TWM595870U - Care system - Google Patents

Abstract

A care system is provided. The care system is suitably applied to a person under care. The system determines the person’s physiological condition through peripheral sensors. The system determines the health condition of the person under care by monitoring his sleep disorder. The system uses an audio receiver to receive sound about the person under care and obtains the sound of the person after eliminating background audio. The system uses an image sensor to capture the person’s images for motion detection, and obtains posture images of the person. In an aspect, an artificial intelligence technology is introduced to the system for establishing a care-warning prediction model when learning various data including data generated by physiological and environmental sensors. The care-warning prediction model is used to determine if condition of the person under care reaches a warning threshold.

Description

Care system

This book is about a care system, especially a care system that monitors the data of the care recipient's sleep or specific state to achieve early warning and reminder effects.

The operation mode of the conventional medical care system is mainly to install a physiological sensor on or around the care recipient, and connect a care center through the network, so that the care center can always obtain the physiological information transmitted by the terminal's physiological sensor, and Feasibility can be judged to get the physiological state of the care recipient.

Whether the conventional medical care system can indeed achieve the purpose of caring for the care-received depends on the immediate research and judgment ability of the care center and the physiological data provided by the physiological sensor on the care-receiver's side. However, it is often worn by the care-receiver The physiological sensors on the body can only provide limited physiological information, and to independently interpret various physiological data. In addition, accurate physiological state judgment should also be based on the past medical records of the care recipient and current environmental factors, and cooperate with the collected The physiological data performs a comprehensive judgment, and the current technology has not been effectively implemented.

As for the overall physiological data, professional physicians need to judge based on experience, but they still cannot take into account all the possibilities. Therefore, they cannot effectively issue warnings or reminders to specific situations, so that the purpose of home care is discounted. Especially for the purpose of home care, in general, the physiological sensor of home care cannot be compared with the equipment of medical institutions, the physiological data obtained is limited, and it cannot be accurately interpreted for specific conditions, such as the sleep quality of the care recipient, This is one of the important items that affects home care, not only affects the health status of the care recipient, but also pays attention to the situation of the care recipient at any time, and also affects the quality of life of the care recipient.

The exposure book discloses a care system, which includes a machine learning method, which is suitable for care recipients who need care in a home or medical institution. The system passes the physiological sensors, image sensors and audio around the care recipient Sensors and other devices achieve the purpose of home care. One of the main purposes is to enable caregivers to perform more efficient care through the care procedures generated by the machine learning mechanism.

According to the main architecture embodiment of the care system, it includes a data processing unit and multiple sensors. The multiple sensors include a first group of sensors. This operates at full time and generates a sense of care for the care recipient The measured data and the second set of sensors can be generated according to an instruction of the data processing unit to generate the sensed data about the care recipient. When the data processing unit judges that the first set of sensor data generated by the first set of sensors reaches the first threshold, the second set of sensors is activated, and the second set of sensors generates the second set of sensed data . Further, when the data processing unit receives the second set of sensing data, it can determine whether the condition for issuing the warning message is met according to the second threshold.

According to an embodiment, the main components of the care system include a data processing unit, and various sensor components coupled to the data processing unit, such as an audio receiver, for receiving the voice of the care receiver, after deducting the background sound signal, Obtain the audio from the cared person; the image sensor is used to obtain the cared person's image and perform motion detection from it to obtain the cared person's posture image.

Further, the care system may further include one or more physiological sensors to sense the physiological state of the care recipient and generate physiological data. In this way, the physiological data can be used to determine whether to generate warning information.

Further, the care system further includes a machine learning unit, which obtains audio and posture images from the care receiver according to the data processing unit, and compares the clinical data to establish a care warning prediction model. When establishing a care warning prediction model, the sensor data of the environment sensor and the information received by the system through the user interface can be incorporated into various factors for the machine learning unit to establish the care warning prediction model.

In order to further understand the features and technical content of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are for reference and explanation only, and are not intended to limit the present invention.

The following is a description of the implementation of the present invention through specific specific examples. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments. Various details in this specification can also be based on different viewpoints and applications, and various modifications and changes can be made without departing from the concept of the present invention. In addition, the drawings of the present invention are merely schematic illustrations, and are not drawn according to actual dimensions, and are declared in advance. The following embodiments will further describe the related technical content of the present invention in detail, but the disclosed content is not intended to limit the protection scope of the present invention.

It should be understood that although terms such as “first”, “second”, and “third” may be used herein to describe various elements or signals, these elements or signals should not be limited by these terms. These terms are mainly used to distinguish one component from another component, or one signal from another signal. In addition, the term "or" as used herein may include any combination of any one or more of the associated listed items, depending on the actual situation.

In view of the fact that the current care system relies on manpower (caregivers) for care and may need to provide 24-hour care services under certain circumstances, the demand for all manpower in medical hospitals places a great burden on home caregivers, even if there are various Sensor technology assists in care, but because there is no care warning mechanism provided according to the personal physiological and environmental conditions of the care recipient, it still needs human constant attention to prevent the system from accurately warning real emergencies, or generating false alarms to allow care Frequently resolve error messages. As such, this disclosure discloses a care system implemented with multiple sensor technologies and a care method implemented by the system. Machine learning technology is also introduced to establish a care warning prediction model to provide automated and personalized care warning services One of the main technical purposes is to relieve the burden of traditional caregivers and avoid the problems caused by traditional manpower or limited sensing technology.

For an embodiment of the care system, refer to the schematic diagram of the operation scenario shown in FIG. 1.

The figure shows a cared person 10 lying on the bed, at home or in a specific care center, at home or in a specific center does not have the same level of monitoring equipment as the hospital, so the care system senses the cared person 10 Physiological data is mainly used to provide preliminary care functions. However, further, the use of artificial intelligence technology and machine learning methods can still derive the ability to judge the physical condition of the care recipient from limited data. In this example, the care receiver 10 is connected to a physiological sensor, such as a wristband 101, and the physiological data such as body temperature, pulse, respiration, blood pressure, heart rate and other data can be obtained, and the corresponding physiological data display 103 can display the data In general, there are warning thresholds for various data to perform the function of preliminary care. Furthermore, the system can be equipped with a blood oxygen detector 102, which is connected to the care receiver 10, such as a finger, to obtain blood oxygen data, and can be monitored on the oximeter 104.

In addition, the care system may also include various other physiological sensing devices. For example, the chest strap can obtain the chest relief of the care receiver 10, the frequency of the chest motion, and the depth difference of the chest frame, etc., as a basis for judging the physiological state of the care receiver 10.

In the schematic diagram of this embodiment, an image sensor is provided beside the bed, such as the camera 105 in the figure, which can take images of the care receiver 10 at all times and perform motion detection during the shooting process. When performing motion detection, you can first create a background image from continuous images, including motionless objects, such as beds, wardrobes, fixed equipment, etc., and then obtain the changed images after comparing the front and back images. The resulting change can determine the posture image of the care receiver 10, and these posture images will be able to derive the behavior of the care receiver 10 and determine the status of the care receiver 10. For example, the posture image can be used to determine whether the care receiver 10 has turned over during sleep, violent hand and foot changes in a short period of time, falling off the bed, getting up, etc., can provide a warning threshold for posture changes, and is generated when there are abnormal changes Warning information.

Furthermore, an audio receiver may be provided beside the bed, such as the microphone 107 in the figure, for receiving the sound of the care receiver 10, especially the sound related to breathing. Among them, when using the microphone 107 to receive ambient sound, you can first create background audio, the background audio is the audio generated around the fixed, such as equipment operation sound, cooling fan, air-conditioning sound, etc. Once the background audio is established, the microphone 107 is recorded Can be deducted from the background audio, and the sound made by the care receiver 10 can be determined.

For example, the care system can be used to detect the sleep quality of the cared person 10, and use the microphone 107 to record the sound emitted by the cared person 10 when sleeping to determine the situation requiring immediate care, such as choking, phlegm, breathing sound, etc. , You can also judge the situation of breathing pause based on continuous sound. In this way, the audio becomes very important information for judging the care recipient 10, especially the physiological conditions related to sleep.

However, the physiological data of the care receiver obtained by using various sensors described in the above embodiments can cooperate with the feedback information generated by the caregiver or the caregiver to actively trigger, and the key data within the critical time can be obtained through the mechanism of the system annotation, different The big data needed to build the model in traditional artificial intelligence. The data generated by this type of care system is personalized data. A care warning prediction model can be established for individual situations, which can provide personalized care warning services and can effectively relieve caregivers. The burden.

It is mentioned here that the automatic care method applied in the above-mentioned care situation can apply various physiological sensors that can sense the care receiver 10, for example, the pulse and the pulse of the care receiver 10 can be sensed through the sensing bracelet 102 The reflected heart rate status and the blood oxygen condition sensed by the blood oxygen detector 102 allow the correlation of data between different sensing data to determine the condition of the care recipient 10 individually or comprehensively, so that the care recipient The person 10 is under more complete care.

FIG. 2 then shows a schematic diagram of an embodiment of each functional module in the care system, wherein the functional module can be implemented by software, or software and hardware devices.

The main components of the care system are various sensing devices around the person being cared for, including a data processing unit 201 that processes data generated by the sensing devices, such as a central processing unit of a computer system, which may be equipped with powerful Data processing capabilities, in addition to storing the data that has been generated, can also effectively process the data generated in real time, and can execute machine learning algorithms, perform big data analysis, and establish a model for predicting the physical condition of the care recipient.

According to the embodiment of the care system, a variety of sensors that are electrically connected to the data processing unit 201 may be included, including various physiological sensors 202, audio receivers 203, image sensors 204, and environmental sensors 205, and It may include a machine learning unit 207 coupled to the data processing unit 201, and may be connected to the user device 210 through the user interface 209 and receive information input by the care recipient or the care provider.

The care system can also connect to the external system 21 or the medical cloud 23 through the network 20 through a specific communication protocol. In addition to providing the data of the care recipient, it can also obtain various clinical data, big data analysis of individuals or groups Results, etc., are used as data for the implementation of artificial intelligence in the care system.

In this care system, the data processing unit 201, such as the central processor of the system, integrates the data generated by each peripheral device, after processing, performs warning judgment, and communicates with external systems 21 (such as medical institutions, care centers, etc.), medical cloud 23, etc. The connection also includes the use of artificial intelligence to build a care warning prediction model to predict the physical condition of the care recipient.

The audio receiver 203 is used to receive the voice of the cared person. As described in the above embodiment, the audio sent by the cared person can be obtained by subtracting the background audio signal. Therefore, in the automatic care method, the system can be based on the audio receiver 203 Obtained audio After obtaining the audio from the cared person, and then comparing the sound samples (including volume and/or audio), it can be determined whether the cared person needs care, such as (but not limited to) apnea condition, or Produce choking sounds, or phlegm sounds.

The image sensor 204 is used to obtain the image of the care receiver, and to perform motion detection therefrom. According to the method described in the above embodiment, the posture image of the care receiver is determined at any time during the full-time shooting, and the motion detection is performed. According to an embodiment, the motion detection result provided by the image sensor 204 can be used as a starting basis for whether to perform audio reception and judgment. For example, when judging from the video, the person being cared for goes to sleep, that is, taking full-time photography and judging the posture changes during sleep. When it is judged that the value of the care receiver's image change exceeds a threshold within this time, it may indicate sleep disturbances, including respiratory distress, or the need for intervention by the caregiver or no intervention, that is, the audio receiver 203 is activated to receive the received News from caregivers. After that, the system can judge the condition of the cared person according to the audio and judge whether a warning threshold is reached. If the warning threshold is reached, it means that the audio shows that the cared person is in a dangerous and embarrassing state, such as severe apnea and breathlessness. Phenomenon and other situations, it generates warning information.

In this way, the care system can perform home or special care tasks based on the audio receiver 203 and the image sensor 204 provided in the care recipient. Furthermore, the care system may also include one or more physiological sensors 202 for sensing the physiological state of the care recipient and generating physiological data, such as: body temperature, pulse, respiration, blood pressure, heart rate, etc., during actual operation The care system may not be limited to these physiological sensors and the data obtained. In the automatic care method performed by the data processing unit 201, after receiving these physiological data, it is possible to individually or comprehensively determine whether or not to generate warning information.

The environmental sensor 205 may be a temperature and humidity sensor, an air quality sensor, a barometric pressure sensor, etc. These environmental data may be used to modify the threshold for judging the physiological condition of the care recipient, because both the care recipient and the instrument have May be affected by these environmental factors, for example, the care recipient is in an environment with poor air quality or the indoor temperature and humidity are abnormal, which is prone to physical discomfort.

In an embodiment, the care system may further include a machine learning unit 207 coupled to the data processing unit 201. The machine learning unit 207 can obtain audio and posture images from the care recipient according to the data processing unit 201, and compare clinical data, A big data analysis method is used to establish a care warning prediction model.

Furthermore, in the process of implementing the care warning prediction model through the machine learning unit 207, the applied data may also include the environmental data of the care receiver sensed by the environmental sensor 205 and the physiological data generated by various physiological sensors 202, It becomes one of the factors for the machine learning unit 207 to establish a care warning prediction model. When performing machine learning according to the machine learning unit 207, the care system receives the care recipient or care provider through the user interface 209, and uses the warning information generated by the user device 210, which is provided for the care recipient, care provider, or others. Measures that the system has not judged to give instructions have also become one of the factors for the machine learning unit 207 to establish a care warning prediction model.

For example, after the care system connects to the user device 210 through the user interface 209, it can receive the information input by the care recipient or the care recipient. That is to say, when the system performs big data analysis through the machine learning unit 207 to establish a care warning prediction model, the actual physiological feelings of the care receiver can be input into the care system through the user interface 209 by operating the user device 210, and become feedback Information from the machine learning unit 207; similarly, caregivers can also generate feedback information based on their own judgment and input it into the care system through the user interface 210 through the user interface 209, so that the system can obtain data other than the data generated by the sensor The information can be used to calibrate the care warning prediction model established by the system using machine learning.

Moreover, the care system can also connect to the external system 21 or the medical cloud 23 through the network 20 through a specific communication protocol. In addition to providing the data of the care recipient, it can also obtain various clinical data, individual or group data Big data analysis results, etc., are used as data for implementing artificial intelligence in care systems.

According to the embodiment of the care system, various image, sound, physiological and environmental sensors are installed on or around the care recipient, such as the physiological sensor (202) mentioned in the embodiment of FIG. 2 and the audio receiver ( 203), an image sensor (204) and various environmental sensors (205), when performing the automatic care method, one or two of these sensors can be in operation at any time, such as a physiological sensor (may Called the first group of sensors), which can be processed by the data processing unit (201, Figure 2) in the care system and judge that there is physiological data that exceeds the threshold, a warning message is generated, that is, other sleep modes (power saving mode) are activated Sensors (can be called the second group of sensors), so that the system can receive a variety of sensing signals at the moment, according to the data collected before and after, including information that the caregiver is actually in an emergency state that is expected to be abnormal, and Including unexpected but cared-for information about an emergency, it becomes very important data for machine learning.

In one embodiment, among the multiple sensors in the care system, two or more sensors (the first group of sensors) may be used for full-time operation. When the signals generated by multiple sensors are combined to determine that the warning conditions are met, in addition to the warning information, other sensors that are not fully operational (second group of sensors) are activated through the system, so that the care system can get More comprehensive information for care recipients.

For the purpose of machine learning, the care system can obtain the conditions for the first group of sensors to generate warning information according to the model derived from the machine learning algorithm modeling, and the machine learning unit (207, Figure 2) obtains the second group of sensors In this way, the data of the device can be updated by the caregiver, the feedback generated by the caregiver, or a condition can be established to determine whether the caregiver is in an emergency state, and a prediction model can be established. The prediction model can be used to predict the physical condition of the care recipient based on limited or complete sensing data.

FIG. 3 shows a flowchart of an embodiment of an automatic care method in a care system. The care system is provided with a plurality of sensors, which are electrically connected to the data processing unit in the system, including the first set of sensors. Generate the sensing data about the cared person, and the second group of sensors, after starting according to the instruction of the data processing unit, generate the sensing data about the cared person.

When the care system is in operation, the first set of sensors (which may include one or more sensors) can be operated at full time, and the first set of sensing data about the care recipient can be continuously generated (step S301 ), through the execution of the data processing unit, according to the first threshold, determine whether the first set of sensing data generated by the first set of sensors reaches this threshold to determine whether the warning conditions are met? (Step S303).

In one embodiment, in addition to the system automatically determining whether the situation meets the threshold for generating an alert in this process, the user device (210) as shown in FIG. 2 can be used to input external commands through the user interface (209) (step S315 ), to assist in judging whether the current situation should meet or not meet the warning conditions, so that the original judgment can be corrected to allow the system to execute or not to continue the following steps. Under this measure, the lack of system operation can be made up, and the quality of care can be increased, which can be reflected in the subsequent machine learning.

When the sensing data has not yet reached the warning condition (No), step S301 is still repeated, and the first group of sensors continues to operate. It is mentioned here that at this time, the caregiver (or specific person) can still intervene in this care process, that is, even if the sensing data does not meet the warning conditions established by the system, if the caregiver finds that there may still be a demand for warning , The user device (210) shown in FIG. 2 can set this stage alert through the user interface (209), so that the process can continue, as in step S305, the second group of sensors is activated. The data generated by the caregiver's intervention is still important information for subsequent analysis and machine learning.

If the first set of sensed data reaches the warning condition (Yes), the data processing unit generates an instruction to activate the second set of sensors (which may include one or more sensors), and generates a second set of senses after sensing Measured data (step S305). At this time, the second set of sensing data is received by the data processing unit to determine whether the condition for issuing the warning message is met. If the second set of sensing data also reaches the second threshold set by another system, the warning message is issued. In a specific embodiment, since the activation of the second set of sensors indicates that the system judges that it is at a critical moment, the first set of sensing data and the second set of senses can be noted and retained for a period of time when the second set of sensors is activated The measured data is fed back to the machine learning unit (see Figure 2, 207).

Moreover, it can also issue warnings based on the judgment of the prediction model. In this example, the care system can store the data generated by the first set of sensors generated around this time according to the time when the above warning information is generated, and also store the data generated by the second set of sensors activated at this time Sensing data (step S307), and input the stored sensing data to the prediction model established by the machine learning algorithm (step S309).

According to yet another embodiment, inputting the sensing data to the prediction model may also add external data (step S317), such as the external data such as the external system (21) or health cloud (23) shown in FIG. External data can make the entire judgment basis more complete.

The prediction model is based on the data of medical institutions, health clouds, etc., plus the personal historical data and clinical data of the care recipient, and the model established by the characteristic correlation is obtained, and the parameters can be optimized according to the actual situation feedback , So that the limited sensing data can be effectively used. For example, the data of the first group of sensors and the second group of sensors can be used by the care receiver to predict the physiological condition of the care receiver to determine whether to issue a warning message? (Step S311).

Through the prediction model, when it is judged that it is not an emergency (No), the step still returns to S301. At this time, the first group of sensors can continue to operate and generate continuous sensing data, and the second group of sensors can be There is no need to work full time and return to power saving mode. Conversely, when it is determined that a warning needs to be issued, in step S313, the system will generate a warning message.

Similarly, in step S311, even if it is judged that there is no emergency state, the system still provides the caregiver to set an alert through the user device (210) shown in FIG. 2 through the user interface (209), including annotation and retention of data for a period of time. Give back to the system.

In another embodiment, except that the second group of sensors is activated only after judging that the warning condition is met, the first group and the second group of sensors can operate in the background at all times and continue to collect data when based on When the data generated by the first group of sensors judges that the warning conditions are met, the system will actively mark and store the sensing data of the first and second groups of sensors for a period of time (such as 1 minute before and after). Time data should be more critical data, which can be used for subsequent data analysis, including the purpose of providing machine analysis.

In yet another embodiment, the image sensor (204, FIG. 2) in the care system can also be used as the first sensor for full-time operation, which is generated when there is any abnormal image that exceeds the motion threshold Warning, and start the operation of other sensors, the embodiment can refer to FIG. 4.

FIG. 4 shows a flowchart of an embodiment of using an image sensor to drive a surrounding device.

When the care system uses an image sensor and an audio receiver as the sensing device for the care recipient, the image sensor, as the first group of sensors referred to in the above embodiments, can be used for full-time shooting ( Step S401), the images of the cared person can be obtained every time, but it is not necessary to store all the images until the specific conditions of the system are met. For example, in full-time shooting, motion detection can be performed at the same time (step S403), and the changes of the front and back images can be obtained from the consecutive images, and the threshold set by the system can be compared (as mentioned in the above embodiment. Threshold), when the change value is greater than this threshold, the image recording starts (step S405). This action can save space for storing images.

When an image change is detected (there is a certain degree of change after the comparison threshold), the degree and mode of image change can be determined (step S407). At this time, the care system will issue a warning by comparing the degree and mode of front-to-back change Conditions, if the degree of change and the method reach the conditions for generating an alert, that is, in step S409, alert information is generated, and at the same time, in one embodiment, the system will actively mark an image for a critical period of time (for example, 3 minutes before and after) , Time and related data. For example, from the image motion detection, the posture image of the cared person can be obtained, and the movement method, such as turning over, falling, and dramatic changes in the body, etc., can trigger the warning, and the warning can be triggered. The data at a critical moment before and after will be retained for other purposes. For example, this part of the critical data will become the data for subsequent analysis and machine learning.

In this example, when the warning message is generated, that is, when the value of the care receiver’s image change exceeds the first threshold set by the system, the audio receiver is activated to receive the audio from the care receiver. As shown in step S411, the audio receiver Then the second group of sensors as mentioned in the above embodiment starts the audio receiver such as a microphone and starts recording breathing audio. Once the background audio is deducted, the cleaner voice of the cared person can be obtained, as in the steps S413: After obtaining the sensor data and storing it in the storage device of the system, perform further judgment. Similarly, the audio recorded continuously according to the warning information will also be annotated and retained, and become the reference data for subsequent analysis and machine learning.

According to the automatic care method performed by the care system, the condition of the care recipient can be judged from the audio, especially for the problem of sleep disturbance. Among them, the sound samples stored by the system can be compared with the sound samples stored by the system, including the frequency and Volume samples, or personalized sound samples can be created from past data, or sound samples generated from group data, to determine whether the care recipient needs immediate care, such as breathing stops, choking sounds, or sputum sounds Waiting for embarrassment. When judging whether the warning threshold is reached (such as the second threshold in the above embodiment), if the warning threshold has been reached, such as judging that the apnea time is too long (beyond a certain time threshold), choking or phlegm sounds indicate a dangerous situation, etc., A warning message is generated.

In particular, the relevant first threshold and second threshold can be set by the system, and the threshold for generating warnings can also be set by the caregiver according to the actual situation.

Further, the care system can also cooperate with the physiological data generated by the physiological state of the care-received by one or more physiological sensors (which can be listed as the second group of sensors) to further determine whether the warning is generated information. Similarly, various physiological and environmental data generated during a critical moment set by the system will be annotated and retained for subsequent analysis and machine learning purposes.

Furthermore, the care system also adopts a machine learning method, which can establish a care warning prediction model based on the audio and posture images sent by the care recipient, and comparing the clinical data.

It is worth mentioning that, as in the automatic care method described in the above embodiment, the audio receiver can also be included in the first set of sensors to receive the care receiver's audio at all times, when the characteristics of the sound (frequency, frequency, When the volume) reaches the threshold, the image sensor listed as the second group of sensors is activated.

In another embodiment, the job of recording the audio of the cared person can be started at all times. Similarly, the system can set warning conditions for effective audio after excluding background audio. When a warning condition occurs, except for keeping critical time In addition to the data, the image recording process is further started to record, annotate and retain the images of the caregiver generated at this critical moment, which can also become data for subsequent analysis and machine learning.

A related embodiment is a flowchart of an embodiment of establishing a prediction model by a machine learning method in the automatic care method shown in FIG. 5.

In this step, initially as in step S501, the system collects the respiratory audio, posture image, physiological data and environmental data of the care receiver through various sensors. During the data processing process, as in step S503, it is necessary to establish a background Data to get clean data. In step S505, the system also sets the warning conditions according to the personal conditions of the care recipient, which are all the work before the care operation in specific occasions such as homes and medical institutions.

In addition to continuously obtaining the respiratory audio, posture image, physiological data and environmental data of the care recipient, and then obtaining the actual records in step S507, these may be from the clinical data provided by the medical institution or health cloud, or may be the care recipient The actual personal event data generated by the feedback mechanism in the system. As such, in step S509, the machine learning method is started to perform big data analysis. Generally speaking, after the big data analysis, a care warning prediction model for predicting the physiological condition of the care recipient can be established. The way to build a care warning prediction model is mainly based on the data collected from the sensors around the care receiver, and also the actual data provided by the health cloud, etc., and the information that the care receiver and its caregiver actively generate to the system. Use software algorithms to learn the characteristics of the data and establish the correlation between the data.

In step S511, after a big data analysis and algorithm, a care warning prediction model is established. In certain applications, the model can determine the care recipient based on the audio, posture image and/or physiological data emitted by the care recipient during sleep Whether it is a sleep disorder such as respiratory distress, or a state in which caregiver intervention is required or no intervention is required, so that verification can be performed repeatedly based on the facts (step S513), and the parameters of the care warning prediction model can be adjusted through actual data.

For details, refer to the schematic diagram of an embodiment of the care system shown in FIG. 6 using artificial intelligence technology to establish a care warning prediction model.

The figure shows an artificial intelligence module 60 implemented with software and hardware calculus in the care system, which includes a specific machine learning method 601, which is trained according to various sensing data provided by the system and can identify the data between the data. Relevance, the goal of prediction can be achieved.

When running the machine learning method 601, first obtain the data, such as the database 603 in the figure, and obtain the image data 61 generated by various sensors, the image includes the posture of the care recipient, for example, the posture such as turning, hand and foot movements; Physiological data can be obtained such as thoracic fluctuations, frequency, depth, etc. After training, it can establish posture recognition ability, and can establish the correlation between the posture image of the caregiver and other data; audio data 62, such as the general rest and sleep of the caregiver The sound emitted can identify the audio generated by specific physiological reactions, such as sputum sound, choking sound, and the sound produced by the phenomenon of apnea, etc., which can establish the correlation between various recognized audio information and other data; physiological data 63, such as body temperature, heart rate, respiration, pulse, blood oxygen and other physiological data of the care recipient, in addition to judging the normal and abnormal conditions of the physiological data, it is necessary to establish correlation with other data at any time; and environmental data 64, Sensing the environmental data of the care recipient through various environmental sensors (which can be listed as the first group of sensors or the second group of sensors) has become one of the factors for the machine learning method to establish the care warning prediction model 605, The data can include temperature, humidity, air quality, climate and other information of the caregiver's daily life, and can also be related to other data through learning. Then analyze the data to find the overall correlation between the data.

In particular, when the artificial intelligence used in the care system proposed in the Revealed Book uses the personalized data such as the image data 61, audio data 62, physiological data 63, and environmental data 64, a personalized care warning prediction model can be established.

After that, the machine learning method 601 cooperates with the historical and real-time data collected by the database 603, and the actual recorded physiological response, and can derive personalized rules, and then establish a care warning prediction model 605 for predicting the physical condition of the care recipient, even if Before any dangerous event occurs, it may be possible to predict what may happen based on various data to achieve the purpose of prevention and warning in advance.

Moreover, the artificial intelligence module 60 can also be trained and learned from data provided by specific sources through the machine learning method 601, for example, from the medical cloud 611, various groups of de-identified clinical data can be obtained, in addition to providing the data required for training In addition, it can also be used to verify the care warning prediction model 605.

The artificial intelligence module 60 receives the feedback information formed by the care recipient, the caregiver or other means through the feedback system 612. As described in the above embodiment, the care system has a user interface that can be used to receive the care recipient or other people The warning information generated by the user device becomes one of the factors for the machine learning method 601 to establish the care warning prediction model 605. Such data will be used to verify the prediction model and be used to adjust the parameters of the care warning prediction model 605 in the artificial intelligence module 60. .

In this way, the machine learning method 601 provided in the artificial intelligence module 60 integrates various data obtained by the data processing unit of the integrated care system, and can set different weights for different data according to the needs of the machine learning method, plus personalization The data, as well as the data from the medical cloud 611 and the feedback system 612, perform big data analysis and establish a care warning prediction model 605, which is used to determine whether the state of the care recipient reaches a specific warning threshold, which is also to meet the conditions for issuing warning information.

It is worth mentioning that the medical cloud 611 can also obtain terminal data from the care system. In this way, the calculation technology in the medical cloud 611 can obtain information with de-identified annotations, including time information and identity information, and collect it at the same time. The related physiological data and environmental data, which have established annotation data, are obviously related information for the medical cloud 611 and the artificial intelligence module 60 of the care system.

These annotated physiological data and various data are one of the effective data that need to be learned in artificial intelligence technology, so that the constructed model can identify and predict physiological information, and the machine learning and verification continue to be performed during the operation process to construct a more complete的医疗云611.

In various implementations of the care system, in particular, audio can be used to determine the physiological condition of the care receiver, as shown in FIG. 7 is a flowchart of an embodiment of the automatic care method.

At the beginning, as in step S701, the audio receiver around the care recipient is received through the audio receiver in the care system. As in step S703, when implemented, background audio can be created in the initial state to form a background sample for the system to be introduced during actual operation After that, you can get clean audio.

In step S705, through the software method, the data processing unit in the care system compares the obtained clean audio with the pre-established sound samples, and after comparing the sound characteristics (such as frequency, volume characteristics, etc.), as in step S707, Identify the voices of the cared person, such as choking sounds, phlegm sounds, audio changes, etc. These audios can be used to determine the physical condition of the cared person, such as step S709, to determine whether there is a situation requiring immediate care, one of which That is, to determine whether there is a condition of respiratory distress (such as apnea, choking), then in step S711, the threshold provided by the system is compared to determine whether the warning condition is reached, and subsequent medical measures can be executed accordingly.

In FIG. 8, a flowchart of an embodiment of applying a care warning prediction model to predict an abnormal condition of a care recipient in an automatic care method is described.

In step S801, the care system obtains real-time recorded audio, image, and sensing data through various sensors. In step S803, the system can first perform preliminary data processing and screening, that is, introduce data into the care warning prediction model. The care warning prediction model judges the current physiological condition of the care recipient according to the correlation between the before and after data and various data, and performs prediction. In step S805, the system determines whether a warning message is generated. If the warning condition has been reached, step S807 is performed to issue the warning message; otherwise, the flow returns to step S801.

It is worth mentioning that in the step of issuing warning information, feedback can be provided on the relevant data, the actual status information can be obtained through the feedback system 80, and the machine learning unit 82 can then optimize the care warning prediction model.

In summary, the care system and the automatic care method described in the above embodiments are applicable to the care environment where people in need of care are located. In this automatic care method, in addition to using various types of sensing in the care system The data generated by the device (such as various contact or non-contact sensors) performs an integration of various data to determine the physiological condition, and can be trained and learned from various data through machine learning to obtain the correlation between the data , Especially the use of personalized physiological and environmental data, and then establish a personalized care warning prediction model for predicting the physical condition of the care recipient, to achieve the purpose of caring at home or on specific occasions, through automatic care methods, or introducing artificial intelligence Technology can help the care provider to effectively care for the task of the care recipient.

The content disclosed above is only a preferred and feasible embodiment of this creation, and does not limit the scope of the patent application for this creation, so any equivalent technical changes made by using this creation specification and graphic content are included in this creation application Within the scope of the patent.

10: Caregiver 101: Sensing bracelet 103: physiological data display 102: Blood oxygen detector 104: oximeter 105: camera 107: microphone 202: Physiological sensor 203: Audio receiver 204: Image sensor 205: Environmental sensor 201: data processing unit 207: Machine Learning Unit 209: User interface 210: user device 20: Internet 21: External system 23: Medical Cloud 60: Artificial intelligence module 601: Machine learning 603: Database 605: Care warning prediction model 61: Image data 62: Audio data 63: physiological data 64: Environmental data 611: Medical Cloud 612: Feedback system 80: Feedback system 82: Machine Learning Unit Steps S301～S317: flow of automatic care method Steps S401-S413: the process flow of an embodiment using an image sensor driving device Steps S501-S513: an embodiment process of establishing a prediction model through a machine learning method Steps S701-S711: an embodiment process of judging the state of the care recipient by audio Steps S801-S807: an embodiment process of applying a care warning prediction model to perform prediction

Figure 1 shows a schematic diagram of the operation situation of the care system;

Figure 2 shows a schematic diagram of an embodiment of each functional module in the care system;

3 shows a flowchart of an embodiment of an automatic care method in a care system;

4 shows a flowchart of an embodiment of using an image sensor to drive a surrounding device;

FIG. 5 shows a flowchart of an embodiment of establishing a prediction model through a machine learning method in an automatic care method;

6 shows a schematic diagram of an embodiment of a care system adopting artificial intelligence technology to establish a care warning prediction model;

7 shows a flowchart of an embodiment of judging the state of a care recipient by audio in the automatic care method; and

FIG. 8 shows a flowchart of an embodiment of applying a care warning prediction model to predict an abnormal condition of a care recipient in an automatic care method.

202: Physiological sensor

203: Audio receiver

204: Image sensor

205: Environmental sensor

201: data processing unit

207: Machine Learning Unit

209: User interface

210: user device

20: Internet

21: External system

23: Medical Cloud

Claims (10)

A care system, including: A data processing unit; A plurality of sensors are electrically connected to the data processing unit, including a first group of sensors, which operate at full time and generate sensing data about a care recipient, and a second group of sensors, according to After an instruction of the data processing unit is activated, it generates sensory data about the cared person; and A machine learning unit executes a machine learning method, is coupled to the data processing unit, obtains the first set of sensed data and the second set of sensed data from the care receiver according to the data processing unit, and creates a care alert Prediction model Wherein, when the data processing unit determines that a first threshold is reached according to the first set of sensing data generated by the first set of sensors, the command is generated and the second set of sensors is activated, the second set of sensors The sensor generates a second set of sensing data; when the data processing unit receives the second set of sensing data, it determines whether the condition for issuing the warning message is met according to a second threshold; and when the second set of sensors is activated , The annotation and retention of data for a period of time to feed back to the machine learning unit. The care system according to claim 1, wherein the machine learning method determines whether the care recipient is in respiratory distress or needs to be based on the audio, posture image and/or physiological data emitted by the care recipient during sleep A state in which caregivers intervene or do not need to intervene. The care system of claim 2, wherein the plurality of sensors include: An audio receiver is electrically connected to the data processing unit for receiving the voice of a cared person, and after deducting the background audio signal, the audio sent by the cared person is obtained. The care system of claim 3, wherein the plurality of sensors further includes: An image sensor, electrically connected to the data processing unit, is used to obtain an image of the care receiver, and perform motion detection therefrom to obtain a posture image of the care receiver. The care system according to claim 4, wherein the image sensor is the first group of sensors, and the audio receiver is the second group of sensors, and the subject is photographed through the image sensor at all times The caregiver's posture image and perform a motion detection; and, when it is determined that the caregiver's image change value exceeds the first threshold, the audio receiver is activated to receive the audio from the caregiver, based on the audio judgment The condition of the cared person determines whether the second threshold is reached. If the second threshold is reached, the warning message is generated. The care system of claim 5, wherein the second set of sensors further includes one or more physiological sensors, electrically connected to the data processing unit, for sensing the physiological state of the care recipient, and Generating physiological data; in the automatic care method executed by the data processing unit, after receiving the physiological data, it is judged whether the warning information is generated. The care system according to claim 5, wherein in the automatic care method, audio samples from the care receiver are compared with sound samples, including volume and/or audio, to determine whether the care receiver needs immediate care Happening. The care system as described in any one of claims 1 to 7, further comprising a user interface for receiving warning information generated by the care receiver or a caregiver through a user device, which is established by the machine learning unit One of the factors of this care warning prediction model. The care system of claim 8, wherein the plurality of sensors further include an environment sensor for sensing the environment data of the care receiver, which becomes a factor for the machine learning unit to establish the care warning prediction model one. The care system according to claim 9, wherein in the automatic care method, the machine learning unit integrates various data acquired by the care system through the data processing unit, and the care-reward prediction model is used to determine the care receiver Whether the status of the system meets the conditions for issuing warning information.

Images