KR20200069264A - System for outputing User-Customizable voice and Driving Method thereof - Google Patents

Abstract

The present invention relates to a user-customized voice selectable voice output system and a method for driving the same, and by a user-customized voice selectable voice output system including: a cloud server including a voice learning part configured to collect and store user voice data from a user terminal, in which a user-customized voice output service-specific application is driven, and to, if the amount of stored voice data of a specific user is a reference value or more, analyze voice data for the specific user to learn characteristics of the voice data of the specific user, and a voice style file generating part configured to generate a voice style file defining voice characteristics of the specific user based on the contents learned by the voice learning part; a text-to-speech (TTS) module part configured to convert data in the form of a text received from a user terminal, on which the user-customized voice output service-specific application is mounted, into data in the form of a voice, and a voice output part configured to apply the data in the form of a voice converted by the TTS module part to the voice style file of the specific user received from the cloud server and output a voice of the specific user, a voice of a general person desired by the user can be learned, and the data in the form of a letter can be output with a voice of a person desired by the user.

Description

System for outputting user-customizable voice and driving method thereof

The present invention relates to a voice output device, and more particularly, to a voice output system capable of selecting a user-customized voice and a driving method thereof.

With the development of artificial intelligence, the development of speech recognition technology to provide more convenience is becoming active. Speech recognition technology is a computer that converts acoustic signals from a sound sensor, such as a microphone, into words or sentences.

For example, in the case of a neural network based method, which is one of the most commonly used voice generation methods in the speech conversion method, each model is created using features expressing a spectral shape, and the output value from each model Using them, we can create another model that maps each other back to perform the transformation.

At this time, as models reflecting the characteristics of each spectral shape, a restricted Boltzmann machine or the like can be used. In addition, an artificial neural network, a Bernoulli bidirectional associative memory, etc. may be used as a model for mapping output values from these models.

In the case of the existing neural network-based speech generation method, data is learned through various models, and thus there is an advantage that can reflect the non-linear characteristics of the data well. However, there is a limitation that the model must be trained through 1:1 mapping. That is, a large number of models are required when converting from one speaker to several speaker voices.

In particular, speech analysis is an essential and very important process for speech recognition or speech synthesis. The voice parameters are divided into a method of extracting in the time domain and a method of extracting in the frequency domain. In the time domain, energy, zero-crossing, pitch period, pitch frequency, linear prediction coefficient, LPC Cepstrum, and MFCC (Mel Frequency) Cepstral Coefficient is used, and in the frequency domain, a spectrum envelope and feature parameters are used.

Meanwhile, a technology for outputting a text message in a voice form from a smart device such as a smartphone has already been popularized. However, in this case, it is limited to the technology of outputting the artificial intelligence voice designated within the terminal, that is, a limited voice.

KR 10-2001-0026402 A KR 10-2004-0013071 A KR 10-2014-0015228 A

The present invention is derived from such a background, and a voice output system capable of selecting a user-defined voice for learning a voice of a general person desired by a user and outputting text data in a voice of a person (general person) desired by the user and its driving The purpose is to provide a method.

Accordingly, it is intended to provide a voice output system capable of selecting a user-customized voice capable of providing a more friendly and interesting TTS (Text to Speech) service and a driving method thereof.

*

The present invention for achieving the above object includes the following configuration.

That is, according to an embodiment of the present invention, a voice output system capable of selecting a user-customized voice collects and stores user voice data from a user terminal in which a user-specific voice output service-only application is driven, and a specific user's voice data storage amount exceeds a reference value. Next, a voice learning unit that analyzes the voice data of the specific user to learn the characteristics of the voice data of the specific user, and a voice style file defining a voice characteristic of the specific user based on the content learned by the voice learning unit A cloud server including a voice style file generating unit, and a text to speech (TTS) module unit that converts text type data received from a user terminal equipped with a dedicated application for user-defined voice output service into voice type data, and the And a smart speaker device including a voice output unit that applies the data in the form of voice converted by the TTS module unit to the voice style file of the specific user received from the cloud server and outputs the voice of the specific user.

In one aspect of the present invention, when a user-specific voice output service-only application generates the voice style file of the specific user in the voice style file generator, a voice selection button for the specific user is generated and output. do.

According to another aspect, the smart speaker device further includes a short-range wireless communication unit for performing short-range wireless communication with the user terminal, and the voice output unit receives text data from a user terminal through the short-range wireless communication unit. Is done.

On the other hand, the method of driving a voice output system capable of selecting a user-customized voice is that the cloud server collects and stores user voice data from a user terminal running a dedicated application for a user-defined voice output service, and if a specific user's voice data storage is greater than a reference value. Analyzing the voice data of the specific user to learn the characteristics of the voice data of the specific user, and generating a voice style file defining the voice characteristics of the specific user based on the learned content by the cloud server, smart A speaker device converts text type data received from a user terminal equipped with a dedicated application for user-specific voice output service into voice form data, and the smart speaker device converts the converted voice form data from the cloud server. And applying the received voice style file of the specific user to output the voice of the specific user.

According to the present invention, by providing a voice output system capable of selecting a user-specified voice and learning the voice of a general person desired by a user and outputting text data in the voice of a person desired by the user, and a driving method thereof, a voice desired by the user As it provides a more friendly TTS (Text to Speech) service, it also has the effect of increasing the degree of interest.

Accordingly, it is possible to read a book in the voice of a parent, teacher, or other nurturer, thereby reducing the rejection and intimacy of infants and toddlers, thereby increasing the stability of infants and toddlers and remarkably increasing the story immersion.

In addition, there is an effect of providing a smart speaker device that outputs various book contents or current affairs contents in a voice selected by a user, not only outputting a simple image or sound.

1 is an exemplary view showing the configuration of a voice output system capable of selecting a user-customized voice according to an aspect of the present invention;
Figure 2 is a block diagram showing the configuration of a cloud server according to an embodiment of the present invention,
Figure 3 is a block diagram showing the configuration of a smart speaker device according to an embodiment of the present invention,
4 is an exemplary view of a voice collection screen in a user-specific voice output service dedicated application according to an embodiment of the present invention;
5 is an exemplary view of a script transmission screen of a user-specific voice output service-only application according to an embodiment of the present invention;
6 is a flowchart illustrating a method of driving a voice output system capable of selecting a user-customized voice according to an embodiment of the present invention.

It should be noted that the technical terms used in the present invention are only used to describe specific embodiments, and are not intended to limit the present invention. In addition, technical terms used in the present invention should be interpreted as meanings generally understood by a person having ordinary knowledge in the technical field to which the present invention belongs, unless otherwise defined in the present invention. It should not be interpreted as a meaning or an excessively reduced meaning.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is an exemplary view showing a configuration of a voice output system capable of selecting a user-customized voice according to an aspect of the present invention.

As shown in FIG. 1, a voice output system according to an embodiment of the present invention includes a user terminal 10, a cloud server 20, and a smart speaker device 30.

The user terminal 10 is driven by a user-specific voice output service dedicated application installed.

In one embodiment, the user terminal 10 is a personal computer such as a laptop, desktop or laptop carried by the user, or a Personal Communication System (PCS), Global System for Mobile communications (GSM), PDC (Personal DigitalCellular), PHS(Personal Handyphone System), PDA(Personal Digital Assistant), IMT(International Mobile Telecommunication)-2000, CDMA(Code Division Multiple Access)-2000, W-CDMA(W-Code Division MultipleAccess), Wibro (Wireless Broadband Internet) terminal, a smart phone (smartphone), a smart pad (smartpad), a tablet PC (Tablet PC) and the like may be one of all types of handheld (Handheld)-based wireless communication device, but the present invention is limited to this It is not done.

2 is a block diagram showing the configuration of a cloud server according to an embodiment of the present invention.

2, the cloud server 20 includes a voice learning unit 220, a voice style file generation unit 230, a voice data storage unit 210, and a communication unit 240.

The cloud server 20 collects user voice data from the user terminal 10 through the communication unit 240 and stores it in the voice data storage unit 210.

The voice data storage unit 210 may include volatile or nonvolatile memory. The voice data storage unit 210 may store, for example, commands or data related to at least one other component of the cloud server 20. According to an embodiment, the voice data storage unit 210 may store software or a program.

Here, the program may include, for example, a kernel, middleware, an application programming interface (API), or an application program dedicated to a user-defined voice output service.

In one embodiment, the voice data storage unit 210 may be implemented as big data for storing voice data.

The voice learning unit 220 analyzes voice data of a specific user when the amount of voice data storage of a specific user stored in the voice data storage 210 is greater than or equal to a reference value.

Here, the reference value of the voice data storage amount may be counted based on the play time (time) of the voice data stored in an amount sufficient to grasp the voice data of a specific user, or may be counted based on the voice data storage capacity. . The voice data of a person stored in the voice data storage unit 210 may be analyzed when voice data of a certain time or more is stored or voice data of a certain capacity is stored.

Then, the voice data of a specific user is analyzed to learn its characteristics. In one embodiment, the voice learning unit 220 may be implemented as an application program that can analyze the wavelength, frequency, and color of user voice data. Alternatively, the voice learning unit 220 is interpreted to include any means for extracting and distinguishing the wavelength, frequency, and tone of the input voice, including a sensing device.

In addition, the voice learning unit 220 may collect the user's voice data while the user makes a voice call to the user terminal 10 to secure the user's voice data. At this time, it is preferable that the user is allowed to collect voice data during a call through a user-specific voice output service-only application.

In one aspect, the voice learning unit 220 performs deep learning-based voice learning. Deep learning (deep learning) and deep learning are high-level abstractions (abstracts, summarizing key content or functions in a large amount of data or complex data) through a combination of several nonlinear transformation methods. It is defined as a set of machine learning algorithms that attempt.

It can be said that in a large frame, it is a field of machine learning that teaches a person's way of thinking to computers. Accordingly, it is possible to define a voice characteristic of a specific user with less voice data than before.

The voice learning unit 220 may learn the voices of ordinary people, such as their families, including the owner of the user terminal 10, and also learn the voices of officials such as entertainers, announcers, politicians, and professors.

The voice style file generation unit 230 generates a voice style file defining a voice characteristic of a specific user based on the content learned by the voice learning unit 220. The voice style file contains information capable of restoring the user's voice based on the voice style file.

In one embodiment, a voice style file defining a voice characteristic of a specific user may be style.xml. Extensible Markup Language (XML) is a versatile markup language developed by W3C that is recommended for use in creating other special purpose markup languages. XML is a simplified subset of SGML, which can be used to describe many different kinds of data. XML mainly makes it easy to exchange data between different types of systems, especially those connected to the Internet.

In one aspect of the present invention, when a user-specific voice output service-only application generates a voice style file of a specific user in the voice style file generation unit 230, a voice selection button for a specific user is generated and output. do.

That is, the user can be provided as a selectable item through a user-specific voice output service-specific application for the voice generated by the voice style file in the cloud server 20.

In addition, according to an additional aspect of the present invention, the voice learning unit 220 may further reflect the user's emotional state information and intonation information in learning characteristics of the user's voice data. At this time, the voice style file includes the user's emotional state information and intonation information.

For example, even one person's voice can express different emotional states by adjusting the speed or size. In a joyful state or an excited state, the volume information can be set to speed up or the voice tone is set to high, and in a calm or sad state, the volume can be set to low, slow, and voice tone. However, it is not limited thereto. It is applicable by varying various elements that can express emotions.

The communication unit 240 is interpreted to cover the technical configuration of performing data transmission and reception with the user terminal 10 and the smart speaker device 30. In one embodiment, the communication unit 240 receives voice data from the user terminal 10 and provides voice information in which a voice style file is generated through a user-specific voice output service-only application.

Further, the communication unit 240 provides a specific user voice style file requested from the user terminal 10 to the smart speaker device 30.

3 is a block diagram showing the configuration of a smart speaker device according to an embodiment of the present invention.

As can be seen from FIG. 3. The smart speaker device 30 includes a communication unit 330 including a TTS module unit 310, a voice output unit 320, and a short-range wireless communication unit 335.

In one embodiment, the smart speaker device 30 may be functionally mounted on an existing Bluetooth speaker or implemented as a Linux OS-Linux PC. However, it is not limited thereto.

The TTS module unit 310 receives text data through a user-specific voice output service-only application from the user terminal 10 equipped with a user-specific voice output service-only application. Then, the received text data is converted into voice data.

The TTS technology builds a pronunciation database for all phonemes in a language and connects them to generate a continuous voice. At this time, the natural voice is synthesized by adjusting the size, length, and height of the voice. For this, natural language processing technology may be included.

The voice output unit 320 applies the data in the form of voice converted by the TTS module unit 310 to a voice style file of a specific user received from the cloud server 20 and outputs the voice in a specific user's voice.

The voice output unit 320 may output a voice signal to the outside. For example, the voice output unit 320 converts and outputs voice (or voice signal) data converted by the TTS module unit 310 into an audible sound.

The communication unit 330 is interpreted to cover both wired and wireless communication modules capable of transmitting and receiving data with the cloud server 20 and the user terminal 10.

In one embodiment, the communication unit 330 receives a voice style file from the cloud server 20 in a Wi-Fi, ie, wireless Internet method. However, it is not limited thereto.

According to an aspect, the communication unit 330 of the smart speaker device 30 according to an embodiment of the present invention includes a short-range wireless communication unit 335 to perform short-range wireless communication with the user terminal 10. In one embodiment, the short-range wireless communication unit 335 may be a Bluetooth module.

Bluetooth is a typical short-range communication method that enables voice and data communication between terminals with low cost and low power. In general, in a Bluetooth communication, a master device and a slave device use a piconet. It is formed.

The master device means a device that initiates Bluetooth communication by transmitting a Bluetooth signal, and the slave device means a device that receives the Bluetooth signal sent from the master device and performs communication with the master device.

In one embodiment, the user terminal 10 may operate as a master device, and the smart speaker device 30 may operate as a slave device.

A predetermined Bluetooth device broadcasts an inquiry signal by setting a frequency hopping sequence to search for Bluetooth devices located in the vicinity. Bluetooth devices that perform an inquiry scan transmit Bluetooth device address (BD_ADDR) and clock information to the Bluetooth device broadcasting the inquiry signal.

However, the short-range wireless communication unit 335 is not limited to a Bluetooth communication module, and is interpreted to cover various modifications.

The voice output unit 320 of the smart speaker device 30 according to an embodiment of the present invention receives data in the form of text from the user terminal 10 through the short-range wireless communication unit 335.

In addition, the smart speaker device 30 may receive a control signal for the overall operation through a user-specific voice output service dedicated application of the user terminal 10. For example, control signals such as power ON/OFF signals and volume control signals are received.

In addition, the user terminal 10 may receive voice selection information, text-type data script, and additional emotion information about the voice.

4 is an exemplary view of a voice collection screen in a user-specific voice output service-only application according to an embodiment of the present invention.

As illustrated in FIG. 4, voice data of a user may be collected by recording an example script. At this time, it is desirable to provide a sufficient amount of script or recording time to generate a user's voice style file.

It is not only possible to register one voice with one user terminal 10, it is also possible to collect multiple voices with one user terminal 10 by receiving user identification information before the voice collection screen of FIG.

5 is an exemplary view of a script transmission screen of an application for a user-specific voice output service according to an embodiment of the present invention.

As shown in FIG. 5, the script can be selected by selecting a file stored in the user terminal 10 or directly inputting the file.

In addition, the user may select a voice through a user-specific voice output service dedicated application according to an embodiment. Voice 1, Voice 2, Voice 3, and Voice 4 may be voices of other people, such as mom, dad, grandmother, and grandfather, respectively.

In one embodiment, a user-specific voice output service-only application further provides a list of categories for selecting voices.

For example, the voice of a celebrity such as a celebrity or an announcer, a politician, or a professor may be provided as a selection list, or the voice of the user who is the owner of the user terminal 10 or a public such as his family or friends may be provided as a selection list. It is possible.

At this time, the user-specific voice output service-only application is to generate a voice selection button for the voice generated by the voice style file generation unit 230 of the cloud server 20.

The user terminal 10 transmits the selected text type script file and voice identification information to the smart speaker device 30 in a short-range wireless communication method.

The smart speaker device 30 is a script in the form of a text in a short-range wireless communication method from the user terminal 10 when the start signal is input as the'send' button is selected through a user-specific voice output service dedicated application of the user terminal 10 To receive.

In addition, the received text-type script can be applied to the voice style file of the specific user selected by the user and read by the selected user's voice by utilizing the TTS function.

As an example, when the child goes to bed, the user may select'Mom Voice' through the user-specific voice output service dedicated application of the user terminal 10 and select'Calm' as the emotion information. In addition, you can select a child's favorite textbook file.

Then, the smart speaker device 30 receives a voice style file for the existing'mother voice' from the cloud server 20. Then, the user terminal 10 receives a file of a fairy tale book through a short-range wireless communication method. The smart speaker device 30 converts the text-type fairy tale book file received by the TTS function into voice data and applies it to the mother voice style file to output it in the mother voice.

At this time, even if the user selects the mother's voice in the same way, it is possible to apply the emotion information such as'calmly' to output the speed by slowing the speed, decreasing the volume or increasing the weight of the bass. Accordingly, it is possible to provide the voice of a parent or primary caregiver to the child from time to time, thereby providing a smart speaker device capable of increasing intimacy.

On the other hand, as another example, in the morning time, the user may select'announcer a voice' through the user-specific voice output service dedicated application of the user terminal 10 and'lively' as the emotion information. Additionally, you can select a newspaper file for the day.

Then, the smart speaker device 30 receives a voice style file for'announcer a voice' from the cloud server 20. Then, the newspaper file of the day is received from the user terminal 10 by wireless communication. The smart speaker device 30 converts the text file of the text received by the TTS function into voice data and outputs it in the voice of announcer a.

At this time, even if the user similarly selects the voice of announcer a, it is possible to apply emotion information such as'lively' to speed up the speed somewhat, increase the volume or increase the weight of the treble. Accordingly, it is possible to provide a smart speaker device that delivers a variety of text-type current affairs in a user-friendly voice.

As another example, when the owner of the user terminal 10 is a student, an'idol b voice' may be selected through a user-specific voice output service-only application of the user terminal 10, and a textbook summary that is a test content may be selected as a script. .

Then, the smart speaker device 30 receives a voice style file for the existing'idol b voice' from the cloud server 20. Then, the text file of the textbook summary is received from the user terminal 10 by wireless communication. At this time, the text file of the textbook arrangement may be manually written by the user.

The smart speaker device 30 converts the text file of the textbook organized contents received by the TTS function into voice data and outputs the voice of the idol b. Accordingly, it is possible to provide a smart speaker device capable of inducing learning interest and simultaneously generating an effective review effect through repetitive learning.

In order to secure storage space efficiency, the smart speaker device 30 is preferably implemented to delete the script received from the user terminal 10 after outputting it in the voice of a specific user.

6 is a flowchart illustrating a method of driving a voice output system capable of selecting a user-customized voice according to an embodiment of the present invention.

First, the cloud server collects and stores user voice data from a user terminal in which a user-specific voice output service dedicated application is driven (S600).

Then, if the storage amount of the voice data of the specific user is greater than or equal to the reference value (S610), the voice data of the specific user is analyzed to learn the characteristics of the voice data of the specific user (S620).

In this case, the reference value of the voice data storage amount may be counted as the time of the voice data stored in an amount sufficient to grasp the voice data of a specific user, or may be counted as the storage capacity.

And learning the characteristics of the voice data of a specific user may be implemented as an application program that can analyze the wavelength, frequency, sound color, etc. of the user voice data in one embodiment. Or it is interpreted to include any means for extracting and distinguishing the wavelength, frequency, sound color, etc. of the input voice as including a sensing device.

It is also possible to collect the user's voice data while the user is making a voice call to the user terminal to secure the user's voice data. At this time, it is preferable that the user is allowed to collect voice data during a call through an application dedicated to a user-specific voice output service.

In one aspect, learning the characteristics of voice data is characterized by performing deep learning-based voice learning.

Deep learning (deep learning) and deep learning are high-level abstractions (abstracts, summarizing key content or functions in a large amount of data or complex data) through a combination of several nonlinear transformation methods. It is defined as a set of machine learning algorithms that attempt.

It can be said that in a large frame, it is a field of machine learning that teaches a person's way of thinking to computers. Accordingly, it is possible to define a voice characteristic of a specific user with less voice data than before.

In addition, the cloud server generates a voice style file defining a voice characteristic of a specific user based on the learned content (S630).

In one embodiment, a voice style file defining a voice characteristic of a specific user may be style.xml.

Extensible Markup Language (XML) is a multipurpose markup language developed by W3C that is recommended for use in creating other special purpose markup languages. XML is a simplified subset of SGML and can be used to describe many different kinds of data. XML mainly makes it easy to exchange data between different types of systems, especially those connected to the Internet.

Thereafter, the smart speaker device converts text-form data received from a user terminal equipped with a user-specific voice output service-only application into voice form data (S640).

The smart speaker device receives text data from a user terminal in a short-range wireless communication method.

The user may select a file stored in the user terminal through a user-specific voice output service-only application, or select a text-type script by directly inputting text to generate a text file.

Then, the smart speaker device applies the converted voice type data to a voice style file of a specific user received from the cloud server and outputs the voice in a specific user (S650).

At this time, when the cloud server generates a voice style file of a specific user, the application for a user-specific voice output service generates and outputs a voice selection button for a specific user.

According to an additional aspect of the present invention, the step of learning the characteristics of the voice data further reflects the user's emotional state information and intonation information in learning the features of the user's voice data, and the voice style file includes the user's emotional state information and It is characterized by including accent information.

In addition, the smart speaker device may apply a text-type script received from the user terminal to a voice style file of a specific user selected by the user and read the voice of a specific user selected by utilizing the TTS function.

At this time, the voice style file of the specific user selected by the user is downloaded from the cloud server.

After this, the smart speaker device is preferably implemented to delete the script after outputting the received script in the voice of a specific user in order to secure storage efficiency.

10: user terminal 20: cloud server
30: smart speaker device 210: voice data storage
220: voice learning unit 230: voice style file generation unit
240: communication unit 310: TTS module unit
320: audio output unit 330: communication unit
335: short-range wireless communication unit

Claims (2)

User voice data is collected and stored from a user terminal in which a user-specific voice output service-only application is driven, and if a specific user's voice data storage is greater than or equal to a reference value, the voice data of the specific user is analyzed to characterize the voice data of the specific user. A cloud server including a learning voice learning unit and a voice style file generating unit for generating a voice style file defining a voice characteristic of the specific user based on the content learned by the voice learning unit; And
The text-to-speech (TTS) module unit converts text-type data received from a user terminal equipped with the application for the user-specific voice output service into speech-type data, and voice-type data converted by the TTS module unit. It includes; a smart speaker device including a voice output unit applied to the voice style file of the specific user received from the cloud server to output the voice of the specific user;
The user-specific voice output service-only application is characterized in that when the voice style file generation unit generates a voice style file of a specific user, a voice selection button for the specific user is generated and output.
The user-specific voice output service-only application further provides a list of categories for selecting voices,
The smart speaker device is characterized in that it receives a voice style file of a specific user selected from the voice selection button for the user generated in the application for the customized voice output service, from the cloud server,
The voice learning unit further reflects the user's emotional state information and intonation information in learning the characteristics of the user's voice data, and wherein the voice style file includes the user's emotional state information and intonation information,
The smart speaker device further includes a short-range wireless communication unit for performing short-range wireless communication with the user terminal,
The voice output unit is characterized in that it receives text data from the user terminal through the short-range wireless communication unit,
The smart speaker device receives a control signal for the overall operation of the smart speaker device through a user-specific voice output service-only application, and adds at least one of voice selection information, text-type data script, and emotion information about the voice. Characterized by receiving input,
The voice learning unit is characterized by performing deep learning-based voice learning,
The voice learning unit further comprises collecting user voice data from the user terminal while a user makes a voice call to the user terminal.
The cloud server collects and stores user voice data from a user terminal in which a user-specific voice output service-only application is driven, and when a specific user's voice data storage amount is greater than a reference value, analyzes the voice data of the specific user and analyzes the voice data of the specific user Learning the features of the;
Generating, by the cloud server, a voice style file defining a voice characteristic of the specific user based on the learned content;
Converting text-type data received from a user terminal equipped with the application dedicated to the user-specific voice output service into voice-type data by a smart speaker device; and
And the smart speaker device applying the converted voice data to the voice style file of the specific user received from the cloud server and outputting the voice to the specific user.
The user-specific voice output service-only application is characterized in that when the cloud server generates a voice style file of a specific user, a voice selection button for the specific user is generated and output.
The user-specific voice output service-only application further provides a list of categories for selecting voices,
The smart speaker device is characterized in that it receives a voice style file of a specific user selected from the voice selection button for the generated user in the application dedicated to the customized voice output service from the cloud server,
The step of learning the characteristics of the voice data further reflects the user's emotional state information and intonation information in learning the characteristics of the user's voice data, and the voice style file includes the user's emotional state information and intonation information. Features,
The smart speaker device further comprises receiving text data from the user terminal in a short-range wireless communication method.
The step of outputting in a voice may further include receiving text data from the user terminal in the short-range wireless communication method.
The smart speaker device receives a control signal for the overall operation of the smart speaker device through a user-specific voice output service-only application, and adds at least one of voice selection information, text-type data script, and emotion information about the voice. Characterized in that the input,
The step of learning the characteristics of the voice data is characterized by performing deep learning-based voice learning,
The step of learning the characteristics of the voice data is a method of driving a voice output system capable of selecting a user-customized voice, further comprising collecting user voice data from the user terminal while a user makes a voice call to the user terminal.

Images