Sleep-aiding pet robot

Abstract

The invention discloses a sleep-aiding pet robot, and belongs to the technical field of sleep aiding, a sensor module, a control module and a sleep-aiding execution module are arranged in the robot body, the sensor module collects a multi-modal real-time interaction signal of a user, the control module preprocesses the multi-modal real-time interaction signal, and the sleep-aiding execution module executes the sleep-aiding execution module. The method comprises the following steps: acquiring a multi-modal real-time interaction signal group, preprocessing the multi-modal real-time interaction signal group to obtain a preprocessed multi-modal real-time interaction signal, matching the real-time interaction signal group based on a signal processing model, the preprocessed multi-modal real-time interaction signal, an AI algorithm and a preset sleep aiding rule to obtain a corresponding bionic interaction instruction, and performing staging processing on a sleep monitoring signal based on a sleep automatic staging algorithm of an attention mechanism. And the sleep-aiding execution module controls the execution of the corresponding module according to the bionic interaction instruction. The bionic interaction instruction is given through the AI algorithm in combination with the preset sleep aiding rule, bionic interaction is brought, sleep staging is carried out on the sleep process of the user, intervention opportunities can be selected conveniently, and the sleep state can be known conveniently.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of sleep aid, and particularly relates to a sleep aid pet robot. BACKGROUND

[0002] Sleep is an indispensable physiological phenomenon for human beings, and sufficient sleep can not only restore physical strength, but also maintain normal immune balance of the human body. Insufficient sleep can increase the risk of chronic diseases and easily lead to mental diseases such as depression. At present, due to the generally fast pace of life in today's society, more than 60% of adults in the world have sleep problems of different degrees, such as sleep disorders, due to excessive emotional stress or environmental factors. Sleep disorders mainly manifest in three situations: difficulty falling asleep, easy wake-up in the middle of the night, and early wake-up before the expected time and inability to fall asleep again. These symptoms can significantly reduce sleep quality.

[0003] Sleep stage staging is an important indicator for measuring sleep quality. Traditional sleep stage staging mainly relies on manual labeling of each stage by sleep experts, which not only has corresponding limitations, but also has high cost. Most of the sleep aid products in the prior art are mainly sleep awakening and medical drugs, which cannot bring corresponding bionic interaction, and the existing sleep aid products cannot automatically and accurately distinguish the sleep stage of the user, resulting in inaccurate intervention timing. Therefore, how to provide an effective solution to solve the problems that the existing sleep aid products in the prior art cannot bring corresponding bionic interaction and cannot automatically and accurately distinguish the sleep stage of the user, resulting in inaccurate intervention timing, has become a difficult problem to be solved in the prior art. SUMMARY

[0004] The purpose of the present application is to provide a sleep aid pet robot to solve the above problems existing in the prior art.

[0005] In order to achieve the above purpose, the present application adopts the following technical solutions: In a first aspect, the present application provides a sleep aid pet robot, comprising a pet main body shell, a sensor module, a control module and a sleep aid execution module are arranged in the pet main body shell, and the control module is in communication connection with the sensor module and the sleep aid execution module respectively: The sensor module is used for collecting multi-modal real-time interaction signals of a user, the multi-modal real-time interaction signals include action signals, sound signals and physiological signals, and the multi-modal real-time interaction signals are uploaded to the control module. The control module is configured to preprocess the multi-modal real-time interaction signal to obtain a preprocessed multi-modal real-time interaction signal, process the preprocessed multi-modal real-time interaction signal based on a signal processing model to obtain a real-time interaction signal group, match the real-time interaction signal group based on an AI algorithm and a preset sleep-aiding rule to obtain a corresponding bionic interaction instruction, and send the corresponding bionic interaction instruction to the sleep-aiding execution module. The control module is further configured to collect sleep monitoring signals within a preset time, process the sleep monitoring signals based on a sleep automatic staging algorithm based on an attention mechanism to obtain sleep staging results, wherein the sleep staging results comprise monitoring signal groups of different sleep periods, and construct a visual sleep staging model based on the sleep staging results. The sleep-aiding execution module is configured to control the breathing module, the brain wave tuning module, the light module, the environmental noise module and / or the music module to execute according to the corresponding bionic interaction instruction.

[0006] In a possible design, the control module is further configured to construct a multi-modal emotion state recognition model, recognize the monitoring signal groups of different sleep periods based on the multi-modal emotion state recognition model to obtain emotion states in different sleep states, and generate corresponding breathing control instructions according to the emotion states.

[0007] In a possible design, the sleep-aiding execution module is in communication connection with the breathing module, the brain wave tuning module, the light module, the environmental noise module and the music module respectively. The breathing module comprises an air bag or a flexible actuator, and is configured to simulate human breathing frequency and breathing action. The brain wave tuning module is configured to release a / b / theta wave band to a user. The light module is configured to turn on and adjust light brightness. The environmental noise module comprises a directional speaker and an environmental microphone, the directional speaker is configured to play sound capable of canceling interference noise or play white noise, and the environmental microphone is configured to collect interference noise in the environment. The music module is configured to play soothing sleep-aiding music or preset audio.

[0008] In a possible design, the physiological signals comprise real-time body temperature data and carbon dioxide concentration values, and the sensor module comprises an accelerometer, an audio sensor and a miniature non-contact physiological sensor. The accelerometer is configured to collect user action signals. The audio sensor is configured to collect user sound signals. The micro non-contact physiological sensor includes an infrared sensor for collecting real-time body surface temperature data of a user and a carbon dioxide sensor for collecting a carbon dioxide concentration value in a space.

[0009] In a possible design, the multimodal real-time interaction signal is preprocessed to obtain a preprocessed multimodal real-time interaction signal, including: The sound signal is pre-emphasized to obtain an emphasized sound signal; The action signal, the emphasized sound signal, the real-time body surface temperature data, and the carbon dioxide concentration value are subjected to outlier processing and missing value filling to obtain a filled action signal, a filled sound signal, a filled real-time body surface temperature data, and a filled carbon dioxide concentration value; The filled action signal, the filled sound signal, the filled real-time body surface temperature data, and the filled carbon dioxide concentration value are subjected to denoising processing to obtain a denoised action signal, a denoised sound signal, a denoised real-time body surface temperature data, and a denoised carbon dioxide concentration value; The denoised action signal, the denoised sound signal, the denoised real-time body surface temperature data, and the denoised carbon dioxide concentration value are subjected to normalization processing to obtain the preprocessed multimodal real-time interaction signal.

[0010] In a possible design, the real-time interaction signal group includes an action signal feature, a sound signal feature, and a physiological signal feature; and the signal processing model is used to process the preprocessed multimodal real-time interaction signal to obtain the real-time interaction signal group, including: The preprocessed action signal is subjected to time domain and frequency domain feature extraction based on the signal processing model to obtain an action signal feature, where the action signal feature includes an action time domain feature and an action frequency domain feature; The preprocessed sound signal is subjected to voiceprint feature extraction based on the signal processing model to obtain a sound signal feature, where the sound signal feature includes a voiceprint feature; The preprocessed physiological signal feature is subjected to frequency domain feature extraction based on the signal processing model to obtain a physiological signal feature, where the physiological signal feature includes a temperature frequency domain feature and a carbon dioxide frequency domain feature.

[0011] In a possible design, a sleep automatic staging algorithm based on an attention mechanism is used to process the sleep monitoring signal to obtain a sleep staging result, including: A CNN model is used to extract features from the sleep monitoring signal to obtain a plurality of monitoring signal features, and the plurality of monitoring signal features are sequentially processed to obtain a monitoring signal feature sequence; The monitoring signal feature sequence is subjected to data fusion to obtain a fused signal feature sequence; The self-attention mechanism is adopted to capture the dependency relationship of the fusion signal feature sequence, so that the processed fusion signal feature sequence is obtained. The sleep automatic staging algorithm based on the attention mechanism is used for staging processing on the processed fusion signal feature sequence, so that a sleep staging result is obtained.

[0012] In a possible design, the real-time interaction signal group is matched based on an AI algorithm and a preset sleep-aiding rule, so that corresponding bionic interaction instructions are obtained, including: The real-time interaction signal group is matched with the preset sleep-aiding rule, and it is judged whether a trigger condition of the preset sleep-aiding rule is reached; If the trigger condition is reached, the real-time interaction signal group and the preset sleep-aiding rule are processed by using the AI algorithm, so that the corresponding bionic interaction instructions are obtained.

[0013] In a possible design, the sleep-aiding execution module further includes a body temperature optimization module, the body temperature optimization module includes a heating assembly, and the body temperature optimization module is used for collecting an ambient temperature, and dynamically controlling the heating assembly to start heating based on a physiological signal and the ambient temperature.

[0014] In a possible design, a display module is further included, the display module is in communication connection with the control module, and the display module is used for visually displaying the visual sleep staging model.

[0015] The present application has the following advantages: The application discloses a sleep-aiding pet robot, which comprises a pet main body shell, a sensor module, a control module and a sleep-aiding execution module are arranged in the pet main body shell, the control module is in communication connection with the sensor module and the sleep-aiding execution module, the sensor module collects multi-modal real-time interaction signals of a user, uploads the multi-modal real-time interaction signals to the control module, the control module pre-processes the multi-modal interaction signals to obtain pre-processed multi-modal real-time interaction signals, processes the pre-processed multi-modal real-time interaction signals based on a signal processing model to obtain a real-time interaction signal group, matches the real-time interaction signal group based on an AI algorithm and a preset sleep-aiding rule to obtain corresponding bionic interaction instructions; meanwhile, sleep monitoring signals in a preset time are collected, a sleep automatic staging algorithm based on an attention mechanism is used to process the sleep monitoring signals to obtain sleep staging results, the sleep staging results comprise monitoring signal groups in different sleep periods, a visual sleep staging model is constructed based on the sleep staging results, and the sleep-aiding execution module controls a breathing module, a brain wave tuning module, a light module, an environmental noise module and / or a music module to execute according to the corresponding bionic interaction instructions. The sleep-aiding pet robot can guarantee corresponding bionic interaction by issuing bionic interaction instructions based on an AI algorithm combined with a preset sleep-aiding rule, can realize automatic division of sleep stages of a user by using a sleep automatic staging algorithm based on an attention mechanism, can generate a visual sleep staging model, and is convenient for selection of intervention timing and understanding of a sleep state of the user, and is convenient for application and promotion. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A module diagram of the sleep-aiding pet robot provided for the embodiment 1. DETAILED DESCRIPTION

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the present application will be briefly introduced below in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the drawings is only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings. It should be noted that the description of these embodiments is used to help understand the present application, but does not constitute a limitation on the present application.

[0018] It should be understood that the term "and / or" that may appear in this document is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B may indicate three situations: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" that may appear in this document describes another type of association object relationship, indicating that two relationships may exist. For example, A / and B may indicate two situations: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this document generally indicates that the previous and subsequent associated objects are in an "or" relationship.

[0019] Example 1: like Figure 1 As shown, this embodiment provides a sleep-aiding pet robot, comprising a pet main body shell, wherein a sensor module, a control module, and a sleep-aiding execution module are provided in the pet main body shell, wherein the control module is communicatively connected with the sensor module and the sleep-aiding execution module respectively; The sensor module is used to collect the user's multimodal real-time interaction signals, which include action signals, sound signals and physiological signals, and upload the multimodal real-time interaction signals to the control module; The control module is used to preprocess the multimodal real-time interaction signal to obtain a preprocessed multimodal real-time interaction signal, process the preprocessed multimodal real-time interaction signal based on the signal processing model to obtain a real-time interaction signal group, match the real-time interaction signal group based on the AI ​​algorithm and preset sleep-aiding rules to obtain a corresponding bionic interaction instruction, and send the corresponding bionic interaction instruction to the sleep-aiding execution module; The control module is further configured to collect sleep monitoring signals within a preset time, perform stage processing on the sleep monitoring signals using an automatic sleep staging algorithm based on an attention mechanism, obtain sleep staging results, wherein the sleep staging results include groups of monitoring signals for different sleep periods, and construct a visual sleep staging model based on the sleep staging results; The sleep-aiding execution module is used to control the execution of the breathing module, the brainwave tuning module, the lighting module, the ambient noise module and / or the music module according to the corresponding bionic interaction instructions.

[0020] Based on the above disclosure, the embodiment cooperates the sensor module, the control module and the sleep-aiding execution module, collects signals through the sensor module, generates bionic interaction instructions according to the collected signals through the control module, controls the respiratory module, the brain wave tuning module, the light module, the environmental noise module and / or the music module to execute the bionic interaction function according to the bionic interaction instructions through the sleep-aiding execution module, brings interactive sleep-aiding to the user, improves the experience of the user, and at the same time, adopts the sleep automatic staging algorithm based on the attention mechanism to perform staging processing on different sleep stages of the user, which is used to measure the sleep quality of the user, and at the same time, constructs a visual sleep staging model according to the sleep staging result, which can directly reflect the change trend of the sleep quality of the user after using the sleep-aiding pet robot, and is convenient for the user to view.

[0021] Embodiment 2: The embodiment provides a sleep-aiding pet robot, which comprises a pet main body shell, a sensor module, a control module and a sleep-aiding execution module are arranged in the pet main body shell, the control module is in communication connection with the sensor module and the sleep-aiding execution module respectively; The sensor module is used for collecting multi-modal real-time interaction signals of a user, the multi-modal real-time interaction signals comprise action signals, sound signals and physiological signals, and the multi-modal real-time interaction signals are uploaded to the control module; The control module is used for pre-processing the multi-modal real-time interaction signals to obtain pre-processed multi-modal real-time interaction signals, processing the pre-processed multi-modal real-time interaction signals based on a signal processing model to obtain a real-time interaction signal group, matching the real-time interaction signal group based on an AI algorithm and a preset sleep-aiding rule to obtain corresponding bionic interaction instructions, and sending the corresponding bionic interaction instructions to the sleep-aiding execution module; The control module is also used for collecting sleep monitoring signals in a preset time, performing staging processing on the sleep monitoring signals by adopting a sleep automatic staging algorithm based on an attention mechanism to obtain sleep staging results, the sleep staging results comprise monitoring signal groups in different sleep periods, and constructing a visual sleep staging model based on the sleep staging results; The sleep-aiding execution module is used for controlling the respiratory module, the brain wave tuning module, the light module, the environmental noise module and / or the music module to execute according to the corresponding bionic interaction instructions.

[0022] Preferably, the control module is also used for constructing a multi-modal emotional state recognition model, recognizing the monitoring signal groups in different sleep periods based on the multi-modal emotional state recognition model to obtain emotional states in different sleep states, and generating corresponding respiratory control instructions according to the emotional states.

[0023] Specifically, the multi-modal emotional state recognition model is constructed based on a deep learning model. After the multi-modal emotional state recognition model is constructed, historical monitoring signal groups of different sleep periods and emotional state label values corresponding to different sleep states are obtained, the historical monitoring signal groups of different sleep periods and the emotional state label values corresponding to different sleep states are taken as a training data set, the training data set is input into the multi-modal emotional state recognition model for training to obtain a prediction value, parameters of the multi-modal emotional state recognition model are updated based on the prediction value, the training data set and a loss function, the above updating step is repeated until a preset iteration termination condition is reached, a trained multi-modal emotional state recognition model is obtained, different sleep states of different sleep periods are recognized based on the trained multi-modal emotional state recognition model, and emotional states in different sleep states are obtained. The preset iteration termination condition can be, for example, that the number of updates reaches a preset threshold.

[0024] Preferably, the sleep-aiding execution module is in communication connection with a breathing module, a brain wave tuning module, a light module, an environmental noise module and a music module respectively. The breathing module includes an air bag or a flexible actuator, and is used to simulate human respiratory frequency and respiratory action. The brain wave tuning module is used to release α / θ wave bands to the user. The light module is used to turn on and adjust the brightness of the light. The environmental noise module includes a directional speaker and an environmental microphone. The directional speaker is used to play sound that can offset interference noise or to play white noise. The environmental microphone is used to collect interference noise in the environment.

[0025] The music module is used to play soothing sleep-aiding music or preset audio.

[0026] Specifically, the breathing module simulates human respiratory frequency and respiratory action through an air bag or a flexible actuator, and guides the user to enter a deeper relaxation state through ups and downs synchronized with human respiratory frequency. The brain wave tuning module releases α / θ wave bands to the user, which can guide the human body to enter a deep relaxation state, reduce external interference, help the user fall asleep quickly, prevent insomnia and improve sleep quality. At the same time, it can also alleviate tense emotions and help the user fall asleep quickly. The light module realizes the simulation of starry sky or slowly changing soft halo through the integration of ultra-low brightness, blue light-free flexible OLED or micro LED array. It can also simulate the sunrise effect through gradually brightening light, and naturally wake up the user. The environmental noise module plays sound that can offset interference noise or white noise, such as rain sound or campfire sound, and can adjust the volume or type according to biological feedback to help the user fall asleep.

[0027] Preferably, the physiological signal comprises real-time body surface temperature data and carbon dioxide concentration value; the sensor module comprises an accelerometer, an audio sensor and a miniature non-contact physiological sensor; The accelerometer is configured to collect the motion signal of the user. The audio sensor is configured to collect the sound signal of the user. The miniature non-contact physiological sensor comprises an infrared sensor and a carbon dioxide sensor, the infrared sensor is configured to collect the real-time body surface temperature data of the user, and the carbon dioxide sensor is configured to collect the carbon dioxide concentration value in the space.

[0028] The motion signal of the user can reflect the body activity of the user during sleep, such as the turning-over action of the user, etc.; the carbon dioxide concentration value in the space can be used to evaluate the air quality in the space, when the carbon dioxide concentration value is too high, it will cause the occurrence of hypoxia, dizziness or drowsiness, etc. discomfort symptoms, affecting the sleep quality of the user.

[0029] Preferably, the multi-modal real-time interaction signal is pre-processed to obtain the pre-processed multi-modal real-time interaction signal, comprising: The sound signal is pre-emphasized to obtain the emphasized sound signal. The motion signal, the emphasized sound signal, the real-time body surface temperature data and the carbon dioxide concentration value are subjected to outlier processing and missing value filling to obtain the filled motion signal, sound signal, real-time body surface temperature data and carbon dioxide concentration value. The filled motion signal, sound signal, real-time body surface temperature data and carbon dioxide concentration value are subjected to denoising processing to obtain the denoised motion signal, sound signal, real-time body surface temperature data and carbon dioxide concentration value. The denoised motion signal, sound signal, real-time body surface temperature data and carbon dioxide concentration value are subjected to normalization processing to obtain the pre-processed multi-modal real-time interaction signal.

[0030] Specifically, the multi-modal real-time interaction signal is subjected to outlier processing, missing value filling and denoising processing to avoid the influence of outlier, missing value and noise value on subsequent signal processing, the denoising processing can select wavelet transform and filtering algorithm, the denoised multi-modal real-time interaction signal is subjected to normalization processing to eliminate the influence of different dimensions, so that the data is in a similar numerical range, facilitating subsequent model processing and data fusion.

[0031] In one possible design, the real-time interaction signal group comprises motion signal features, sound signal features and physiological signal features; the signal processing model is configured to process the pre-processed multi-modal real-time interaction signal to obtain the real-time interaction signal group, comprising: extract the action signal features based on the signal processing model, the action signal features including action time domain features and action frequency domain features; extract the voiceprint features based on the signal processing model, the voice signal features including the voiceprint features; extract the physiological signal features based on the signal processing model, the physiological signal features including temperature frequency domain features and carbon dioxide frequency domain features.

[0032] In this embodiment, the action signal features, the voice signal features, and the physiological signal features are obtained by performing feature processing on the preprocessed multi-modal real-time interaction signals, and the action signal, the voice signal, and the physiological signal are amplified for subsequent matching operations.

[0033] In a possible design, a sleep automatic staging algorithm based on an attention mechanism is used to perform staging processing on the sleep monitoring signals to obtain sleep staging results, including: a CNN (Convolutional Neural Network) model is used to extract features from the sleep monitoring signals to obtain a plurality of monitoring signal features, and the plurality of monitoring signal features are sequentially processed to obtain a monitoring signal feature sequence; data fusion is performed on the monitoring signal feature sequence to obtain a fused signal feature sequence; a self-attention mechanism is used to capture the dependency relationship of the fused signal feature sequence to obtain a processed fused signal feature sequence; the sleep automatic staging algorithm based on the attention mechanism is used to perform staging processing on the processed fused signal feature sequence to obtain the sleep staging results.

[0034] In this embodiment, before data fusion is performed on the monitoring signal feature sequence, a strategy enhancement model based on multi-task learning is used to learn the representation of a specific modality, so that the feature data can be better fused when data fusion is performed on the monitoring signal feature sequence. In this embodiment, a Transformer model is used to perform data fusion on the monitoring signal feature sequence.

[0035] In a possible design, the real-time interaction signal group is matched based on an AI algorithm and a preset sleep-aiding rule to obtain corresponding bionic interaction instructions, including: The real-time interaction signal group is matched with the preset sleep-aiding rule to determine whether the triggering condition of the preset sleep-aiding rule is met; If the triggering condition is met, the real-time interaction signal group and the preset sleep-aiding rule are processed using the AI algorithm to obtain the corresponding bionic interaction instructions.

[0036] Specifically, the preset sleep-aiding rule can be exemplified as follows: when the signals reflect high heart rate, rapid breathing and frequent movement, it indicates that the user state is anxious or nervous, at this time, the generation of playing soothing music, starting slow breathing module and other bionic interaction instructions can be generated; when the signals reflect stable heart rate, slow and deep breathing and slight or no movement, it indicates that the user state is relaxed or in sleep, at this time, the generation of reducing light brightness to dim light, playing white noise and other bionic interaction instructions can be generated; when the signals reflect snoring sound or slight decrease in physiological state, it indicates that the user state may be breathing difficulty, at this time, the generation of playing prompt sound guiding lateral recumbency and other bionic interaction instructions can be generated.

[0037] Further, the AI algorithm can be selected as a classification model or a pattern matching model, such as SVM (Support Vector Machine) or LSTM (Long Short-Term Memory), etc. The above algorithms are prior art and will not be described in detail here.

[0038] Preferably, in the embodiment, the AI built-in learns the user's interaction habits, preferences and sleep patterns to form a personalized "digital personality" and enhance the sense of companionship of the real pet.

[0039] In a possible design, the sleep-aiding execution module further includes a body temperature optimization module, the body temperature optimization module includes a heating assembly, and the body temperature optimization module is configured to collect an ambient temperature and dynamically control the heating assembly to start heating based on the physiological signals and the ambient temperature.

[0040] Specifically, the heating assembly dynamically adjusts its temperature according to the physiological signals and the ambient temperature to achieve the most comfortable temperature for the user.

[0041] In a possible design, a display module is further included, which is in communication connection with the control module, and the display module is configured to visually display the visual sleep staging model.

[0042] In the embodiment, the visual sleep staging model is visually displayed, which facilitates the user to understand his / her own sleep condition, and a sleep report is generated according to the visual sleep staging model, which is stored in the cloud and can be viewed and called by the user at any time. The sleep report includes the sleep-in time, the number of night awakenings, the number of interactions with the robot and the effectiveness of the interactions.

[0043] In a possible design, a micro aromatherapy diffuser is further arranged inside the pet body shell, which can release a small amount of sleep-aiding aroma at a specific sleep stage according to the user's preference or preset program, and can be cooperatively adjusted with the body temperature optimization module to increase the environmental comfort.

[0044] In a possible design, the outer layer of the pet body shell is a detachable protective layer, and the protective layer is composed of a flexible material, which includes but is not limited to cotton, plush, air cushion and silica gel, so as to bring a comfortable tactile feeling to the user.

[0045] Finally, it should be noted that: the above only for the preferred embodiments of the present application, and not for limiting the scope of protection of the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A sleep-aiding pet robot, characterized in that: The pet main body shell includes a sensor module, a control module and a sleep-aiding execution module, wherein the control module is respectively connected to the sensor module and the sleep-aiding execution module: The sensor module is used to collect the user's multimodal real-time interaction signals, which include action signals, sound signals and physiological signals, and upload the multimodal real-time interaction signals to the control module; The control module is used to preprocess the multimodal real-time interaction signal to obtain a preprocessed multimodal real-time interaction signal, process the preprocessed multimodal real-time interaction signal based on the signal processing model to obtain a real-time interaction signal group, match the real-time interaction signal group based on the AI ​​algorithm and preset sleep-aiding rules to obtain a corresponding bionic interaction instruction, and send the corresponding bionic interaction instruction to the sleep-aiding execution module; The control module is further configured to collect sleep monitoring signals within a preset time, perform stage processing on the sleep monitoring signals using an automatic sleep staging algorithm based on an attention mechanism, obtain sleep staging results, wherein the sleep staging results include groups of monitoring signals for different sleep periods, and construct a visual sleep staging model based on the sleep staging results; The sleep-aiding execution module is used to control the execution of the breathing module, the brainwave tuning module, the lighting module, the ambient noise module and / or the music module according to the corresponding bionic interaction instructions.

2. The sleep-aiding pet robot according to claim 1, characterized in that: The control module is also used to construct a multimodal emotional state recognition model, identify monitoring signal groups of different sleep periods based on the multimodal emotional state recognition model, obtain emotional states under different sleep states, and generate corresponding breathing control instructions according to the emotional states.

3. The sleep-aiding pet robot according to claim 1, characterized in that: The sleep-aiding execution module is respectively connected to the breathing module, the brainwave tuning module, the lighting module, the ambient noise module and the music module; The breathing module includes an air bag or a flexible actuator, and is used to simulate the breathing frequency and breathing action of the human body; The brainwave tuning module is used to release ɑ / θ wave bands to the user; The lighting module is used to turn on and adjust the brightness of the light; The environmental noise module includes a directional speaker and an environmental microphone. The directional speaker is used to play sounds that can offset interference noise or play white noise. The environmental microphone is used to collect interference noise in the environment. The music module is used to play soothing sleep-inducing music or preset audio.

4. The sleep-aiding pet robot according to claim 1, characterized in that: The physiological signal includes real-time body surface temperature data and carbon dioxide concentration value; the sensor module includes an accelerometer, an audio sensor and a miniature non-contact physiological sensor; The accelerometer is used to collect the user's motion signal; The audio sensor is used to collect the user's voice signal; The miniature non-contact physiological sensor includes an infrared sensor and a carbon dioxide sensor. The infrared sensor is used to collect real-time body surface temperature data of the user, and the carbon dioxide sensor is used to collect carbon dioxide concentration values ​​in a space.

5. The sleep-aiding pet robot according to claim 4, characterized in that: Preprocessing the multimodal real-time interaction signal to obtain a preprocessed multimodal real-time interaction signal includes: Performing pre-emphasis processing on the sound signal to obtain an emphasized sound signal; Performing outlier processing and missing value filling on the motion signal, the aggravated sound signal, the real-time body surface temperature data, and the carbon dioxide concentration value to obtain the filled motion signal, the sound signal, the real-time body surface temperature data, and the carbon dioxide concentration value; De-noising the filled motion signal, sound signal, real-time body surface temperature data, and carbon dioxide concentration value to obtain de-noised motion signal, sound signal, real-time body surface temperature data, and carbon dioxide concentration value; The denoised motion signals, sound signals, real-time body surface temperature data and carbon dioxide concentration values ​​are normalized to obtain the preprocessed multimodal real-time interaction signals.

6. The sleep-aiding pet robot according to claim 1, characterized in that: The real-time interaction signal group includes motion signal features, sound signal features and physiological signal features; The pre-processed multimodal real-time interaction signal is processed based on the signal processing model to obtain a real-time interaction signal group, including: Extracting time domain and frequency domain features of the preprocessed motion signal based on the signal processing model to obtain motion signal features, wherein the motion signal features include motion time domain features and motion frequency domain features; Extracting voiceprint features from the preprocessed sound signal based on the signal processing model to obtain sound signal features, wherein the sound signal features include voiceprint features; Frequency domain features are extracted from the preprocessed physiological signal features based on the signal processing model to obtain physiological signal features, wherein the physiological signal features include temperature frequency domain features and carbon dioxide frequency domain features.

7. The sleep-aiding pet robot according to claim 1, characterized in that: The sleep automatic staging algorithm based on the attention mechanism is used to process the sleep monitoring signals into stages to obtain the sleep staging results, including: A CNN model is used to extract features of sleep monitoring signals to obtain multiple monitoring signal features, and multiple monitoring signal features are serialized to obtain a monitoring signal feature sequence; Performing data fusion on the monitoring signal feature sequence to obtain a fused signal feature sequence; The self-attention mechanism is used to capture the dependency of the fusion signal feature sequence to obtain the processed fusion signal feature sequence; The automatic sleep staging algorithm based on the attention mechanism stages the processed fusion signal feature sequence to obtain the sleep staging results.

8. The sleep-aiding pet robot according to claim 1, characterized in that: Based on AI algorithms and preset sleep-aiding rules, the real-time interaction signal group is matched to obtain corresponding bionic interaction instructions, including: Match the real-time interactive signal group with the preset sleep-aiding rules to determine whether the triggering conditions of the preset sleep-aiding rules are met; If it is achieved, the AI ​​algorithm is used to process the real-time interaction signal group and the preset sleep-aiding rules to obtain the corresponding bionic interaction instructions.

9. The sleep-aiding pet robot according to claim 1, characterized in that: The sleep aid execution module also includes a body temperature optimization module, which includes a heating component. The body temperature optimization module is used to collect ambient temperature and dynamically control the heating component to start heating based on physiological signals and ambient temperature.

10. The sleep-aiding pet robot according to claim 1, characterized in that: It also includes a display module, which is communicatively connected to the control module and is used to visually display the visualized sleep staging model.

Images