KR101429138B1 - Speech recognition method at an apparatus for a plurality of users - Google Patents

Abstract

The method includes the steps of: (a) extracting a voice feature for identifying a user from a received voice from a user; (b) searching for whether the extracted voice feature is present in a voice feature stored in the device; and ) Determining the text corresponding to the user ' s voice if the extracted voice feature is present. ≪ RTI ID = 0.0 > [0002] < / RTI >
By using the present invention, more accurate voice recognition can be performed at low cost in a multi-shop apparatus, and a multi-shop apparatus can be conveniently controlled.

Description

[0001] SPEECH RECOGNITION METHOD AT APPARATUS FOR A PLURALITY OF USERS [0002]

The present invention relates to a method for recognizing a voice in a device for a plurality of users, and more particularly, to a voice recognition method using a voice feature of each user extracted from a voice of a user, To a method for speech recognition in a device for a plurality of users.

A multi-function shop, such as a karaoke room, a DVD room, a multi-room, a coffee shop, etc., is provided with a specific device in accordance with a service or function provided by each shop. Such a device may be, for example, a song accompaniment device, a DVD player, or a device capable of reproducing a combination of video and / or music.

On the other hand, as the development of electronic technology, voice recognition is being commercialized as an alternative to or in parallel with a letter or key input. Such a speech recognition technology is installed in, for example, a mobile phone or a navigation system, and receives a voice from a user, converts the voice to text, displays the converted result to the user, or re-transmits the voice converted result.

On the other hand, there are various restrictions on the use of such speech recognition technology in devices used in multi-businesses.

First, since such a device is not a device for a specific user but a device for a plurality of users, it is necessary to accurately recognize the voice, which is also a problem related to the reliability of the product. For example, when speech recognition is inaccurate, it may cause inconvenience not only to the user who entered the speech but also to other persons in the same space.

In addition, the use environment of multiple businesses is not favorable for voice recognition. For example, it is an environment in which video, music, or voice is generated as a large output by each device disposed in multiple businesses, and voices emitted by a plurality of users are mixed. Speech recognition in such an environment necessarily has difficulties in speech recognition.

In addition, users of multiple businesses are not an individual who owns and use devices of multiple businesses but an unspecified number of users. In consideration of such an unspecified number of users, it is necessary to provide an expensive speech recognition processor (for example, DSP) for more accurate speech recognition, which increases the production and development cost of a multi-site device.

Therefore, there is a need for a method of recognizing speech in a device for a plurality of users, which enables speech recognition to be provided accurately and inexpensively in consideration of the environment of such a multi-shop.

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above-described problems, and it is an object of the present invention to provide a voice recognition apparatus and a voice recognition method which automatically identifies a user using a multi- It is an object of the present invention to provide a method for recognizing a speech in a device for a plurality of users.

In addition, the present invention is not limited to the case where the voice characteristic recognized from the voice is different from the stored voice characteristic by using the voice characteristic of the user, so that the filtering of the received voice is simplified and the conversion into unnecessary text is eliminated And a voice recognition method in a device for a plurality of users.

The present invention also provides a method and apparatus for sharing a voice characteristic of each user in multiple devices of multiple businesses without exposing the user's personal important information and converting the voice from each user into text so that the multi- And an object of the present invention is to provide a speech recognition method in a device for a plurality of users.

Further, the present invention provides a method for recognizing a speech in a device for a plurality of users, which enables conversion from voice to text without increasing the performance load of a multi-occupancy device, so that a multi-occupancy device can be provided at low cost It has its purpose.

It is another object of the present invention to provide a method for recognizing speech in a device for a plurality of users, which enables a more accurate speech recognition by correcting a user's speech characteristic in real time.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not restrictive of the invention, unless further departing from the spirit and scope of the invention as defined by the appended claims. It will be possible.

According to another aspect of the present invention, there is provided a speech recognition method for a plurality of users, comprising the steps of: (a) extracting a speech feature for identifying a user from a received speech from a user; (b) Retrieving whether a feature is present in the voice feature stored in the device; and (c) if the extracted voice feature is present, determining the text corresponding to the user's voice.

In order to achieve the above object, there is also provided a method of recognizing speech in a device for a plurality of users, comprising the steps of: (a) receiving a registration request for a voice feature from a user; Extracting the voice feature from the voice received from the user, mapping the extracted voice feature to the user identifier received from the user, and registering the user by storing the mapped voice feature and the user identifier do.

According to still another aspect of the present invention, there is provided a method for recognizing speech in a device for a plurality of users, the method comprising: (d) identifying a control command from a determined text; and (e) .

According to another aspect of the present invention, there is provided a speech recognition method for a plurality of users, comprising the steps of: determining a difference between an extracted speech feature and a corresponding stored speech feature when the extracted speech feature exists; And modifying the corresponding stored voice feature using the determined difference.

The method for recognizing a speech in a device for a plurality of users according to the present invention as described above automatically identifies a user using a multi-user device by using a speech feature that can be determined from the speech, So that the accuracy of speech recognition can be improved and noise can be removed.

In addition, the method of recognizing speech in a plurality of user apparatuses according to the present invention may further comprise the steps of: It is possible to simplify the filtering of the voice and to eliminate the conversion into unnecessary text.

In addition, the method of recognizing speech in a device for a plurality of users according to the present invention, as described above, can share voice characteristics of each user in various devices of multiple businesses without exposing personal important information of the user, Is converted into a text so that the devices of multiple businesses can be controlled.

In addition, the speech recognition method for a plurality of users according to the present invention as described above can convert a speech-to-text into a text without increasing the performance load of the multi-tasking apparatus, It is effective.

In addition, the speech recognition method for a plurality of users according to the present invention as described above has the effect of enabling more accurate speech recognition by correcting the speech characteristics of the user in real time.

The effects obtained by the present invention are not limited to the above-mentioned effects, and other effects not mentioned can be clearly understood by those skilled in the art from the following description will be.

1 is a diagram illustrating a system environment in which a speech recognition method according to the present invention is implemented.
2 is a diagram showing an exemplary hardware block diagram of a multi-functional apparatus.
3 is a diagram showing a general processing procedure of a speech recognition method performed in a multi-business apparatus and / or a personal apparatus.
4 is a diagram showing an exemplary process flow for registering user-specific voice features according to S100.
5 is a diagram showing an exemplary process flow for controlling the apparatus using speech recognition according to S200.
6 is a diagram illustrating an exemplary process flow for modifying a voice feature in accordance with S300.

The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings, in which: It can be easily carried out. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram illustrating a system environment in which a speech recognition method according to the present invention is implemented.

1, the system environment includes one or more multi-premises devices 100 and / or one or more personal devices 200 and may also be connected to the Internet 100 Lt; RTI ID = 0.0 > 300 < / RTI >

Such a system environment may omit some system blocks as needed or may be configured to further include other system blocks not shown in Fig.

Referring to the respective blocks in the system environment of FIG. 1, the multi-functional apparatus 100 is an apparatus that can be used by an unspecified number of users or individuals in multiple businesses and can provide a function dedicated to the multiple businesses.

The multi-entertainment apparatus 100 may be, for example, a song accompaniment device, a DVD player, or a device capable of reproducing a combination of video and / or music.

The multi-premises device (100) can receive voice from a plurality of unspecified users by using a microphone or the like, and can extract voice features from the received voice. And the extracted voice feature is within the threshold range according to the voice feature stored in comparison with the voice feature stored in the multi-user device 100 or the remote server 300 or according to the specified critical range, (Or a character string) containing more than words or characters, and determines a control command for controlling the multi-functional apparatus 100 in the converted text, thereby controlling the multi-functional apparatus 100. [

Also, the multi-premises device 100 can use a simple identifier such as a telephone number or a mobile phone number as an identifier for the user. The identifier for this user may be received from the user and the received identifier may be used to determine the voice characteristics previously stored for the user stored in the multi-user device 100 or the remote server 300.

The multi-premises device (100) is configured to determine whether to perform speech recognition using the voice characteristics of a registered user and to convert the text into text, thereby improving the accuracy of speech recognition and preventing unnecessary conversion to text Thereby eliminating many problems that may be caused by noise or conversion into text by unregistered users.

The speech recognition method in the multi-premises device 100 using the speech feature will be described in detail with reference to FIG. 3 to FIG.

The personal device 200 is a device that is connected to the Internet and is used by an individual user carried or carried. For example, the personal device 200 may be a mobile phone, a smart phone, a tablet PC, a notebook, a stationary personal PC, and the like.

The personal device 200 may be connected to the personal server 200 by using a processor included in the personal device 200 and a network interface and an input interface for receiving voice. It is used to register voice features.

Accordingly, the personal device 200 can be used to identify a user identifier and a series of voices used to extract voice features, such as, for example, an identification (ID) or a mobile phone number or a telephone number, Voice features extracted from a series of voices can be delivered to the remote server 300 over the Internet. Accordingly, the remote server 300 receives or extracts the voice feature corresponding to the user identifier, maps the voice feature corresponding to the user identifier, and stores the voice feature in the database or the mass storage medium.

In order to register a user in this personal device 200, a program for registration may be stored in the mass storage medium of the personal device 200. [

The specific process flow for user registration, which may occur in the personal device 200, will be described in more detail with reference to FIG. 3 and FIG.

The remote server 300 includes a processor, a network interface, and a mass storage medium, and is capable of storing user identifiers and user information for user registration from one or more personal devices 200 and / A voice feature extracted from a voice or voice data of the user, and a voice feature corresponding to the received user identifier (extracting a voice feature from voice data when voice data is received) to a mass storage medium Hard disk) in the form of a database, for example.

The remote server 300 searches the database for the voice characteristic corresponding to the user identifier included in the request in response to the transmission request of the voice characteristic by the multi-party device 100, To the device (100).

Also, the database stored in the remote server 300 can be simply stored in the mapping relationship of the user identifier and the voice characteristic, thereby excluding the storage of additional information related to the other users.

For example, the database stores only the voice features corresponding to the user's phone number, mobile phone number, or user's ID (Identification), and stores the mapped voice characteristics, It is possible to prevent unauthorized hacking.

2 is a diagram showing an exemplary hardware block diagram of the multi-functional apparatus 100. As shown in FIG.

2, the multi-premises equipment 100 includes an input unit 110, an audio output unit 120, a video output unit 130, a memory 140, a hard disk 150, and a control processor 190 And also includes a microphone connection port 160, an analog-to-digital converter 170 and an audio processing processor 180, which are used for speech recognition.

Although the control processor 190 and the audio processing processor 180 are shown here as separate blocks for the sake of understanding, the control processor 190 does not need to be limited to the control processor 190, The hardware of the multi-functional apparatus 100 may be configured or integrated into one chipset (integrated circuit) by integrating an audio processing processor core and a control processor core.

Of course, the multi-functional apparatus 100 according to FIG. 2 may omit some hardware blocks as needed, or may be configured to further include other hardware blocks not shown in FIG.

2, the input unit 110 includes a touch panel, an input key button, and / or a remote control receiver port, and receives an input from a user.

Also, a specific key for setting and registering the voice characteristic of the user may be arranged on the remote controller through the touch panel, the input key button and / or the remote control receiver port. Through this specific key, Allows you to set up user mapping relationships.

Of course, this particular key can be replaced by the recognition of a control command for registration of a voice feature through the microphone, or it can be used in parallel.

The audio output unit 120 converts a digital audio signal output from the control processor 190, the audio processing processor 180, or the like into an analog audio signal and outputs the analog audio signal to a speaker or an earphone. The audio output unit 120 may include a digital analog converter for audio.

The video output unit 130 changes the digital video signal output from the control processor 190 or the like to the video format of the external display (not shown) and outputs the converted digital video signal to the external display. The video output unit 130 may include a digital-to-analog converter for video.

The memory 140 includes volatile memory such as DRAM (dynamic random access memory) and / or non-volatile memory such as Flash. The memory 140 may permanently or temporarily store various programs and image data or music or voice data driven by the control processor 190 or the audio processing processor 180.

The hard disk 150 stores various programs and various contents that can be used in the multi-functional apparatus 100. Such content may be different depending on the type of the multi-functional apparatus 100, and may include, for example, a compressed video movie file, an audio file in a MIDI format or a compressed format.

The microphone connection port 160 is a port for connecting a microphone that can be connected to the multi-functional apparatus 100. The microphone connected to the microphone connection port 160 may be a linear microphone capable of removing or minimizing external noise.

Such a straight-line microphone is configured to maximize and amplify the voice of a user using a microphone, and to minimize other noises (music sound, other user or human voice).

An analog-to-digital converter (ADC) 170 converts an analog voice signal from a microphone connected through a microphone connection port 160 into a digital voice signal, and converts the converted digital audio stream according to a specified format into audio And outputs it to the processing processor 180.

The analog-to-digital converter 170 can output a continuous digital audio stream by converting it into, for example, 16-bit 1-channel I2S format or the like.

The audio processing processor 180 extracts voice features for identifying the user from the received digital audio stream. The audio processing processor 180 may be a digital signal processor (DSP) for processing a digital audio stream.

The audio processing processor 180 may extract a voice feature by loading a voice feature extraction program for voice feature extraction stored in the memory 140 or the hard disk 150 and driving the extraction program.

These voice features can be extracted for each connected microphone.

Here, the voice characteristic is a frequency parameter for distinguishing the voice in the frequency domain determined from the voice received from the user (for example, using FFT, etc.) from the voice of the other user, Or may be a time parameter on the time domain that can be determined in the digital audio stream before or separately from the conversion, or a combination of the frequency parameter and the time parameter.

Such a voice feature constitutes a feature for distinguishing a specific user from another user. This voice feature can be different for each user even if the same word or sentence is read. By matching these voice features with the user, it is possible to simply filter the voice and enable more accurate conversion of the text from the voice.

The processing and voice characteristics in the audio processing processor 180 will be described in more detail in FIG. 4 and FIG.

The control processor 190 controls each hardware block. For example, the control processor 190 performs a function assigned to the multi-functional apparatus 100 by using a program stored in the memory 140 or the hard disk 150 or the like.

The control processor 190 loads a voice recognition program for voice recognition of the user stored in the memory 140 or the hard disk 150 and outputs the voice characteristic extracted by the audio processor 180 to the memory 140 or the hard disk 150. [ A search is made in the mapping relation between the user and the voice feature stored in the disk 150 or the like to determine whether the extracted voice feature exists and if there is a voice from the user is converted into text.

If there is no speech feature, the control processor 190 is configured not to convert to text to avoid unnecessary or incorrect conversion to text. Accordingly, it is possible to eliminate inconveniences and the like in the multi-functional apparatus 100 due to speech recognition by simple filtering.

Of course, the control processor 190 may perform the processing performed in the audio processing processor 180 instead of the audio processing processor 180, depending on the configuration example of the hardware block diagram. At this time, the audio processing processor 180 may be omitted.

Various specific process flows or control flows performed by the control processor 190 will be described with reference to FIGS. 3 through 6. FIG.

A system bus / control bus (not shown) connects the hardware blocks to transmit and receive data. For example, the system bus / control bus may be a serial bus or a parallel bus.

3 is a diagram showing a general processing procedure of a speech recognition method performed in the multi-premises device 100 and / or the personal device 200. In FIG.

This speech recognition method according to FIG. 3 may be performed by the multi-premises device 100, for example, using a program that uses each hardware block under the control of the control processor 190 and / or the audio processing processor 180 . Also, the personal device 200 can process the registration (S100) of the voice feature of FIG. 3, for example.

3 are not necessarily sequentially performed. For example, during the process of the control using the voice feature (S200), the registration of the voice feature (S100) and the modification of the voice feature (S300) are performed simultaneously Or may be configured to be performed. It will be appreciated that the process of FIG. 3 is just a deterioration of clock for a clear understanding.

3, a voice is received from a user who wishes to register a voice feature in step S100, and a voice feature of the user is extracted from the received voice. Also, the user identifier of the user is received and mapped through the input unit 110 or the microphone (via voice recognition or the like), and the mapped user's voice characteristic and user identifier are then transmitted to the remote server 300 and / Is stored in the multi-functional apparatus (100) implementing the speech recognition method.

An exemplary detailed description of step S100 will be described in more detail with reference to FIG.

In step S200, a voice is received from a user who intends to control the multi-functional device 100 using the voice feature, and the voice feature is extracted from the received voice. If the extracted voice feature coincides with the stored voice feature, And controls the multi-functional apparatus 100 accordingly.

The stored voice feature may be voice features stored in the memory 140 or hard disk 150 of the multi-user device 100 that the user is using or voice features stored in the remote server 300 .

An exemplary detailed description of step S200 will be described in more detail with reference to FIG.

Then, in step S300, the voice feature extracted from the registered user may also be used to modify the stored voice feature. Accordingly, the voice characteristic of the user can be changed and stored according to the use (frequency) of the multi-functional apparatus 100.

An exemplary detailed description of step S300 will be described in more detail with reference to FIG.

4 is a diagram showing an exemplary process flow for registering user-specific voice features according to S100. This control flow of Fig. 4 is performed by the multi-premises device 100 and / or the personal device 200. [

First, in step S103, a voice feature registration request is received from the user. Such a registration request may be made by receiving a specific key code via the input unit 110 or from receiving a specific control command via a microphone.

In step S105, the user identifier is received from the user through the input unit 110, the microphone, or the like. Such a user identifier may be, for example, a telephone number or a mobile phone number, or a user's ID (identification) for registering in a database of the remote server 300. [ This user identifier allows the user to be uniquely identified in a system environment in which the speech recognition method according to the present invention is implemented.

Of course, the step S105 need not be performed in a predetermined order in time with other steps.

Then, in step S107, one or more sentences are displayed on the display through the video output unit 130 or the like. These one or more sentences are sentences for the purpose of extracting the voice characteristic of the user.

Here, the one or more sentences may include, for example, words used for controlling the multi-functional apparatus 100 or the like. For example, the sentence may include a character or a word assigned to or used by a specific multi-functional device 100 according to speech recognition such as "start ", &

Thereafter, in step S109, the user reads aloud one or more sentences aloud using the microphone, and the voiced speech corresponding to the one or more sentences is received via the microphone and via the analog-to-digital converter 170. [

Here, in step S109, although speech reception through a sentence is mentioned, it need not be limited to a sentence, and may be constituted by voice reception for a plurality of words.

In step S111, a voice feature is extracted from the voice received from the user.

Herein, the voice characteristic may be a voice characteristic extracted for the entire one or more sentences, or a voice characteristic extracted corresponding to a designated character or word. Accordingly, it may be a voice feature derived by taking an average of voice feature values for the entire sentence, a voice feature derived by taking an average of voice feature values for a specified word or character, and the like.

Here, the voice features for each designated character or word may also be stored separately, and the voice features for each designated character or word may be used for controlling the multi-premises device 100.

In addition, in addition to the extraction of speech features, each designated character or word is identified using a commonly known speech recognition technique, and in accordance with an acoustic model in a speech recognition technique corresponding to the specified character or word, For example, acoustic model parameters (for example, frequency characteristic vectors) to be used may be extracted for each user by using FFT or the like.

The acoustic model parameters corresponding to each of these designated characters or words may also be stored in the multi-premises device 100 together with the user identifier for each user or transmitted to the remote server 300 and coupled with the user identifier to the remote server 300 Lt; / RTI >

Such acoustic model parameters may further be used to easily identify corresponding control commands from each user or parameters (e.g., numerical numbers, etc.) included in the control commands.

The speech feature corresponding to the user is composed of a combination of frequency parameters on the frequency domain and / or time parameters on the time domain, for example.

For example, the frequency parameter may be a user-specific pitch frequency of the user in the frequency domain from the user's voice and / or a loudspeaker frequency for determining the tone of the user.

Such a pitch frequency may be different for each user and the vocal frequency may be a noise frequency that can be different between each user who is guided to the vocal cords through inhalation or exhalation during vocalization. These pitch frequencies and loudspeaker frequencies may further include other frequency characteristics for identifying the voice of a user other than the other, so that each user can be easily identified as a voice map of each user.

This frequency parameter may be determined by analyzing the frequency converted with the frequency domain transform (e.g., FFT (Fast Fourier Transform), Cosine Transform) for the converted digital audio stream from the received user voice Can be obtained.

For example, the time parameter may be the magnitude (or intensity) of the voice, the amplitude (or range) of the voice, the voice rhythm determined from the level of the voice (change in size), and / or the speed of voice.

The time parameter may be determined from the digital audio stream continuously received in time. For example, the size of the voice may be determined by the magnitude value of each audio data in the digital audio stream (e.g., the quantized output value from the analog-to-digital converter 170) or the average thereof, Or a mean value thereof. The voice rhythm is a variation pattern of the magnitude value of audio data in a sentence or a word, and the speed of voice can be determined by a time interval between each character and a following character.

Of course, these exemplary frequency parameters or time parameters are only examples for identifying a particular user, and in addition to that, different frequency features or time domain features may be used to identify the user.

And such time and frequency parameters may be included as a component of a voice feature for filtering speech.

For example, the voice feature may include one or more frequency parameters, or may include one or more temporal parameters, or may include (in combination) the one or more frequency and time parameters.

The inclusion of these multiple parameters makes it easier and more precise to identify a particular user and also allows the comparison range of each parameter to be compared over a wider range.

Thereafter, in step S113, the extracted voice feature is mapped to the received user identifier. In step S115, the voice feature and the user identifier mapped in step S115, or in addition to the designated character or word for control of the multi- The acoustic model parameters of each user according to the acoustic model for recognition are stored in the hard disk 150 or the memory 140 and the registration is completed.

Where only the user identifier and voice features can be configured to be used for registration of the user, thereby preventing exposure of certain important information to the user.

The acoustic model parameters corresponding to the mapped speech features and the user identifiers or in addition to the designated characters or words for control of the multi-functional apparatus 100 and corresponding to the acoustic models for speech recognition are stored in the multi- (S117) to the remote server 300 connected via the Internet or the like so that it can be used in the other multi-functional apparatuses 100 or the like.

5 is a diagram showing an exemplary process flow for controlling the apparatus using speech recognition according to S200.

And receives an identifier of the user through the input unit 110 or the microphone in step S203 according to a start request of speech recognition using the voice feature of the user through the input unit 110 or the like (i.e., a voice recognition request for each user). And determines a voice feature corresponding to the user identifier.

Accordingly, the multi-premises equipment 100 determines a plurality of corresponding voice features in response to a start request of voice recognition of a plurality of users, and enables only voice recognition of the user corresponding to the plurality of voice features.

The determined voice features are voice characteristics stored in the memory 140 or the hard disk 150 of the multi-functional apparatus 100 performing the step S200 or may be stored in the remote server 300). ≪ / RTI >

Of course, if the voice feature is not present in the multi-user device 100 or the remote server 300, the user can request the registration process S100.

Then, in step S205, the microphones and the analog-to-digital converter 170 receive voice from the user.

Then, the speech feature is extracted from the speech received in step S207. Such speech features may include frequency parameters and / or time parameters.

Then, in step S209, it is searched whether the extracted voice feature is stored in the multi-functional apparatus 100. [

This step S209 searches to recognize if there is a voice feature extracted from the determined one or more voice features corresponding to the one or more user identifiers that requested the start of voice recognition.

Of course, the one or more voice features may be voice features already stored in the multi-premises device 100, or dynamically received from the remote server 300 in response to a start request of the user's voice recognition in step S203, 100 or the memory 140 or the hard disk 150 of the mobile terminal 100. [

The retrieval of the extracted voice feature is performed only for a plurality of voice features for which a start request for voice recognition is made according to, for example, step S203, so that a simple search can identify a specific user.

Alternatively, an alternative may be to search for a plurality of voice features already stored in the multi-premises device 100 without using the user's identifier (i.e., without performing step S203).

The search for speech features may be performed by comparing each parameter of the extracted speech feature (on the frequency and / or time domain) with each of the parameters included in the plurality of speech features, for example, The threshold range of each corresponding parameter of the plurality of speech features (for example, + 5% to -5% with respect to the parameter, this threshold range may vary depending on the characteristics of each parameter per parameter) Can be determined.

Even if the user is the same user, the voice characteristic can be changed according to the specific time and the specific situation. Therefore, the user can be easily identified regardless of such a change simply by using the threshold range, and also using a plurality of parameters, Or to identify the corresponding voice features of this user.

Then, in step S211, it is determined whether or not the voice feature exists in, for example, a plurality of voice features for which a start request for voice recognition is made. If the voice feature is not present, the voice is not subjected to voice recognition conversion and the process proceeds to step S205 .

On the other hand, if there is a voice feature, the process proceeds to S213.

In step S213, the text corresponding to the received voice is determined. These texts may constitute one sentence or may consist of one or more words or letters.

The determination of such speech as text is accomplished using well-known speech recognition conversion techniques. For example, the speech recognition conversion technique converts a received speech into a frequency band through an FFT or the like, and converts the received speech into a corresponding word, character, or sentence using a given acoustic model.

Here, the voice characteristic corresponding to the extracted voice characteristic or the extracted voice characteristic of the user and stored in the multi-functional apparatus 100 can also be used for determination with this text.

Accordingly, the speech recognition conversion technology enables the text to be recognized more accurately by using the extracted or stored speech features. This cognitive conversion technique can be performed in the frequency domain and each individual character or word can be determined by further reflecting the individual voice characteristic.

(Or in addition to) the speech recognition conversion technique may be applied to each of the user's voices to control the multi-user device 100, which may be stored in the multi-user device 100 or received with voice features from a remote server Acoustic model parameters corresponding to the determined designated character or word may be used.

These designated characters or words constitute the control command used in the multi-functional apparatus 100 or the input parameters of the control command, and accordingly, the corresponding character or word is used for matching between words or characters in an acoustic model for speech recognition conversion Acoustic model parameters can be used so that it can be easily converted from speech containing control commands to text in a simple match with reduced performance load.

Here, conversion or determination into text is important because it is important to precisely determine the control command and input parameters required in the multi-premises equipment 100, so that the accuracy of the other words or characters may not be a problem.

Then, in step S215, the control command is identified in the determined text. These control commands may be words such as "start", "end", "reservation", "OK", "select", etc., which are used to control the multi-

After identifying each of these control commands, the multi-functional apparatus 100 is controlled in accordance with the respective control commands in step S217.

Here, a specific control command may require, for example, a series of numbers provided as input parameters of the control command. Accordingly, it is possible to identify the words preceding or following the specific control command in the text, and to control the multi-functional apparatus 100 using the identified word as an input parameter.

6 is a diagram illustrating an exemplary process flow for modifying a voice feature in accordance with S300. The step S300 may be performed separately from the step S200, and may be performed using the processing result of the step S200.

First, in step S303, if a voice feature exists in step S211, the user identifier corresponding to the retrieved voice feature and voice feature extracted in step S207 is temporarily stored in the memory 140 or the like.

Such temporary storage may be configured to distinguish by user identifiers and may comprise one or more extracted voice features for each user. Such one or more extracted voice features may be the case where voice features are extracted more than once according to a time difference in a single multi-functional apparatus 100.

Then, using the temporarily stored user identifier in step S305, a difference between the temporarily stored one or more voice features and the corresponding voice feature stored in the multi-premises device 100 is determined

These differences can be calculated differently depending on the frequency parameters or time parameters included in each speech feature.

Where the difference may be determined by the difference between the average value of one or more temporally stored speech features and the corresponding value of the stored speech feature.

In step S307, the stored voice characteristic corresponding to the user identifier is corrected by reflecting the difference, and the modified voice characteristic is stored in the memory 140 or the hard disk 150 of the multi-functional device 100. [

This modification of the speech feature enables more accurate speech recognition than later.

Then, in step S309, this modified voice feature is transmitted with the user identifier to the remote server 300 so that it can be used in other multi-functional apparatuses 100. [

As can be seen from FIGS. 3 to 6, the present invention enables voice recognition for only a designated user, enables voice recognition using voice characteristics of each user, improves the accuracy of voice recognition, Thereby preventing a malfunction due to unacceptable noise.

Further, the present invention eliminates unnecessary conversion to text with simple filtering, allows only a voice feature corresponding to a user identifier to be shared, and utilizes a low-specification voice recognition technology.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the invention. The present invention is not limited to the drawings.

100: Multi-functional device
110: input unit 120: audio output unit
130: video output unit 140: memory
150: Hard disk 160: Microphone connection port
170: analog-to-digital converter 180: audio processing processor
190: Control processor
200: Personal device 300: Remote server

Claims (7)

A speech recognition method in a device for a plurality of users,
(a) extracting a voice feature for identifying a user from a received voice from a user;
(b) retrieving whether the extracted voice feature is present in the voice feature stored in the device; And
(c) if the extracted voice feature is present, determining a text corresponding to the user's voice,
Further comprising: prior to step (a), receiving a user identifier from one or more users,
Wherein step (b) comprises: (b-1) determining a voice feature corresponding to a user identifier; And (b-2) searching for a speech feature extracted from the determined one or more speech features.
Speech recognition method. The method according to claim 1,
Wherein the voice feature comprises a combination of a frequency parameter determined by converting a user's voice into a frequency domain and a time parameter determined from the user's voice in a time domain,
Speech recognition method. delete A speech recognition method in a device for a plurality of users,
(a) extracting a voice feature for identifying a user from a received voice from a user;
(b) retrieving whether the extracted voice feature is present in the voice feature stored in the device; And
(c) if the extracted voice feature is present, determining a text corresponding to the user's voice,
Prior to step (a)
Receiving a registration request for a voice feature from a user;
Extracting a voice feature from a voice received from a user to determine a voice feature corresponding to the user;
Mapping extracted voice features to a user identifier received from a user; And
Further comprising registering the user by storing the mapped voice features and the user identifier,
Speech recognition method. 5. The method of claim 4,
And transmitting the mapped speech feature and user identifier to a remote server connected to the device,
Wherein the step (b) comprises receiving and storing a voice feature corresponding to a user identifier received from a user for dynamic voice recognition from the remote server and searching for the presence of the extracted voice feature,
Speech recognition method. The method according to claim 1,
(d) identifying a control command from the determined text; And
(e) controlling the device in accordance with the identified control command.
Speech recognition method. A speech recognition method in a device for a plurality of users,
(a) extracting a voice feature for identifying a user from a received voice from a user;
(b) retrieving whether the extracted voice feature is present in the voice feature stored in the device; And
(c) if the extracted voice feature is present, determining a text corresponding to the user's voice,
Determining a difference between the extracted speech characteristic and the corresponding stored speech characteristic when the extracted speech characteristic is present; And modifying the corresponding stored voice feature using the determined difference.
Speech recognition method.

Images