KR20200095668A - Elderly care system and method using AI speakers - Google Patents

Abstract

Disclosed are an elderly care system and method using an AI speaker. The elderly care system of the present invention includes a doll device having a microphone collecting a voice, a speaker outputting the voice, and a transmission unit transmitting the collected voice data to a management server. The management server may include a preprocessing unit normalizing by excluding non-natural language elements from the received voice data, a voice analysis unit analyzing the voice data preprocessed by the preprocessing unit based on big data, and a message transmission unit transmitting a message corresponding to the analyzed result to the doll device.

Description

Elderly care system and method using AI speakers

The present invention relates to an elderly care system and method using an AI speaker that analyzes the voice of a user in need of care to determine the presence or absence of a disease or to be used and managed as medical information.

The silver industry and similar care industry are expected to continue to develop. Research has been conducted to lower the incidence rate and alleviate symptoms through pre-diagnosis and prevention by using it as medical information to determine the progression of various diseases or dementia through the results of speech analysis of the elderly in need of care. In this regard, several studies have been conducted for a silver generation-friendly voice interface optimized for vocal behavior. According to domestic and international research on voice interface, Apple (SIRI) and Google (voice search) are building cloud-based infrastructure and evolving into personalized services. For example, Google released its own voice recognition service'Google Now' in June 2012 due to the influence of the Android operating system in the mobile market, and a new search engine algorithm optimized for voice search. 'Hummingbird' was also released. In Korea, a voice analysis method has been used to extract a voice variable and a non-parametric statistical method is applied thereto to classify the constitution through voice analysis. (See Korean Patent Publication No. 10-0926771). However, care-related technologies using voice analysis technology monitor the daily life of vulnerable groups who need social safety nets such as the elderly living alone in real time, and use the voice recognition function to directly communicate with users and provide emotional stability, or through voice analysis. Since it cannot determine a condition or provide a response method according to a health condition, there is a limit to use as a remote management and medical treatment or a routine health management method. Accordingly, using a voice recognition tool for users in need of care, the characteristics and changes of the voice recognition rate according to the user's utterance rate adjustment for each syllable are tracked, and diseases such as dementia appearing in voice recognition can be detected early, diagnosed, and responded to user care. A possible proposal is being demanded.

Patent Document 1. Korean Patent Publication No. 10-2015-0112901 (published on 2015.10.07)

Patent Document 2. Korean Patent Publication No. 10-1894722 (announcement date 2018.10.04)

One of the technical problems to be solved in the present invention is to identify the presence or absence of a disease by analyzing the voice of a user in need of care or to use it as medical information.

One of the technical problems to be solved in the present invention is to solve the problem of storing and managing voice data by a duplicate user rather than a designated user in analyzing the voice data.

The above objects, according to the present invention, a microphone for collecting voice; A speaker outputting sound; A doll device including a transmission unit for classifying the collected voice data and transmitting the classified voice data to the management server; And a preprocessor for normalizing by excluding non-natural languages from the received voice data. A voice analysis unit that analyzes the voice data preprocessed by the preprocessor based on big data; A message transmission unit for transmitting a message corresponding to the analyzed result to the doll device; including, can be achieved from the elderly care system using an AI speaker.

According to an embodiment of the present invention, the voice data collected through the doll device may be transmitted to a management server, and a relay unit configured to transmit a message transmitted from the management server to the doll device.

According to an embodiment of the present invention, the management server may include a disease determination unit that determines whether a user has a disease using an IR sensor image of the doll device and a collected voice data signal.

According to an embodiment of the present invention, the management server further comprises a disease determination unit configured to determine whether a disease has progressed by measuring a length between a syllable and a vowel of the IR sensor image of the doll device and the collected voice data signal. Can be.

The above objects are, according to the present invention, the steps of collecting, storing and managing voice data through a doll device including a microphone for collecting the user's voice and a speaker for outputting voice, and transmitting the voice data to a management server; Normalizing by excluding non-natural language elements among voice data signals received from the step through a preprocessor of the management server; A voice analysis step of analyzing the voice data preprocessed from the step based on big data; And a message transmission step of transmitting a message corresponding to the result analyzed from the step to the doll device. It may be achieved from the elderly care method using an AI speaker.

According to an embodiment of the present invention, the transmitting of the voice data includes: a relay step of transmitting the voice data collected from the doll device to a management server, and transmitting a message transmitted from the management server to the doll device; It can be configured to include.

According to an embodiment of the present invention, the voice analysis step may include a disease determination step of determining whether a user has a disease using an IR sensor image of the doll device and a collected voice data signal.

According to an embodiment of the present invention, the voice analysis step further includes a disease determination step of determining whether the disease progresses by measuring a length between a syllable and a vowel of the IR sensor image of the doll device and the collected voice data signal. Can be configured.

The present invention has the effect of being able to determine the presence or absence of a disease or obtain medical information by analyzing vocalization at all times by placing it in a residential space of the elderly in need of care.

The present invention has an effect of improving the reliability of an analysis result by solving the problem of analyzing, storing, and managing voice data by a duplicate user or an external sound signal other than a designated user.

The present invention has the effect of providing emotional stability while communicating with users of the vulnerable who need a social safety net, such as the elderly living alone.

The present invention has an effect of determining a health condition through voice analysis or providing a response method according to the health condition, and utilizing it for remote medical treatment or daily health management.

1 is an example of an elderly care system network using an AI speaker according to an embodiment of the present invention.
2 is an example of a doll device system constituting the elderly care system using an AI speaker according to an embodiment of the present invention.
3 is an example of a design type of a doll device constituting an elderly care system using an AI speaker according to an embodiment of the present invention.
4 is an example of a management server configuring an elderly care system using an AI speaker according to an embodiment of the present invention.
5 is an example of communication data processing between a doll device and a management server constituting an elderly care system using an AI speaker according to an embodiment of the present invention.
6 is an example of a flow chart related to controlling an elderly care system using an AI speaker according to an embodiment of the present invention.

Hereinafter, a'senior care system and method using an AI speaker' according to a preferred embodiment of the present invention will be described as follows.

Among the terms used in the present invention,'doll device' may be used interchangeably as'A1 speaker' or'smart speaker'. The'artificial intelligence speaker' may be a speaker to which the function of artificial intelligence (AI) is added. For example, it may be a speaker capable of providing a voice recognition function in addition to listening to music or radio.

In the present invention, in managing and diagnosing the health of the elderly in need of care, it was conceived that syllables, pronunciation, and voice speed patterns for each age group have unique points, and each peculiar point can be used as health information related to the presence or progress of diseases. It uses a voice recognition tool for the elderly who need care to track the characteristics and changes of the voice recognition rate according to the speech speed pattern of each syllable of the elderly, and detect and diagnose diseases such as dementia early, identified through voice recognition, or quickly. It is presented to be able to respond.

1 is an example of an elderly care system network using an AI speaker according to an embodiment of the present invention.

2 is an example of a doll device system constituting the elderly care system using an AI speaker according to an embodiment of the present invention.

3 is an example of a design type of a doll device constituting an elderly care system using an AI speaker according to an embodiment of the present invention.

4 is an example of a management server configuring an elderly care system using an AI speaker according to an embodiment of the present invention.

Elderly care system (A) using an AI speaker according to an embodiment of the present invention, as shown in Figures 1 to 4, wired and wireless communication network 100, communication for transmitting and receiving data with the wired and wireless communication network 100 It may be composed of a doll device 200 having a module 210 and a management server 300 connected through the communication module 310 to enable communication with the doll device 200.

In addition, voice data collected from the doll device 200 is transmitted to the management server 300, and the message transmitted from the management server 300 is transmitted and received by the doll device 200 as a diagnosis result sound or a message such as a push message. It may be configured to include a relay unit 400.

2 is an example of a doll device system constituting the elderly care system using an AI speaker according to an embodiment of the present invention.

As shown in FIG. 2, the doll device 200 includes a microphone 220 that collects voice uttered from a user and inputs it to the controller 250, a speaker 230 that receives and outputs voice from the controller 250, and a speaker A voice information processing unit 240 for normalizing the voice output from the voice information 230 and a transmission unit 270 for transmitting voice data processed from the voice information processing unit 240 to the management server 300 may be provided. . Here, the speech information processing unit 240 that normalizes and processes speech may include a classification processing unit 260 that separates and processes speech in natural and non-natural languages.

The doll device 200 may perform the function of a smart speaker (artificial intelligence speaker). For example, voice recognition and recognized voice information may be processed and transmitted to the management server 300.

The doll device 200 may collect more voice data through artificial intelligence and accumulate it as big data. The voice information processing unit 240 and the control unit 250 that may be embedded in the doll device 200 may learn the intention and pattern of conversation with artificial intelligence. This means that the more data you need for learning, the faster you can learn. The voice information processing unit 240 and the control unit 250 that may be mounted on the doll device 200 may collect and manage more data or learn and store voices through the database 280 cloud.

Through this, it is possible to have discernment to distinguish and determine whether the user of the doll device 200 is the intended original user, or whether it is a user other than the original user.

The main voice input/output devices of the doll device 200 based on voice recognition are a microphone 220 and a speaker 230. Since the doll device 200 is composed of only such main devices, it may be more efficient than other smart devices compared to other smart devices, for example, a smart phone or a TV in terms of collecting voice data. For example, in the case of other smart devices, voice data can be recorded and processed only after the user activates the voice function, but the doll device 200 is advantageous in that it can always record and accumulate voice data except for the use of the mute function. I can.

The voice recognition doll device 200 moves the existing touch type platform to a voice-based platform when processing voice data. At the same time, the voice recognition service that can be provided by the doll device 200 can have excellent command execution ability, and since it does not require manual manipulation, the operation method is simple, and in particular, users of the elderly people who are not free to move and have low touch manipulation power. It can be provided with an interface suitable for the field.

The voice recognition through the doll device 200 may classify and process the voice by age group based on voice data collected when the voice is recognized, and automatically classify and process the user by comparing it with the learned accumulated data. As described above, this classification process can discriminate between an intended designated user and a non-designated user who may be intervened unintended, thereby providing reliability in information processing and management of voice recognition related to the user.

The doll device 200 may include a filtering module that automatically classifies and processes voice data of non-designated users except for the voice of a designated user learned and stored in the voice data accumulation process to block errors in existing accumulated big data.

In the voice information processing by the doll device 200, for example, the currently collected voice data is accumulated and compared with the learned voice data, and the currently input voice data can be classified and processed by age group divided into adult-middle-aged-old, etc. have.

For example, it is possible to classify adults, middle-aged, and elderly people into groups, and set the stuttering and speed patterns as a comparison reference value parameter during conversation, and the voice data currently collected by the doll device 200 may be classified and determined by age group. .

According to a number of studies, adults 21-30 years old (average 25 years old), middle aged 1 45-54 years old (average 49 years old), middle aged 2 55-64 years old (average 60 years old), and aged 1 65-74 years old. When experimenting by dividing the categories into age (average 68 years old) and aged 2, 75-91 years old (average 80 years), young adult males speak faster than middle-aged and elderly adults, and middle-aged 2 adults are aged 2 It is derived as an average result that young adults, middle-aged, and old 1 speak faster than old 2.

Therefore, it is possible to automatically classify and process the current user age with the voice data collected by the doll device 200, and at the same time, it is possible to determine whether it is an originally designated user or an intervening non-designated user, and only the voice data of the designated user is accumulated Since it can be managed as big data, it can block errors in the analysis result of voice data due to non-designated user intervention, and provide reliability for voice information processing and management.

On the other hand, the voice data that is classified in the process of classifying the voice data collected through the doll device 200 into a natural language-non-natural language through the voice information processing unit 240 and the data learned as big data of the database 280 If matched, the voice data collected by the doll device 200 may be transmitted to the management server 300 through the transmission unit 270 and the communication module 210.

The voice data collected through the doll device 200 is classified and processed in natural language-non-natural language through the voice information processing unit 240, and the classified voice data does not match the data learned as big data of the database 280. If not, the control unit 250 may cut off its own power to end voice data collection.

Existing user big data accumulated by automatically classifying and processing all non-natural languages that may be intervened from outside, including the voice data of non-designated users, excluding the voice data of the designated user learned and memorized during the voice data accumulation process through a separate filtering module. Will block the error.

Here, the natural language may include pure'pronunciation','syllables', and'sound' produced by humans as all'language' accumulated or learned as big data. Unnatural languages are'intervals','sounds of grasping','sounds of bumps','sounds made by animals','sounds of operation of various machines and devices', and'sounds from nature' during speech during the pronunciation process. It may contain external sounds other than people including the back.

Table 1 below is an example of classification and classification of voice data that can be classified through the voice government processing unit 240 or a filtering module and stored and managed as big data or learned.

Acquisition data signal Classification field Classification field Processing result Plain language Natural language - Save Consonant vowel) Natural language - Save Syllable Natural language - Save Pronunciation Natural language - Save cough Natural language - Save breath Natural language - Save Animal (beast) sound - Unnatural language Discard Sound of water - Unnatural language Discard Mechanical sound - Unnatural language Discard The sound of the wind - Unnatural language Discard Cracking sound - Unnatural language Discard Forest sound - Unnatural language Discard

As shown in Table 1, all voice data collected through the doll device can be automatically classified as available voice data and unusable data and stored in a database or classified as discarded values.

In addition, the doll device 200 constituting the elderly care system (A) using an AI speaker according to an embodiment of the present invention receives and receives health management information including a disease-related message analyzed from the management server 300 A voice synthesizer 281 that synthesizes the generated message into voice and outputs it through the speaker 230 to directly sound information related to disease occurrence, a plurality of touch sensors 282, and a shock sensor 283 are provided. Thus, it can be configured to recognize the user's behavior and manage it as the user's singularity information.

In addition, an IR sensor 284 is embedded in the doll device 200 to sense a user's activity and synthesize a voice signal in response thereto through the controller 250 and output to the speaker 230.

A display unit (not shown) for receiving and displaying a message transmitted from the management server 300 and a power supply unit (not shown) for supplying power may be provided at the right place of the doll device 200.

3 is an example of a design type of a doll device constituting an elderly care system using an AI speaker according to an embodiment of the present invention.

Preferably, as the design of the doll device 200, as shown in FIG. 4, a design in the shape of an animal that can provide emotional stability and friendliness to a person may be selected. For example, a doll in which a dog, a bear, etc. is shaped into a design shape may be selected as a desirable external design of the doll device 200.

In this way, in order to configure the elderly care system using the AI speaker, an example of the cognitive and emotional evaluation results for the doll device 200 in which the shape of the dog as shown in FIG. 3 is shaped as a design is as follows.

The purpose of the test is to determine the effect of the doll device, a silver care robot, on cognitive function, depression, and ability to perform daily life.

As a test method, participants were divided into an experiment group and a control group with 17 subjects in the experimental group and 25 subjects in the control group according to the criteria, and conducted in the form of a doll device and a play with a built-in silver care service. The test period was conducted for 5 weeks, twice a week, approximately 30 minutes per session per elderly person.

Experimental tools include cognitive function-dementia simple screening test (K-MMSE), depression-short-form Geriatric Depression Scale (GDS), daily life performance ability-elderly skill assessment questionnaire. It was carried out by selection.

As a result of the test, the silver care doll device was evaluated to be helpful in slowing down the cognitive decline in the experimental group and improving the ability to perform daily life. There was no significant difference in depression score.

In this way, it can be seen that the use of the doll device for elderly care increases the psychological stability.

On the other hand, a representative example of disease diagnosis that can be recognized by the voice data of the doll device 200 through the big data voice analysis of the management server 300 is shown in Table 2 below.

Table 2 may be an example of disease diagnosis based on human language (natural language) excluding non-natural languages, but is not necessarily consistent and may be updated as a clinical trial.

Disease classification Symptom Content of the push message
Laryngeal cancer When speaking out loud, it makes a rough and squeezing sound Suspected disease
Lung cancer, thyroid cancer Noise mixes as if the wind is falling, regardless of the volume of the voice Suspected disease
Parkinson's disease The volume of the voice is reduced and the trembling is severe. The pronunciation is not accurate and the length between the syllables and vowels of the speech signal is lengthened. Suspected disease
Bipolar disorder Speaking quickly when in a good mood, slow when feeling depressed Suspected disease Hyperactivity disorder syndrome Speaking in short syllables Suspected disease Post-traumatic stress disorder Pronouncing certain notes for a long time Suspected disease

The puppet device 200 is how accurately the user pronounces at a location placed around the user, whether the speaking speed is not fast, how many intervals in the middle when speaking, how the pattern is, and how the voice is unstable. It can be classified and extracted into non-verbal elements such as whether or not the words are not obscured, as well as natural and non-natural language elements, which can be learned and managed as big data.

4 is an example of a management server configuring an elderly care system using an AI speaker according to an embodiment of the present invention.

The management server 300 includes a preprocessor 320 that normalizes voice data received through the communication module 310, a big data-based voice analysis unit 330 that analyzes the preprocessed voice data, and corresponds to the analyzed result. It may be configured to include a message transmission unit 350 for transmitting a message to the doll device 200.

In addition, the management server 300 may include a disease determination unit 340 that determines whether a user has a disease using the IR sensor image of the doll device 200 and the collected voice data signal.

In addition, the management server 300 measures the length between the syllables and vowels of the image of the IR sensor 284 received from the doll device 200 and the collected voice data signal, and the disease classification reference value of the database 360 It may be configured to include a disease determination unit 340 for diagnosing and determining whether the disease progresses or not.

5 is an example of communication data processing between a doll device and a management server constituting an elderly care system using an AI speaker according to an embodiment of the present invention.

A method of care for the elderly through the elderly care system using an AI speaker according to an embodiment of the present invention configured as described above will be described as follows.

Collecting and managing voice data through a doll device including a microphone for collecting user's voice and a speaker for outputting voice, and transmitting voice data to a management server; Normalizing by excluding non-natural language elements among voice data signals received from the step through a preprocessor of the management server; A voice analysis step of analyzing the voice data preprocessed from the step based on big data; And a message transmission step of transmitting a message corresponding to a result analyzed from the step to the doll device.

In addition, the transmitting of the voice data may include a relay step of transmitting the voice data received from the doll device to the management server, and transmitting the message transmitted from the management server to the doll device.

In addition, the voice analysis step may include a disease determination step of determining whether the user has a disease using the IR sensor image of the doll device and the collected voice data signal.

In addition, the voice analysis step may further include a disease determination step of determining whether the disease progresses by measuring a length between a syllable and a vowel of the IR sensor image of the doll device and the collected voice data signal.

6 is a representative example of a flow chart related to controlling an elderly care system using an AI speaker according to an embodiment of the present invention.

First, it is determined whether the designated user has applied power to drive the doll device (S110).

As a result of the determination in step S110, when it is determined that the power to drive the doll device is'ON', collection of external voice data is started through the microphone of the doll device (S120).

When the voice data collection starts in step S120, at the same time, the controller of the doll device determines whether or not the user is a designated user (S130).

As a result of the determination in step S130, if it is determined that the user is not the designated user, the initial step is returned.

As a result of the determination in step S130, if it is determined that the user is a designated user, the age of the user is continuously determined (S140). Subsequently, the collected voice data is classified into natural language, which is a voice produced by the user, and ambient noise excluding the user, and classification processing is performed on the voice data in a natural/unnatural language system (S150).

Among the voice data classified in step S150, only'natural language' excluding non-natural language is stored in the database, and big data management is started through learning and accumulation (S160).

The voice data managed in step S160 is transmitted to the management server, and the management server starts comparative evaluation or calculation determination based on big data of the voice data signal (S180).

It is determined whether a disease or a disease is suspected based on the big data determined in step S180 (S190).

When the determination result in step S190 is not determined as a disease or disease suspicion, the received voice data information processing results are accumulated as big data of the management server (S200).

As a result of the determination in step S190, when it is determined that a disease or a disease is suspected, the voice analysis routine for the voice data is terminated by sending a push message to the user's doll device and notifying it (S210).

Meanwhile, an application area of the elderly care system using an AI speaker according to an embodiment of the present invention and an example of its evaluation are as follows.

The voices of the elderly were recorded through the doll device 200, and the speech speed of the elderly was adjusted using a Synchronized Overlap-Add algorithm (SOLA), and the difference in recognition rate before and after the adjustment was confirmed. As a result, it was confirmed that the speech recognition rate increased by 14.5% for men and 10.1% for women after the speech rate was adjusted. In other words, the average recognition rate of elderly men was 78.6%, and the recognition rate improved by 93.1% after adjustment by about 14.5%, and voice recognition by elderly women was improved from 77.0% before adjustment to 87.1% after adjustment to 10.1%. .

It can be seen that the voice recognition rate of the young people is quite similar to the voice recognition rate of the young when adjusting the speech speed of each syllable in the elderly. In addition, in the case of the elderly, voice recognition failure rate of about 9.9% was confirmed after the speech rate was adjusted. Accordingly, through analysis of the cause of speech recognition failure, three causes were identified: initial weakness, silence problem, and pronunciation error.

Among them, pronunciation errors were identified as the most common errors. It was confirmed that it was successful before the ignition speed adjustment, but it did not appear in the error that failed after the adjustment, and it was a common error only when both before and after the adjustment failed. In the case of initial weakening, if the length and energy value of the syllable for the initial voice are secured, a more accurate speech recognition success rate can be expected. Finally, the silent problem mainly appeared in compound nouns (eg download management). After adding silence directly to the voice that lacks silence, 84% of the errors in the silence problem appear as the speech recognition success rate.

In the process of using the doll device 200 without burden through this test, the doll device 200 of the elderly by supplementing the voice interface function that corrects the speed and pattern of speech closer to the youth based on the voice data extracted by age group. A result that could attract interest in use was obtained, and the present invention is used to promote brain function activity that delays dementia or aging related to the rate of ignition, including diagnosis and prescription of diseases based on big data It was confirmed that it can also be used as.

The present invention has been described with reference to the embodiments shown in the drawings, but is not limited to the embodiments, and can be carried out with modifications and variations without departing from the gist of the present invention, and modifications and variations are made in the spirit of the present invention. Included.

100: wired/wireless communication network (network)
200: doll device 210: communication module
220: microphone 230: speaker
240: voice information processing unit 250: control unit
260: classification processing unit 270: transmission unit
280: database 281: speech synthesis unit
282: touch sensor 283: shock sensor
284: IR sensor 300: management server
310: communication module 320: preprocessor
330: voice analysis unit 340: disease judgment unit
350: message transmission unit 360: database

Claims (8)

A microphone for collecting voice; A speaker outputting sound; A doll device including a transmission unit for classifying the collected voice data and transmitting the classified voice data to the management server; And
The management server may include a preprocessor for normalizing by excluding non-natural languages from among received voice data; A voice analysis unit that analyzes the voice data preprocessed by the preprocessor based on big data; A message transmission unit for transmitting a message corresponding to the analyzed result to the doll device; including, elderly care system using an AI speaker. The method of claim 1,
A relay unit for transmitting the voice data collected through the doll device to the management server, and transmitting the message transmitted from the management server to the doll device; the elderly care system using an AI speaker further comprising. The method of claim 1,
The management server is a disease determination unit for determining whether the user is sick using the IR sensor image of the doll device and the collected voice data signal; and an aged care system using an AI speaker comprising a. The method of claim 1,
The management server further comprises a disease determination unit for determining whether the disease progresses by measuring the length between the syllables and vowels of the IR sensor image of the doll device and the collected voice data signal. Collecting and managing voice data through a doll device including a microphone for collecting user's voice and a speaker for outputting voice, and transmitting voice data to a management server;
Normalizing by excluding non-natural language elements among voice data signals received from the step through a preprocessor of the management server;
A voice analysis step of analyzing the voice data preprocessed from the step based on big data; And a message transmitting step of transmitting a message corresponding to the result analyzed from the step to the doll device. Including, the elderly care method using an AI speaker. The method of claim 4,
The transmitting of the voice data may include a relay step of transmitting the voice data collected from the doll device to a management server, and transmitting a message transmitted from the management server to the doll device; Way. The method of claim 4,
The voice analysis step includes a disease determination step of determining whether a user has a disease using an IR sensor image of the doll device and a collected voice data signal. The method of claim 4,
The voice analysis step further comprises: a disease determination step of determining whether the disease progresses by measuring the length between the syllables and vowels of the IR sensor image of the doll device and the collected voice data signal; .

Images