KR20200114606A - Methode and aparatus of providing voice - Google Patents

Abstract

According to one embodiment, provided is a method and device for providing a voice, wherein the method comprises in receiving a first voice uttered by a speaker, obtaining speaker information corresponding to the speaker recognized based on the first voice, generating a second voice corresponding to the first voice based on an analysis result of the first voice, and providing a third voice for the speaker by correcting a frequency of the second voice based on the speaker information.

Description

TECHNICAL FIELD [Method and Aparatus of Providing Voice]

Embodiments relate to a method and apparatus for providing voice.

With the development of artificial intelligence technology, various services (for example, sound source provision, control through voice recognition, autonomous driving, augmented reality, etc.) are provided through smart devices such as an AI (Artificial Intelligence) secretary or AI speaker. These services are generally provided through audio or video. However, there are users who have difficulty listening to a signal in a specific frequency band due to a congenital disorder or aging. When a sound source or voice out of the audible frequency band of these users is provided, there is a problem in that it is difficult for the users to accurately recognize the provided sound source or voice, or they cannot recognize it at all.

According to an embodiment, it is possible to provide a personalized speech synthesis (Text-To-Speech; TTS) service through speaker recognition.

According to an embodiment, by generating and providing voice in consideration of the speaker's audible frequency band when servicing through a smart device, it is possible to provide a voice or sound source in the optimal frequency band to a user whose audible frequency is reduced due to congenital disorder or aging. have.

According to an embodiment, a method of providing a voice includes receiving a first voice spoken by a speaker; Recognizing the speaker based on the first voice; Obtaining speaker information corresponding to the recognized speaker; Generating a second voice responsive to the first voice based on the analysis result of the first voice; And providing a third voice for the speaker by correcting the frequency of the second voice based on the speaker information.

The providing of the third voice may include determining an audible frequency band of the speaker based on the speaker information; And generating the third voice by correcting the frequency of the second voice based on a comparison result of the frequency band of the second voice and the audible frequency band of the speaker.

The step of generating the third voice based on the comparison result includes: when the second voice is out of the speaker's audible frequency band, the second voice is shifted to the speaker's audible frequency band and sampled. It may include generating a third voice.

The generating of the third voice based on the comparison result may include determining whether the second voice is a signal of a high frequency band outside the audible frequency band of the speaker; And generating the third voice by reinforcing energy of the high frequency band of the second voice according to the determination that the second voice is a signal of the high frequency band.

The method of providing the voice may further include receiving a selection of the speaker as to whether or not to use the third voice.

In the providing of the third voice for the speaker, if the speaker information does not include the speaker's audible frequency band, the frequency of the second voice is corrected based on the average audible frequency for each age group of the speaker. It may include generating a third voice for the speaker.

The speaker information may include at least one of the speaker's identification information, the speaker's age, the speaker's gender, the speaker's audible frequency band, the speaker's average audible frequency by age group, and the speaker's preferred voice frequency band. I can.

The generating of the second voice may include generating an answer sentence in response to the first voice based on an analysis result of the first voice; And generating a second voice in response to the first voice based on the answer sentence.

The generating of the answer sentence may include converting the first voice into a first text by STT (Speech To Text); And generating an answer sentence in response to the first text through natural language processing (NLP) on the first text.

The method of providing the voice may further include authenticating the recognized speaker.

Recognizing the speaker may include determining whether the first voice includes a predetermined pager; And recognizing the speaker by applying the first voice to a speaker recognition engine according to a determination that the first voice includes the call word.

Recognizing the speaker may include extracting features from the first voice; And recognizing the speaker by applying the feature to the previously learned speaker-specific model.

According to an embodiment, an apparatus for providing a voice includes a communication interface for receiving a first voice spoken by a speaker; Based on the first voice, the speaker is recognized, speaker information corresponding to the recognized speaker is obtained, and a second voice responsive to the first voice is generated based on the analysis result of the first voice And a processor for generating a third voice for the speaker by correcting a frequency of the second voice based on the speaker information; And a speaker providing the third voice.

The processor determines the audible frequency band of the speaker based on the speaker information, and corrects the frequency of the second voice based on a comparison result of the frequency band of the second voice and the audible frequency band of the speaker. A third voice can be generated.

If the second voice is out of the speaker's audible frequency band, the processor may generate the third voice by moving the second voice to the speaker's audible frequency band and sampling the second voice.

The processor determines whether the second voice is a signal of a high frequency band outside the speaker's audible frequency band, and according to the determination that the second voice is a signal of the high frequency band, the energy of the high frequency band of the second voice It is possible to generate the third voice by reinforcing it.

The apparatus for providing the voice may further include a user interface for receiving a selection of the speaker as to whether or not to use the third voice.

When the speaker information does not include the speaker's audible frequency band, the processor corrects the frequency of the second voice based on the average audible frequency for each age group of the speaker to provide a third voice for the speaker. I can.

The processor determines whether the first voice includes a predetermined call word, and applies the first voice to the speaker recognition engine according to the determination that the first voice includes the call word to recognize the speaker. I can.

According to one side, it is possible to provide a personalized speech synthesis (Text-To-Speech; TTS) service through speaker recognition.

According to one side, by generating and providing voice in consideration of the speaker's audible frequency band when servicing through a smart device, it is possible to provide a voice or sound source in the optimal frequency band to a user whose audible frequency is reduced due to congenital disorder or aging. .

1 is a flowchart illustrating a method of providing voice according to an embodiment.
2 is a diagram for describing a configuration of an apparatus for providing voice according to an embodiment.
3 is a diagram for describing an operation between a smart device and a speaker recognition server according to an exemplary embodiment.
4 is a diagram for explaining an operation of a speech synthesis (TTS) apparatus according to an embodiment.
5 is a flowchart illustrating a method for providing voice according to another embodiment.
Fig. 6 is a block diagram of a device for providing voice according to an embodiment.

Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings. The same reference numerals in each drawing indicate the same members.

Various changes may be made to the embodiments described below. The embodiments described below are not intended to be limited to the embodiments, and should be understood to include all changes, equivalents, and substitutes thereto.

The terms used in the examples are used only to describe specific embodiments, and are not intended to limit the embodiments. Singular expressions include plural expressions unless the context clearly indicates otherwise. In the present specification, terms such as "comprise" or "have" are intended to designate the presence of features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, but one or more other features. It is to be understood that the presence or addition of elements or numbers, steps, actions, components, parts, or combinations thereof, does not preclude in advance.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiment belongs. Terms as defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and should not be interpreted as an ideal or excessively formal meaning unless explicitly defined in this application. Does not.

In addition, in the description with reference to the accompanying drawings, the same reference numerals are assigned to the same components regardless of the reference numerals, and redundant descriptions thereof will be omitted. In describing the embodiments, when it is determined that a detailed description of related known technologies may unnecessarily obscure the subject matter of the embodiments, the detailed description thereof will be omitted.

1 is a flowchart illustrating a method of providing voice according to an exemplary embodiment. Referring to FIG. 1, a device for providing a voice according to an exemplary embodiment (hereinafter, a “providing device”) receives a first voice spoken by a speaker (110). The'first voice' may correspond to, for example, a voice uttered by a speaker directly or a voice recorded from a speaker. The first voice may include, for example, a question or a request for a smart device.

The providing device recognizes a speaker based on the first voice (120). The providing device may determine, for example, whether the first voice includes a predetermined call word. The predetermined caller may be, for example, a common conversational language such as "Hi OO", or may be a call name such as "Alec O". The providing device may recognize a speaker by applying the first voice to the speaker recognition engine according to a determination that the first voice includes a predetermined call word. The providing device may recognize a speaker by extracting a feature from, for example, the first voice and applying the feature to a pre-learned speaker-specific model. The providing device may also perform speaker authentication as to whether the corresponding speaker is a previously registered speaker. The providing device may perform speaker recognition and/or speaker authentication by, for example, the speaker recognition engine 231 described in FIG. 2 below. The speaker recognition engine will be described in detail with reference to FIGS. 2 and 3 below.

The providing device acquires speaker information corresponding to the speaker recognized in step 120 (130). The speaker information may include, for example, identification information of the speaker, the age of the speaker, the sex of the speaker, an audible frequency band of the speaker, an average audible frequency by age group of the speaker, and a voice frequency band preferred by the speaker. The providing device may obtain speaker information on the recognized speaker, for example, from the user database 239 described in FIG. 2 below.

The providing device generates a second voice in response to the first voice based on the analysis result of the first voice (140). The providing device may generate an answer sentence in response to the first voice based on the analysis result of the first voice. The providing device may convert the first voice into the first text by, for example, Speech To Text (STT) technology. The providing device may generate an answer sentence in response to the first text through natural language processing (NLP) on the first text. The providing device may generate a second voice in response to the first voice based on the answer sentence. Hereinafter, the'second voice' may be understood as a voice generated through a providing device, more specifically, a voice synthesis (TTS) module (refer to reference numeral 250 in FIG. 2) of the providing device.

The providing device corrects the frequency of the second voice based on the speaker information and provides a third voice for the speaker (150). Hereinafter, the'third voice' may be a voice obtained by frequency-correcting the second voice. The providing device may determine the audible frequency band of the speaker based on, for example, speaker information. The providing apparatus may generate the third voice by correcting the frequency of the second voice based on the comparison result of the frequency band of the second voice and the speaker's audible frequency band. For example, when the second voice is out of the speaker's audible frequency band, the providing apparatus may generate the third voice by moving the second voice to the speaker's audible frequency band and performing sampling. Alternatively, the providing device may determine whether the second voice is a signal of a high frequency band outside the speaker's audible frequency band. The providing device may generate the third voice by reinforcing energy in the high frequency band of the second voice according to the determination that the second voice is a signal in the high frequency band. The providing device may, for example, increase the energy value of a high frequency to make it sound like another frequency band. For example, let's say there are users who can't listen to the 6Khz frequency band well. The user may not be able to hear the content well if it is provided at the same sound pressure level (SPL) as in the general frequency band. Accordingly, the providing device may increase and reinforce the energy value of the corresponding frequency so that the user can listen to the 6Khz frequency band at the level of the existing sound pressure.

2 is a diagram illustrating a configuration of an apparatus for providing voice according to an exemplary embodiment. Referring to FIG. 2, a smart device 210 and a voice providing apparatus 230 according to an embodiment are shown. Hereinafter, for convenience of description, the smart device 210 and the voice providing apparatus 230 are described as being separately configured as shown in FIG. 2, but are not limited thereto. The voice providing device 230 may be configured as a single device merged with the smart device 210. The speech providing device 230 includes, for example, a speaker recognition engine 231, an STT/NLP module 233, a speech synthesis (TTS) module 235, a frequency correction module 237, and a user database 239. It may include. The speaker recognition engine 231 may be implemented as a separate speaker recognition server, for example.

The smart device 210 may be a device including an AI assistant function, such as a smart phone, an AI speaker, an Internet of Things (IoT) device, and a set-top box. The smart device 210 may receive the first voice spoken by the speaker (201). The smart device 210 may transmit 202 a file including the first voice to the speaker recognition engine 231 to request speaker recognition (speaker identification) and/or speaker authentication of the corresponding voice. The operation between the smart device 210 and the speaker recognition engine 231 for speaker recognition and/or speaker authentication will be described in detail with reference to FIG. 3 below.

For example, the speaker recognition engine 231 may recognize a speaker of the first voice and authenticate whether the recognized speaker is a pre-registered speaker. When authentication for the corresponding speaker is completed, that is, if it is determined that the corresponding speaker is an authenticated speaker, the speaker recognition engine 231 may request the smart device 210 to activate a microphone for speech recognition (203). The speaker recognition engine 231 may request the smart device 210 to activate the microphone through, for example, a wake-up request. In addition, the speaker recognition engine 231 may transmit the user information of the corresponding user and the TTS setting information of the corresponding user to the speech synthesis (TTS) module 235 based on the user identifier identified through speaker authentication (204). The user information may include, for example, a user ID and audible frequency information of the user. User information of a corresponding user and TTS setting information of a corresponding user may also be referred to as'speaker information'. The TTS configuration information may include configuration information for a frequency band desired by a corresponding user. Speaker information may be stored in the user database 239. In this case, the user database 239 may be included in the frequency correction module 237 or may be configured separately from the frequency correction module 237.

After activation of the microphone, the smart device 210 may transmit a spoken voice (eg, “How about today's fine dust?”) from the user to the STT/NLP module 233 (205).

The STT/NLP module 233, for example, analyzes the intent and/or meaning of the spoken speech to respond to the spoken speech (for example, "fine dust today is at a very bad level. Please wear a mask") ) Can be created as text. More specifically, the STT/NLP module 233 may convert the spoken speech into the first text by the speech synthesis (TTS) module 235 based on the analysis result of the spoken speech. The STT/NLP module 233 may generate a text response sentence in response to the first text through Natural Language Processing (NLP) on the first text. The STT/NLP module 233 may transmit a text-type answer sentence to the speech synthesis (TTS) module 235 (206).

The speech synthesis (TTS) module 235 may generate a text-type answer sentence (“fine dust today is very bad. Please wear a mask”) as a second voice. The process of converting text into speech by the speech synthesis (TTS) module 235 is as follows.

For example, suppose you want to convert a text such as "How's the weather?" into speech. In this case, the speech synthesis (TTS) module 235 may classify a proper noun included in the user dictionary among terms included in the text through text preprocessing. The speech synthesis (TTS) module 235 may analyze a phrase of a corresponding text using, for example, a part-of-speech dictionary database, and tag the part-of-speech of each word according to the analysis result. The speech synthesis (TTS) module 235 sets a boundary for classifying each word using, for example, a delimiter, and converts a character string and pronunciation using a pronunciation dictionary to generate an entire sentence Can be provided.

The second voice generated by the voice synthesis (TTS) module 235 may be, for example, a signal in a frequency band of 22.02 Khz. The voice synthesis (TTS) module 235 may transmit the second voice to the frequency correction module 237. The configuration of the speech synthesis (TTS) module 235 may be referred to in FIG. 4 below.

The frequency correction module 237 may correct the frequency of the second voice based on speaker information stored in the user database 239, for example. For example, assume that the frequency of the second voice is a 4Khz band, and the audible frequency band according to the speaker information is a 3Khz band. In this case, the frequency correction module 237 may generate a third voice by correcting the second voice having a frequency in the 4Khz band into a signal in the 3Khz band that the speaker can hear. The frequency correction module 237 may provide the third voice to the smart device 210 through the voice synthesis (TTS) module 235 (207). The smart device 210 may provide a third voice ("fine dust today is very bad. Please wear a mask") to the speaker.

The frequency correction module 237 identifies the audible frequency for each speaker through speaker information stored in the user database 239, and moves the voice signal in a frequency band that each speaker cannot hear well to a frequency band where the speaker can hear well. Can be sampled. Alternatively, the frequency correction module 237 may generate a third voice by reinforcing energy in a frequency band that the speaker does not hear well, or may generate a third voice by moving a frequency for a frequency band that the speaker does not hear well.

The providing device 230 may play each third voice generated through energy reinforcement or frequency shift to the speaker, and allow the speaker to select any one of the third voices they like. For example, when the speaker does not input information on the audible frequency band or does not make separate settings related to the audible frequency, the providing device 230 may perform a speech synthesis (TTS) module 235 without separate frequency correction. The generated second voice of the basic frequency band may be provided to the smart device 210.

3 is a diagram for explaining an operation between a smart device and a speaker recognition server according to an exemplary embodiment. Referring to FIG. 3, a process in which the speaker recognition server 300 learns a speaker-specific model to identify speakers based on a calling name input through the smart device 210 is illustrated. The speaker recognition server 300 may correspond to the speaker recognition engine 231 described above.

For example, suppose that a call name A spoken by a speaker is received by the smart device 210. The smart device 210 may recognize whether the call name A is, for example, a pre-arranged call name that triggers the operation of the speaker recognition server 300. The smart device 210 may link (or drive) the speaker recognition server 300 by transmitting the calling name A to the speaker recognition server according to the determination that the calling name A is a predetermined calling name.

For example, when a call name A is first input, the speaker recognition server 300 may perform speaker registration 310 with a voice corresponding to the call name A. The speaker recognition server 300 may generate a final speaker model 320 by, for example, extracting 311 features from the calling name A and learning 313 a speaker model using the extracted features. According to an embodiment, the speaker model 320 may be generated by learning a speaker model based on features extracted from the recorded voice of the speaker. The recorded voice of the speaker may be stored in a recording database, for example.

In addition, for example, when the call name A is input after the speaker registration 310, the speaker recognition server 300 may perform the speaker recognition 330 by the call name A. The speaker recognition server 300 may perform speaker recognition 333 by extracting 331 features from the calling name A and comparing the extracted features with information registered in the speaker model 320.

4 is a diagram illustrating an operation of a speech synthesis (TTS) apparatus according to an embodiment. 4, a speech synthesis (TSS) apparatus 400 according to an embodiment transmits text to speech through an end-to-end-based deep neural network (DNN). It can be synthesized. The deep neural network may include, for example, an encoder 410, an attention model 420, a decoder 430, and a vocoder 440.

It is assumed that text is input to the speech synthesis (TSS) device 400. In this case, the encoder 410 may convert the input text into a number or vector representing the characteristics of the text. The encoder 410 may determine a feature vector representing a characteristic of the text from the input text. The encoder 410 may be configured as a CBHG network including, for example, a primary convolution bank, a highway network, and a bidirectional gated recurrent unit (GRU). For example, when the speech synthesis (TSS) apparatus 400 is learned, text and a correct answer sound source may be input together.

The attention model 420 may determine context information of text from the feature vector determined by the encoder 410. The attention model 420 may predict a next spectrogram from context information of text. The spectrogram is, for example, a sound spectrogram, and may represent the distribution of the stimulus intensity and frequency of a word.

The decoder 430 may generate a next spectrogram based on the spectrogram after prediction by the attention model 420. The attention model 420 and the decoder 430 may be configured as, for example, an attention-recurrent neural network (RNN).

The vocoder 440 may convert parameter values for each phoneme into a sound source or voice. The vocoder 440 may convert the next spectrogram generated by the decoder 430 into a sound source or voice. The vocoder 440 may include, for example, a Griffin-Lim vocoder algorithm.

5 is a flowchart illustrating a method of providing voice according to another embodiment. Referring to FIG. 5, the providing device according to an embodiment receives a first voice spoken by a speaker (505 ).

The providing device recognizes a speaker based on the first voice (510).

The providing device acquires speaker information corresponding to the recognized speaker (515 ).

The providing device generates a second voice in response to the first voice based on the analysis result of the first voice (520).

The providing device may determine whether the speaker information includes information on the speaker's audible frequency band (525 ). For example, in step 525, if the speaker information includes information on the speaker's audible frequency band, the providing device may determine the speaker's audible frequency band based on the speaker information (530 ).

The providing apparatus may generate the third voice by correcting the frequency of the second voice based on the comparison result of the frequency band of the second voice and the audible frequency band of the speaker (535).

The providing device may receive a speaker's selection as to whether or not to use the third voice (540 ). The providing device may provide the third voice according to the speaker's selection in step 540 (545 ).

Alternatively, according to an embodiment, in step 525, if the speaker information does not include information on the speaker's audible frequency band, the providing device corrects the frequency of the second voice based on the average audible frequency for each age group of the speaker. Thus, a third voice for the speaker may be generated (550). Thereafter, the providing device may receive a speaker's selection as to whether to use the third voice (operation 540), and provide the third voice according to the speaker's selection (operation 545).

6 is a block diagram of an apparatus for providing voice according to an exemplary embodiment. Referring to FIG. 6, an apparatus for providing voice ('providing device') 600 according to an embodiment includes a communication interface 610, a processor 630, and a speaker 670. The providing device 600 may further include a recorder 620, a memory 650, and a user interface (not shown). The communication interface 610, the recorder 620, the processor 630, the memory 650, and the speaker 670 may communicate with each other through a communication bus 605.

The communication interface 610 receives the first voice spoken by the speaker.

The recorder 620 may record the first voice spoken by the speaker.

The processor 630 recognizes a speaker based on, for example, the first voice received through the communication interface 610 or recorded by the recorder 620. The processor 630 acquires speaker information corresponding to the recognized speaker. The processor 630 generates a second voice in response to the first voice based on the analysis result of the first voice. The processor 630 generates a third voice for the speaker by correcting the frequency of the second voice based on the speaker information.

The processor 630 determines the speaker's audible frequency band based on the speaker information. The processor 630 generates a third voice by correcting the frequency of the second voice based on the comparison result of the frequency band of the second voice and the speaker's audible frequency band.

For example, when the second voice is out of the speaker's audible frequency band, the processor 630 generates a third voice by moving and sampling the second voice into the speaker's audible frequency band. Alternatively, the processor 630 determines whether the second voice is a signal in a high frequency band outside the speaker's audible frequency band. The processor 630 generates a third voice by reinforcing energy in the high frequency band of the second voice according to the determination that the second voice is a signal in the high frequency band. The processor 630 may determine whether the speaker information does not include the speaker's audible frequency band. For example, if the speaker information does not include the speaker's audible frequency band, the processor 630 provides a third voice for the speaker by correcting the frequency of the second voice based on the average audible frequency for each age group of the speaker. do.

The processor 630 may determine, for example, whether the first voice includes a predetermined call word. The processor 630 may recognize the speaker by applying the first voice to the speaker recognition engine according to the determination that the first voice includes a pager.

The memory 650 may store speaker information acquired by the processor 630.

The speaker 670 provides the third voice generated by the processor 630.

The providing device 600 may receive, for example, a speaker's selection of whether to use the third voice through a user interface displayed on a screen of a touch display (not shown).

In addition, the processor 630 may perform at least one method or an algorithm corresponding to at least one method described above through FIGS. 1 to 5. The processor 630 may be a data processing device implemented in hardware having a circuit having a physical structure for executing desired operations. For example, desired operations may include code or instructions included in a program. For example, a data processing device implemented in hardware is a microprocessor, a central processing unit, a processor core, a multi-core processor, and a multiprocessor. , Application-Specific Integrated Circuit (ASIC), and Field Programmable Gate Array (FPGA).

The processor 630 may execute a program and control the providing device 600. The program code executed by the processor 630 may be stored in the memory 650.

The memory 650 may store various types of information received through the communication interface 610. The memory 650 may store at least one of a first voice, a second voice, and a third voice.

In addition, the memory 650 may store various types of information generated during processing in the processor 630 described above. In addition, the memory 650 may store various types of data and programs. The memory 650 may include a volatile memory or a nonvolatile memory. The memory 650 may include a mass storage medium such as a hard disk and may store various types of data.

The method according to an embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the present invention, or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -A hardware device specially configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of the program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The above-described hardware device may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

As described above, although the present invention has been described by the limited embodiments and drawings, the present invention is not limited to the above embodiments, and various modifications and variations from these descriptions are those of ordinary skill in the field to which the present invention belongs. This is possible.

Therefore, the scope of the present invention is limited to the described embodiments and should not be defined, and should be defined by the claims and equivalents as well as the claims to be described later.

600: providing device
605: communication bus
610: communication interface
620: recorder
630: processor
650: memory
670: speaker

Claims (20)

Receiving a first voice spoken by a speaker;
Recognizing the speaker based on the first voice;
Obtaining speaker information corresponding to the recognized speaker;
Generating a second voice responsive to the first voice based on the analysis result of the first voice; And
Providing a third voice for the speaker by correcting a frequency of the second voice based on the speaker information
Containing, a method of providing a voice. The method of claim 1,
Providing the third voice
Determining an audible frequency band of the speaker based on the speaker information; And
Generating the third voice by correcting the frequency of the second voice based on a comparison result of the frequency band of the second voice and the audible frequency band of the speaker
Containing, a method of providing a voice. The method of claim 2,
Generating the third voice based on the comparison result
When the second voice is out of the speaker's audible frequency band, generating the third voice by moving the second voice to the speaker's audible frequency band and performing sampling
Containing, a method of providing a voice. The method of claim 2,
Generating the third voice based on the comparison result
Determining whether the second voice is a signal of a high frequency band outside the audible frequency band of the speaker; And
Generating the third voice by reinforcing energy in the high frequency band of the second voice according to the determination that the second voice is a signal in the high frequency band
Containing, a method of providing a voice. The method of claim 2,
Receiving a selection of the speaker as to whether to use the third voice
The method of providing a voice further comprising. The method of claim 2,
Providing the third voice for the speaker comprises:
If the speaker information does not include the speaker's audible frequency band,
Generating a third voice for the speaker by correcting the frequency of the second voice based on the average audible frequency for each age group of the speaker
Containing, a method of providing a voice. The method of claim 1,
The speaker information above is
Providing a voice including at least one of the speaker's identification information, the speaker's age, the speaker's gender, the speaker's audible frequency band, the speaker's average audible frequency by age group, and the speaker's preferred voice frequency band How to. The method of claim 1,
The step of generating the second voice
Generating an answer sentence in response to the first voice based on the analysis result of the first voice; And
Generating a second voice in response to the first voice based on the answer sentence
Containing, a method of providing a voice. The method of claim 8,
The step of generating the answer sentence
Converting the first voice into first text by STT (Speech To Text); And
Generating an answer sentence in response to the first text through natural language processing (NLP) on the first text
Containing, a method of providing a voice. The method of claim 1,
Authenticating the recognized speaker
The method of providing a voice further comprising. The method of claim 1,
Recognizing the speaker is
Determining whether the first voice includes a predetermined call word; And
Recognizing the speaker by applying the first voice to a speaker recognition engine according to a determination that the first voice includes the call word
Containing, a method of providing a voice. The method of claim 1,
Recognizing the speaker is
Extracting features from the first voice; And
Recognizing the speaker by applying the feature to a previously learned speaker-specific model
Containing, a method of providing a voice. A computer program stored in a computer-readable recording medium in combination with hardware to execute the method of claim 1. A communication interface for receiving the first spoken voice of the speaker;
Based on the first voice, the speaker is recognized, speaker information corresponding to the recognized speaker is obtained, and a second voice responsive to the first voice is generated based on the analysis result of the first voice And a processor for generating a third voice for the speaker by correcting a frequency of the second voice based on the speaker information; And
Speaker providing the third voice
Containing, a device for providing a voice. The method of claim 14,
The processor is
Based on the speaker information, an audible frequency band of the speaker is identified, and the frequency of the second voice is corrected based on a comparison result of the frequency band of the second voice and the audible frequency band of the speaker. A device that provides a voice to generate. The method of claim 15,
The processor is
When the second voice is out of the speaker's audible frequency band, the second voice is moved to the speaker's audible frequency band and sampled to generate the third voice. The method of claim 15,
The processor is
By determining whether the second voice is a signal of a high frequency band outside the speaker's audible frequency band, and reinforcing energy of the high frequency band of the second voice according to the determination that the second voice is a signal of the high frequency band An apparatus for providing a voice for generating the third voice. The method of claim 15,
User interface for receiving input of the speaker's selection regarding whether to use the third voice
The apparatus further comprising a voice providing. The method of claim 15,
The processor is
If the speaker information does not include the speaker's audible frequency band,
An apparatus for providing a voice, for providing a third voice for the speaker by correcting a frequency of the second voice based on an average audible frequency for each age group of the speaker. The method of claim 15,
The processor is
A voice that recognizes the speaker by determining whether the first voice includes a predetermined call word, and applying the first voice to a speaker recognition engine according to a determination that the first voice includes the call word. Device to provide.

Images