, each independent DNN network consists of an input layer and a plurality of hidden layers, in the DNN network, neurons in the middle of each layer are not connected with each other, each layer of neurons are only connected with the adjacent front and back layers, and the layers are of a full-connection structure, in the final network layer, the DNN network structure consists of a fusion layer, a plurality of hidden layers and an output layer, and the output layer is only one neuron and is responsible for outputting an analysis result.
And the DNN of the final network layer outputs the judgment result of the sleep emotion of the subject, and the stimulation mode and parameters for regulating and controlling the sleep of the subject.
The workflow of the hybrid deep neural network model is specifically as follows:
(1) The original data extraction layer performs:
All basic information and sleep sign data are classified in the first step into presence type features (classified variables such as presence/absence, normal/abnormal, too high/normal/too low, etc.) and discrete type features (specific values), and then the original data are decoded, decompiled and disassembled to be converted into data types which are more readable and easier to understand.
(2) The feature extraction layer performs:
Feature extraction is performed from the converted data, and as a preferred mode, a specific method of feature extraction may use a CNN model or an RNN model. As a preferred manner, examples of extraction of the individual sleep sign data and the basic information are shown in the following table:
TABLE 1 examples of feature extraction of sleep sign data and basic information
(3) The feature vector generation layer performs:
The feature information extracted in this embodiment is classified into various types, and two main types, i.e., existing type features and discrete type features, according to the difference of expression forms.
Specifically, the method for generating the corresponding feature vector by the presence type feature and the discrete type feature comprises the following steps:
(1) Presence type feature vector generation algorithm
The generation process of the existing feature vector is visual and simple. First, all the presence features extracted by the feature extraction layer are formed into a large presence feature database. Then, it is sequentially queried whether the corresponding feature of each sample appears in the feature database, if so, the feature is assigned to 1, otherwise, 0, and the system is symbiotic to a plurality of existing feature vectors (for example, 2 feature vectors of respiratory frequency and combined feature vectors composed of heart rate feature, electrocardiogram feature and body temperature feature).
(2) Discrete feature vector generation algorithm
The characteristic of discrete data is a group of combinations of frequency with great difference, therefore, before generating a characteristic vector, data normalization is needed first, the system uses a linear function normalization (MinMax Scaling) method to normalize the frequency data of the characteristic into the range of [0,1] in sequence, the normalization equation is shown in the following formula, wherein x and y respectively represent the characteristic data before and after normalization, and Max and Min respectively represent the maximum value and the minimum value of the frequency:
the feature data are then separately K-means clustered (reference ：I.B.Mohamad,D.Usman.Standardization and its effects on k-means clustering algorithm[J].Research Journal of Applied Sciences,Engineering and Technology,2013,6(17):3299-3303), gets a set of centroids for each feature data. The feature data extracted from the sample will be separately calculated for distance to the different centroids of the corresponding type, using the Euclidean distance calculation formula (below), t represents the coordinate dimensions of points M and N.
The shortest centroid is selected, and the length of the projection of the vector directed to the current feature point by the centroid in different dimensions represents the similarity of the data and the centroid in the features in different dimensions. A longer projection length in one dimension indicates a lower similarity of features in that dimension, and a shorter opposite projection length indicates a higher similarity. In the system, the projection length of 0.4 is used as a threshold value, and elements in the feature vector corresponding to the feature in the dimension of which the projection length exceeds 0.4 are set to be 0, otherwise, the elements are set to be 1, so that the large influence of the feature of partial weak correlation on the subsequent prediction task is avoided, and the calculated amount in the deep learning process is simplified.
(4) The multi-mode sleep emotion detection layer performs:
The feature vector generated by the feature vector generation layer is sent to an initial network layer composed of a plurality of DNN networks independently, the multi-mode sleep emotion detection layer adopts a pre-fusion mode to fuse deep features, and a judgment result of the sleep emotion of the subject and a stimulation mode and parameters for regulating and controlling the sleep of the subject are output.
Through the mixed depth neural network model, the sleep emotion of the subject can be judged in real time, and the result of the stimulation mode and parameters for regulating and controlling the sleep of the patient can be obtained.
The mixed deep neural network model is used for judging the sleep emotion of the subject, and has the following characteristics:
The embodiment is based on the existing multi-Mode DNN (MDNN) model, and the partial features are spliced together in a feature vector generation layer through a fuzzy algorithm to generate feature vectors (namely, fuzzy integration is added),
There are 2 special cases when feature stitching is performed in the feature vector generation layer:
(1) When there is an interaction effect between the modalities, the joint effect between the two modalities is not simply equivalent to the sum of the single-modality effects
(2) When the classification result is not clear, and means that one sample cannot be clearly judged as a positive class or a negative class.
At this time, we can use a fuzzy integrator-fuse (as shown in fig. 5) to effectively fuse the multi-modal results.
Compared with the existing fusion operator, the fuzzy integration has the outstanding advantages that the importance of the classifier corresponding to each mode can be better reflected, and the interaction relation among the modes can be better represented. There are three forms of interaction between general classifiers Di, dj:
(1) Negative synergy if the overall importance of the sub-classifier Di, dj is less than the sum of the individual sub-classifiers. In other words, the fusion classifier Ci, cj cannot improve the classification accuracy of a certain class compared to the single classifier Di, dj.
(2) And in cooperation, if the total meaning of the sub-classifiers Di and cfj to the classification is larger than the sum of the meanings of the single classifier. In other words, compared with the single classifiers Di, dj, the fusion classifier Ci, cj can improve the classification accuracy of a certain class.
(3) Uncorrelated (INDEPENDENCY) is to add the effects of the individual classifiers Di, dj on the classification accuracy.
When interaction exists between different modes, the embodiment uses fuzzy measure to express the interaction, and f is known by Choquet fuzzy integral formula, namely, f is the membership of the modes, namely, the expressive ability of the different modes on the classifier, and then multiplies the measure, and finally sums to obtain the score of the input on the class.
X is taken as any set, i.eMu is a fuzzy measure defined on the set X, fXfwdarw 0,1 is a non-negative real valued measurable function defined on X, the values are plotted on function fOrdering is performed and the minimum value of a may be 0, a i representing the set of a i. Choquet the fuzzy integral formula is as follows:
as can be seen from the fuzzy metric table below, if the input is in three modes, the metrics include not only all single-mode metrics, but also bimodal combinations and trimodal combination metrics, and interactions among multiple modes are shown on the metrics while representing importance of the single modes. Fuzzy integration is mainly characterized by the calculation of measures to characterize interactions between different modalities. Therefore, the embodiment models the multi-modal fusion by means of the fuzzy system, and the importance and interactivity characteristics among the multiple modalities can be simultaneously represented by using the measure, so that an optimal measure calculation method for the multi-modal fusion is found.
Table 2 trimodal fuzzy metric table
In summary, the fuzzy integration added in this embodiment makes the interaction relationship between different data be considered on the basis of two data analysis of "mutually independent unit system" and "pre-fusion", because there is a correlation between data, the single-mode measure is weighted and summed (Choquet fuzzy integration) through the degree of elevation to form a multi-mode measure, and then a decision fusion framework of the multi-mode multi-classifier is constructed according to the fuzzy measure calculation result, and the framework trains and optimizes a plurality of different types of emotion state basic classifiers for each mode respectively, so as to obtain the decision information of a plurality of basic classifiers on different mode data. And then, respectively integrating and calculating the decision information of each mode by using a newly designed multi-basis classifier decision information integrating method, and providing input information for the calculation of the subsequent fuzzy measure, so that the method is far superior to the existing MDNN model.
The "mutually independent unit system" refers to an independent unit system (because one electrocardio contains a lot of contents, such as frequency and waveform, atrial flutter is F wave, atrial fibrillation is F wave) which are all indexes of "electroencephalogram", "myoelectricity" and "electrocardio". The pre-fusion refers to the step of fusing different feature extraction algorithms and classification algorithms in the feature extraction and classification stage of the data set so as to obtain a better data analysis result, and the post-fusion refers to the step of fusing a plurality of classifiers after the feature extraction and classification stage so as to obtain a better classification result.
In the practical model training and detecting use process, it is not necessarily ensured that all basic information and sleep sign data can be completely collected, and the hybrid deep neural network model in this embodiment also has an appropriate coping scheme for such a situation. The method comprises the following steps:
① Training scheme for incomplete feature extraction
The hybrid deep neural network model of the embodiment adopts a multi-mode characteristic input structure, and each DNN in the initial network layer is mutually independent and only fused in the final network layer. This makes it possible for each individual DNN network in the initial network layer to train individually in turn. In the model training process, the feature information which is not extracted is not subjected to 0-vector completion, and a plurality of types of successfully extracted feature vectors are sequentially input into a DNN (digital network) of a corresponding initial network layer for training. And finally training a final network layer on the basis of the completion of the training of the whole initial network layer. The training strategy ensures that the system maintains feature independence to the greatest extent, simultaneously reduces the dependence on success of simultaneous extraction of all features, and maximally uses all extracted feature information.
② Test scheme for incomplete feature extraction
Under the condition that the basic information or sleep sign data of the detected subject is extracted incompletely, the detection link is completed by using a 0 vector, and then all feature vectors are input into a mixed depth neural network model for detection. Because the final output of the initial network corresponding to the 0 vector input only calculates the bias term, the influence of the 0 vector input on the final classification result is small, and the accuracy of the whole detection result is ensured.
3. Multi-sense stimulation intervention strengthening equipment
The multisensory stimulation intervention enhancement device may be implemented by prior art techniques. As a preferred manner, the multisensory stimulation intervention enhancement device of this embodiment comprises the following components:
1. VR visual stimulus equipment, namely, visual stimulus, such as teaching sheets, emotion videos and the like, which are watched during waking, are played to a sleeper to watch on the premise of not influencing sleeping, so that the effect of enhancing relevant memories is achieved.
2. Auditory stimulation devices (which may be external playback boxes or in-ear players) are used to provide a user with soothing, pleasant or exciting music, sounds or speech. The device selects different types, styles, rhythms, and volumes of audio content.
3. Tactile stimulation apparatus:
The specific form can be as follows:
(1) The glove type glove is wearable, and a multipoint stimulation module is arranged in the glove;
(2) A sole massage device;
(3) A mattress integrated massage device;
For providing a soft, comfortable or stimulated tactile sensation to the user. The device can adjust the haptic stimulus for different locations, intensities, frequencies and patterns.
4. An olfactory stimulating device (which may be an indoor aromatherapy machine) for providing a fresh, aromatic or exciting scent to a user. The device may select different kinds, concentrations and durations of olfactory stimuli.
5. Treatment room configuration equipment, including a temperature regulation system (central air conditioner) and an air humidity regulation equipment (indoor humidifier).
When the sleep cabin system of the embodiment works, the sleep cabin system can further comprise the following characteristics:
1. Selection of timing (sleep stage):
human sleep can be divided into two major parts, rapid Eye Movement (REM) sleep and non-rapid eye movement (NREM) sleep.
In combination with the above theory, the regulation of the sleep and memory of a patient during Rapid Eye Movement (REM) sleep is a better option or will achieve more pronounced therapeutic effects. In this embodiment, the sleep emotion multi-module monitoring device can determine when the patient enters the rapid eye movement sleep stage, so that VR multi-sensory stimulation is performed by the multi-sensory stimulation intervention strengthening device at this stage.
2. Selection of parameters of a multisensory stimulation intervention enhancement module:
The specific stimulation parameters are determined according to the discrimination results of the data discrimination and decision-making discrimination equipment, so as to aim at bringing better treatment effect to the patient by using the optimal stimulation mode and parameters.
The parameters of the various stimulation modes are selected as follows:
tactile stimulation, namely massaging hands for 0.5-2 times/S, and the strength is gentle;
The sole massage is performed for 0.5-2 times/S, and the force is gentle to moderate and strong;
auditory stimulus, namely 20-40 dB of sound intensity and graceful and soft melody of sound property;
olfactory stimulus, namely comfortable and pleasant smell, such as elegant flower fragrance, sweet fruit fragrance and fresh grass fragrance;
the indoor environment temperature is 18-25 ℃;
the indoor environment humidity is generally suitable for 45-65%, the relative humidity is 40-80% in summer refrigeration, 30-60% in winter heating, 45-50% for old people and children, and 40-50% for respiratory disease patients such as asthma.
According to the embodiment, the sleep cabin device can monitor sleep sign data of a subject in a sleep process in real time, judge emotion of the subject in the sleep process in real time, and provide appropriate stimulus in the sleep process of the subject by a stimulus model and stimulus parameters which need to be received. The invention can regulate and control emotion memory sleep in sleep of patients with depression (also can be used for common people and people with sleep disorder), eliminate negative memory, strengthen positive memory, improve sleep, adjust emotion, relieve depression symptoms and improve life quality, and has good application prospect.
Claims (10)
1. A sleep compartment for mood memory regulation comprising:
A sleep bed for the subject to sleep;
the sleep emotion multi-module monitoring device is used for monitoring sleep sign data of a subject in real time;
The data discrimination and decision judgment equipment is integrated with a decision operation unit and is used for obtaining a stimulation mode and parameters for regulating and controlling the sleep of the subject according to the sleep sign data;
The multisensory stimulation intervention strengthening device is used for stimulating the subject according to the stimulation mode and the parameters so as to realize the regulation and control of sleep;
Wherein the decision operation unit comprises:
an information receiving port configured to receive sleep sign data of a subject in real time;
the central computer processing system is configured to input the sleep sign data into a machine learning model, obtain a judging result of the sleep emotion of the subject through the machine learning model, and regulate and control a stimulus mode and parameters of the sleep of the subject, wherein the machine learning model is a mixed depth neural network model;
and the information sending port is configured to output the judging result of the sleep emotion of the subject, the stimulation mode and the parameters.
2. The sleep compartment of claim 1 wherein the sleep emotion multi-module monitoring device comprises at least one of a touch polysomnography, a fiber optic perceived sleep monitoring device, and a sensor-integrated smart glove.
3. The sleep compartment of claim 1 wherein the multi-sensory stimulation intervention enhancement device comprises at least one of a VR visual stimulation device, an auditory stimulation device, a tactile stimulation device, an olfactory stimulation device, a temperature regulating system, and an air humidity regulating device.
4. The sleep compartment of claim 3 wherein the multi-sensory stimulation intervention enhancement device is configured to stimulate the subject when the subject enters a rapid eye movement sleep stage.
5. The sleep compartment of claim 1, wherein the sleep sign data comprises at least one of an electroencephalogram, an electrocardiogram, a heart rate, a blood pressure, FVC, FEV1/FVC, an electromyogram, a body temperature;
And/or, in the central computer processing system, inputting basic information of the patient into the machine learning model, the basic information including at least one of age, sex, height, weight, waist circumference, smoking history, drinking history, medical history/hospitalization history/accident, operation history, transfusion history, allergy history, long-term medication condition, family genetic disease;
And/or the stimulation mode comprises at least one of a touch stimulation, a temperature sense stimulation, an auditory stimulation, an olfactory stimulation, an air humidity, a transcranial direct current stimulation.
6. The sleep compartment of claim 1 wherein the hybrid deep neural network model comprises:
The original data extraction layer is used for decoding, decompiling and disassembling the input original data and converting the original data into data types which are more readable and easier to understand;
The feature extraction layer is used for extracting features from the processing results of the original data extraction layer;
the feature vector generation layer is used for generating a plurality of feature vectors by respectively using different algorithms aiming at the existing type features and the discrete type features, wherein part of the features are spliced by a fuzzy algorithm to jointly generate the feature vectors;
The multi-mode sleep emotion detection layer is used for obtaining a judging result of the sleep emotion of the subject according to the feature vector, and a stimulation mode and parameters for regulating and controlling the sleep of the subject;
the multi-mode sleep emotion detection layer comprises an initial network layer and a final network layer, wherein the initial network layer comprises a plurality of mutually unconnected DNN networks, the feature vectors are mutually and independently sent into the initial network layer, the final network layer comprises a DNN network structure, the last layer of each DNN network in the initial network layer is fully connected with the first layer of the DNN network in the final network layer, the multi-mode sleep emotion detection layer adopts a pre-fusion mode to fuse deep features, and the DNN network structure of the final network layer outputs a judging result of sleep emotion of a subject and a stimulation mode and parameters for regulating and controlling the sleep of the subject.
7. The sleep module as in claim 6, wherein each independent DNN network comprises an input layer and a plurality of hidden layers, wherein neurons in the middle of each layer are not connected with each other, each layer of neurons is connected with only the adjacent front and back layers, and the layers are in a fully connected structure;
In the final network layer, the DNN network structure consists of a fusion layer, a plurality of hidden layers and an output layer, wherein the output layer has only one neuron and is responsible for outputting an analysis result.
8. The sleep compartment of claim 6, wherein the feature vector generation layer comprises the following steps:
Sequentially inquiring whether the corresponding feature of each sample appears in the feature database, if so, the feature is assigned as 1, otherwise, the feature is assigned as 0;
The discrete feature vector generation method comprises the following steps:
The method comprises the steps of normalizing feature data, carrying out K-means clustering on the normalized data to obtain a group of centroids of each feature data, respectively calculating distances between the feature data extracted from a sample and different centroids of corresponding types, selecting the centroid with the shortest distance, wherein the length of a vector, which points to a current feature point, of the centroid projected on different dimensions represents similarity of the feature of the data and the centroid on the different dimensions, taking the projection length of 0.4 as a threshold value, and setting elements in feature vectors corresponding to features with the projection length exceeding 0.4 as 0, otherwise setting the elements as 1.
9. The sleep module as set forth in claim 6, wherein:
The feature vector generation layer extracts a feature vector for each discrete feature, and splices all existing features to generate a feature vector.
10. The sleep compartment of claim 6 wherein the feature vector generation layer is complemented with a 0 vector for extracting incomplete features.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510037148.0A CN119950941B (en) | 2025-01-09 | 2025-01-09 | A sleep pod for regulating emotional memory |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510037148.0A CN119950941B (en) | 2025-01-09 | 2025-01-09 | A sleep pod for regulating emotional memory |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN119950941A true CN119950941A (en) | 2025-05-09 |
| CN119950941B CN119950941B (en) | 2025-11-07 |
Family
ID=95592999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202510037148.0A Active CN119950941B (en) | 2025-01-09 | 2025-01-09 | A sleep pod for regulating emotional memory |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119950941B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119964733A (en) * | 2025-01-09 | 2025-05-09 | 四川大学华西医院 | Emotion prediction and control system and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102293656A (en) * | 2011-05-25 | 2011-12-28 | 四川大学华西医院 | Emotional stability evaluation system based on magnetic resonance imaging and evaluation method thereof |
| KR20150061107A (en) * | 2013-11-25 | 2015-06-04 | 재단법인대구경북과학기술원 | Bed system for the detection of sleep and sleep state detection method |
| US20200107766A1 (en) * | 2018-10-09 | 2020-04-09 | Sony Corporation | Electronic device for recognition of mental behavioral attributes based on deep neural networks |
| US20200346016A1 (en) * | 2019-05-02 | 2020-11-05 | Enhale Medical, Inc. | Systems and methods to improve sleep disordered breathing using closed-loop feedback |
| KR20220017083A (en) * | 2020-08-04 | 2022-02-11 | 주식회사 에이슬립 | Sleep analysis system |
| US11501794B1 (en) * | 2020-05-15 | 2022-11-15 | Amazon Technologies, Inc. | Multimodal sentiment detection |
| CN116649980A (en) * | 2023-06-06 | 2023-08-29 | 四川大学 | Emotion monitoring method, system, device and storage medium based on artificial intelligence |
| JP2024056292A (en) * | 2022-10-11 | 2024-04-23 | 国立大学法人大阪大学 | Diagnostic support information providing system, diagnostic support device, symptom index providing device, and symptom index providing program |
| KR20240077493A (en) * | 2022-11-14 | 2024-06-03 | 충북대학교병원 | Electronic apparatus for determining need for intubation to neonatal based on multimodal deep neural network, and control method thereof |
| CN118750000A (en) * | 2024-06-26 | 2024-10-11 | 长春理工大学 | A multimodal sleep stage emotion recognition method based on Dual-CNN+GAN |
-
2025
- 2025-01-09 CN CN202510037148.0A patent/CN119950941B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102293656A (en) * | 2011-05-25 | 2011-12-28 | 四川大学华西医院 | Emotional stability evaluation system based on magnetic resonance imaging and evaluation method thereof |
| KR20150061107A (en) * | 2013-11-25 | 2015-06-04 | 재단법인대구경북과학기술원 | Bed system for the detection of sleep and sleep state detection method |
| US20200107766A1 (en) * | 2018-10-09 | 2020-04-09 | Sony Corporation | Electronic device for recognition of mental behavioral attributes based on deep neural networks |
| US20200346016A1 (en) * | 2019-05-02 | 2020-11-05 | Enhale Medical, Inc. | Systems and methods to improve sleep disordered breathing using closed-loop feedback |
| US11501794B1 (en) * | 2020-05-15 | 2022-11-15 | Amazon Technologies, Inc. | Multimodal sentiment detection |
| KR20220017083A (en) * | 2020-08-04 | 2022-02-11 | 주식회사 에이슬립 | Sleep analysis system |
| JP2024056292A (en) * | 2022-10-11 | 2024-04-23 | 国立大学法人大阪大学 | Diagnostic support information providing system, diagnostic support device, symptom index providing device, and symptom index providing program |
| KR20240077493A (en) * | 2022-11-14 | 2024-06-03 | 충북대학교병원 | Electronic apparatus for determining need for intubation to neonatal based on multimodal deep neural network, and control method thereof |
| CN116649980A (en) * | 2023-06-06 | 2023-08-29 | 四川大学 | Emotion monitoring method, system, device and storage medium based on artificial intelligence |
| CN118750000A (en) * | 2024-06-26 | 2024-10-11 | 长春理工大学 | A multimodal sleep stage emotion recognition method based on Dual-CNN+GAN |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119964733A (en) * | 2025-01-09 | 2025-05-09 | 四川大学华西医院 | Emotion prediction and control system and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN119950941B (en) | 2025-11-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11577043B2 (en) | Brain stimulation system, method and apparatus based on artificial intelligence and storage medium | |
| KR102383921B1 (en) | Methods and system for customized sleep management | |
| CN109859570A (en) | Brain training method and system | |
| CN119950941B (en) | A sleep pod for regulating emotional memory | |
| CN109757926A (en) | The hardness adjusting method and device of intelligent mattress | |
| CN115381413B (en) | Self-adaptive bimodal emotion adjustment method and system | |
| CN119384243A (en) | Intelligent mattress lining system and associated methods | |
| KR20220089913A (en) | System and method for improving development disorder using deep learning module | |
| CN116327195A (en) | Wearable equipment and recommendation system based on EEG and fNIRS | |
| CN120783952A (en) | Virtual reality driven child tic syndrome multi-modal data evaluation and intervention system | |
| CN120280076A (en) | Breathing rehabilitation training guiding system based on virtual reality | |
| CN120526939A (en) | Health data recommendation method and system based on big data analysis | |
| CN119732802A (en) | Active intelligent bed interaction system and method based on biofeedback technology | |
| CN120597644B (en) | A Smart Adjustment Method and System for Shared Massage Chairs Based on Big Data | |
| CN119964733A (en) | Emotion prediction and control system and storage medium | |
| CN115316953B (en) | Control methods and related devices for physiological monitoring systems | |
| CN119650026B (en) | Control method and system of intelligent physiotherapy health care bed | |
| KR102478102B1 (en) | Method of treatment for insomnia | |
| CN121490230A (en) | A smart sleep aid system and method based on multi-band infrared spectroscopy and physiological feedback | |
| CN120837810A (en) | Collaborative intervention system and method based on multimodal emotion perception and five-tone therapy | |
| KR20110135702A (en) | Health care system using complex biosignal recognition and health care method using the same | |
| CN119455263A (en) | A biomagnetic therapy management system and magnetic therapy equipment | |
| Polo | Multimodal assessment of emotional responses by physiological monitoring: novel auditory and visual elicitation strategies in traditional and virtual reality environments | |
| Sharrab et al. | Deep Learning-Based Prediction of Music's Physiological Effects Using Muse2 Heart Rate, Mind Calm, and Stillness | |
| CN120531351A (en) | Insomnia intervention method based on virtual reality technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |