DEVICE AND METHOD FOR SPEECH RECOGNITION

Abstract

A speech recognition device includes a user prompt playback unit configured to change a user prompt inducing an utterance for speech recognition into a sound source and play the sound source. The device further includes a microphone detection signal extraction unit configured to extract a microphone detection signal when a user's speech is received as a speech signal by a microphone. The device also includes a user prompt removal unit configured to remove the user prompt from the microphone detection signal. The device further includes a speech recognition unit configured to recognize a speech based on a value from which the user prompt is removed.The device also includes a response output unit configured to output a response related to the speech based on a result of recognizing the speech.

Description

TECHNICAL FIELD

[0001] The invention relates to a device and a method for speech recognition. BACKGROUND OF THE INVENTION

[0002] The content described below provides background information related to the present embodiment and does not constitute prior art.

[0003] Generally, when speech recognition begins with a voice device in response to a request from a person, i.e., a speaker in a vehicle, all sounds in the vehicle are muted, and silence builds up inside. If other passengers are in the vehicle, they may look at the speaker and hear the speaker's speech. Most passengers, including the speaker, may find this situation uncomfortable.

[0004] The speaker may feel uncomfortable or anxious about speaking and may wonder, "What should I do if my speech isn't recognized while driving?" Therefore, the time the speaker spends talking to the voice device may be reduced. Furthermore, instructions may be uncomfortable spoken by the speaker, resulting in a low recognition rate. As a result, the speaker may stop using speech recognition technology in vehicles, and technologies developed at considerable expense may be left unused.

[0005] Additionally, there are other factors that make it difficult to use existing speech recognition technology. However, there is a lack of development to improve the speech recognition experience. Only new technologies are competitively developed based on the people who use speech recognition, rather than using existing technologies. Ultimately, to achieve good speech recognition performance, a large amount of data related to users' speech needs to be collected. However, as the number of customers who do not use the technology increases, the newly created features may not reflect diverse speech patterns, and performance may eventually reach a threshold limit. SUMMARY

[0006] Against this background, the invention provides a device and a method for speech recognition based on user tastes and preferences.

[0007] The invention provides a device and a method for providing a user-friendly speech environment.

[0008] The problems to be solved by the invention are not limited to those mentioned above. Other problems not mentioned here will be more clearly understood by those skilled in the art from the following description.

[0009] According to one embodiment of the invention, a device for speech recognition comprises a user prompt reproduction unit configured to change a user prompt inducing an utterance for speech recognition into a sound source and play the sound source. The device further comprises a microphone detection signal extraction unit configured to extract a microphone detection signal when a user's speech is received as a speech signal by a microphone. The device also comprises a user prompt removal unit configured to remove the user prompt from the microphone detection signal. The device further comprises a speech recognition unit configured to recognize a speech based on a value from which the user prompt is removed.The device also includes a response output unit configured to output a response related to the speech based on a result of recognizing the speech.

[0010] According to one embodiment of the invention, a method for speech recognition comprises changing a user prompt that induces an utterance for speech recognition into a sound source and playing the sound source. The method further comprises extracting a microphone detection signal when a user's speech is received as a speech signal by a microphone. The method also comprises removing the user prompt from the microphone detection signal. The method further comprises recognizing a speech based on a value from which the user prompt is removed. The method also comprises outputting a response related to the speech based on a result of recognizing the speech.

[0011] The invention can improve existing rigid, e.g., inflexible, speech recognition experiences or speech recognition that customers find unpleasant and uncomfortable.

[0012] The invention can provide an experience that can reduce the user's concerns about speech recognition by adding the user's preferences to the beginning and end of speech recognition (recognition results).

[0013] In the invention, various speech patterns can be collected by enhancing the user experience, which can lead to an efficient development cycle in which the speech recognition performance improves.

[0014] The invention can provide a faster speech recognition service by reducing unnecessary waiting times.

[0015] The features of the invention are not limited to the features mentioned above. Other features not mentioned will be more clearly understood by those skilled in the art from the following description. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram of a speech recognition device according to an embodiment of the invention. Fig. 2 is a block diagram of a speech recognition device according to another embodiment of the invention. Fig. 3 is a flowchart of a speech recognition device according to an embodiment of the invention. Fig. 4 is a flowchart of a speech recognition device according to another embodiment of the invention. DETAILED DESCRIPTION

[0016] Various embodiments of the invention will now be described in detail with reference to the accompanying drawings. In the following description, like reference numerals refer to like elements even though the elements are shown in different drawings. Furthermore, in the following description of the embodiments, a detailed description of known functions and configurations included therein has been omitted for clarity and brevity.

[0017] Additionally, terms such as first, second, A, B, (a), (b) may be used to describe components of the present invention. These terms are intended only to distinguish one component from another, and the type, order, or sequence of components is not limited by the terms. Throughout the specification, whenever any part is intended to "include" or "comprise" any component, it is intended to include other components and not to exclude other components unless expressly stated to the contrary. Additionally, terms such as "part," "module," and the like throughout the specification refer to a unit that handles at least one function or operation, which may be implemented in hardware or software, or a combination of hardware and software.When a controller, component, device, element, part, unit, module, or the like of the present invention is described as having a purpose or performing an operation, function, or the like, the controller, component, device, element, part, unit, or module should be considered "configured" herein to fulfill that purpose or perform that operation or function. Each controller, component, device, element, part, unit, module, and the like may separately embody or be included as part of the device with a processor and memory, such as a non-transitory computer-readable medium.

[0018] In this description, “sound source”, “music” and “sound” have the same meaning and are used interchangeably.

[0019] In this description, a Push-to-Talk (PTT) service refers to a service that enables the transmission of voice messages by pressing a specific key for communication.

[0020] In this specification, a music prompt or a user prompt refers to changing a prompt, such as a user beep sound or a "say a command" sound, that causes a user to formulate a voice recognition command, i.e., requests input from the user, to a sound source selected by the user or stored / recorded to play the same.

[0021] Embodiments of the invention may be applied to a sound processing technology that includes a user experience (UX) that provides users with an environment for using speech recognition in a vehicle.

[0022] The user environments to which an embodiment of the invention can be applied are as follows. 1 The user can set a “user prompt” and a “user background sound source” through applications that can connect to a vehicle (e.g., Blue Link, KIA Connect, and the like) or by setting a voice recognition of the vehicle. ② When setting the vehicle's voice recognition, a default sound source provided by the vehicle can be selected. 3 A voice selection via an application may include a function for transmitting user-generated and user-recorded sound sources to the vehicle by the user, including a selection of a default sound source.

[0023] Fig. 1 is a block diagram of a speech recognition device according to an embodiment of the invention.

[0024] With reference to Fig.1, the speech recognition device according to an embodiment of the invention may include a user prompt playback unit 101, a microphone detection signal extraction unit 102, a user prompt removal unit 103, a voice / speech recognition unit 104, a response output unit 105, and a controller 106. Although not shown in the drawings, the speech recognition device according to an embodiment of the invention may include at least one microphone (MIC) capable of receiving a user's speech including a speech signal, i.e., a voice signal. The user prompt playback unit 101 plays a sound source previously selected by the user when sound source detection begins (PTT or wake-up).In other words, the user prompt playback unit 101 may change a prompt, such as a user beep sound or a "say a command" sound, that prompts the user to utter a command for voice recognition, into a sound source selected by the user or stored / recorded as shown in a database of Table 1 to play the same. [Table 1] DB ID user-selected sound source Media properties 1 Good News (Mac Milller) Rd1 (News) 2 Good weather (Sandol) Rd2(Weather) 3 Today's Happiness (Kim Hyung-joong) Rd3(luck) 4 Server-linked music source Rdn(x) 5 Vivaldi Four Seasons (Spring) Classical MP

[0025] The database of Table 1 can store sound sources in various ways other than the sound sources selected or stored / recorded by the user, and is not limited to a specific format. The database can provide standard music sources that can be uploaded from a terminal application or selected by the user. In addition, the user prompt playback unit 101 can select and use a plurality of music prompts for each user when changing the prompt that prompts the user to utter an utterance for speech recognition to a sound source selected or stored / recorded by the user to play the same.

[0026] Methods for storing sound sources selected or stored / recorded by the user may include a method for sampling and using the user's favorite music, a method for recording narration by the user, and a method for creating and using the user's own prompts.

[0027] The user prompt playback unit 101 can provide a user-friendly utterance environment by allowing the user to change the prompt, such as the existing beep sound or the "say the command" sound, that prompts the user to utter an utterance for speech recognition to a sound source selected or stored / recorded by the user for use. By using these sound sources, the beginning and end of an utterance can be more precisely specified by changing the existing non-specific sound (wind noise or road noise) to a specific sound (user-selected sound source).

[0028] Since the sound source played by the user prompt playback unit 101 is a sound source already known to the vehicle and a timing of playback is accurately detected, the sound source can be classified as a predictable sound different from general ambient noise and road noise.

[0029] When the microphone detection signal extraction unit 102 receives the user's utterance as a speech signal via the vehicle's microphone, the microphone detection signal extraction unit 102 may extract a microphone detection signal (En+Mp+Vs) including wind noise or road noise En flowing into the vehicle, a "user prompt" (Mp), and an utterance (Vs) selected by a beginning of speech (BOS) / end of speech (EOS) extraction unit (not shown). The BOS / EOS extraction unit may extract a starting or ending point of the speaker's utterance. Specifically, the BOS / EOS extraction unit may monitor and determine a time at which a sound of Mp input to the microphone is generated as the beginning of speech BOS.When no speech is detected for about 0.8 seconds (when the sound of Mp lasts for 0.8 seconds or longer), the BOS / EOS extraction unit can determine that the end of speech EOS has been reached.

[0030] Because a music prompt is used, a sound input (music prompt) is already known during speech recognition. Therefore, the BOS and EOS can be measured more accurately when the user utters a command.

[0031] Additionally, when a speech comes into the microphone, if an input with another specific timing that is already known is introduced, it can be set as BOS. When only a signal from the sound source is input to the microphone, it can be set as EOS. Accordingly, a speech start time and a speech end time can be defined more precisely than using a previous speech recognition device. Therefore, unnecessary waiting time (the time to wait to see if the user is not speaking) is reduced because the speech recognition device does not classify unspecified noise and speech, but specific noise and speech, so a determination time is inevitably fast. As a result, a faster speech recognition service is provided by the speech recognition device.

[0032] The microphone detection signal extraction unit 102 can provide a user-friendly utterance environment by allowing the user to change the prompt that prompts the user to utter a command for speech recognition, such as an existing beep sound or the "say the command" sound, to a sound source selected by the user or stored / recorded to use the same.

[0033] The user prompt removal unit 103 can remove the user prompt from the microphone detection signal (En + Mp + Vs). Since the user prompt uses vehicle sound sources, the user prompt can be easily removed.

[0034] The speech recognition unit 104 may recognize a speech based on the value (En + Vs) from which the user prompt has been removed.

[0035] The response output unit 105 may transmit a response related to the speech to the terminal or the server based on a result of speech recognition by the speech recognition unit 104.

[0036] The controller 106 may control each of the user prompt playback unit 101, the microphone detection signal extraction unit 102, the user prompt removal unit 103, the speech recognition unit 104, and the response output unit 105 for speech recognition according to an embodiment of the invention.

[0037] The user prompt playback unit 101, the microphone detection signal extraction unit 102, the user prompt removal unit 103, the voice / speech recognition unit 104, the response output unit 105, and the controller 106 may be implemented by one or more hardware components of a computer system (i.e., a computer processor, a microprocessor) that performs the operations described above.

[0038] Fig. 2 is a block diagram of a speech recognition device according to another embodiment of the invention.

[0039] The speech recognition device according to another embodiment of the invention may comprise a selection domain extraction unit 107, a text-to-speech (TTS) result acquisition unit 108, a keyword extraction unit 109, a sound source acquisition unit 110, a mixing unit 111, and a result output unit 112 in addition to the components of Fig. 1 include.

[0040] The selection domain extraction unit 107 may extract a domain value (Rdn) and a speech recognition result value from the result of speech recognition by the speech recognition unit 104.

[0041] The TTS result acquisition unit 108 may acquire a result value (Rt) output as text from the speech recognition result value of the selection domain extraction unit 107.

[0042] The speech recognition result can be achieved by executing commands and transmitting information. Among these, in the case of a domain that transmits the result as a TTS, among the results corresponding to the transmission of information, the vehicle can provide a suitable environment for information delivery by allowing the user to manually adjust background sounds (music, etc.) for each domain, or can automatically adjust background sounds based on user experience (e.g., usage history, etc.), deviating from the rigid information transmission method of transmitting only voice sounds.

[0043] The keyword extraction unit 109 may extract a keyword from the result value (Rt) output as text and transmit the extracted keyword to the server.

[0044] The sound source acquisition unit 110 may acquire a sound source Rdn(x) corresponding to the keyword extracted from the server. A recommended sound source corresponding to the keyword may be extracted from the server and stored in the database of Table 1 as Rdn(x).

[0045] The sound source acquisition unit 110 may acquire a background sound for each domain from the database of Table 1 when a domain has the result value (Rt) output as text from the result of recognizing the speech.

[0046] The mixing unit 111 can mix the sound source Rdn(x) corresponding to the keyword extracted from the server, the user prompt (Mp), and the result value (Rt) output as text, and output them. Media sources set for each domain can be selected based on a domain value, as shown in the database of Table 1.

[0047] The result output unit 112 can output the result value of the mixing unit 111.

[0048] The Controller 106 from Fig. 1 can control each of the selection domain extraction unit 107, the TTS result acquisition unit 108, the keyword extraction unit 109, the sound source acquisition unit 110, the mixing unit 111, the result output unit 112, and the like for speech recognition according to an embodiment of the invention.

[0049] The selection domain extraction unit 107, the TTS result acquisition unit 108, the keyword extraction unit 109, the sound source acquisition unit 110, the mixing unit 111, the result output unit 112 may be implemented by one or more hardware components of a computer system (ie, a computer processor, a microprocessor) that performs the operations described above.

[0050] Fig.3 is a flowchart of a speech recognition device according to an embodiment of the invention.

[0051] The voice recognition device plays a sound source selected by the user at a time (PTT or wake-up) when the sound source recognition begins in step S301. In other words, the voice recognition device can change the prompt, such as a user beep sound or a "say a command" sound, that prompts the user to utter a voice recognition command to a sound source selected by the user or stored / recorded to play the same, as shown in a database of Table 1.

[0052] In step S302, when the speech recognition device receives the user's utterance as a speech signal through the vehicle's microphone, the speech recognition device may extract a microphone detection signal (En+Mp+Vs) including wind noise or road noise (En) flowing into the vehicle, a "user prompt" (Mp), and an utterance (Vs) selected by a BOS / EOS extraction unit (not shown). The BOS / EOS extraction unit may extract a start or end point of the speaker's utterance. Specifically, the BOS / EOS extraction unit may monitor and determine a time at which a sound of Mp input to the microphone is generated as the start of speech (BOS). When no speech is detected for about 0.8 seconds (when the sound of Mp continues for 0.8 seconds or longer), the BOS / EOS extraction unit may determine that the end of speech (EOS) has been reached.The content of the utterance (Vs) can be expressed as Equation 1 and can include the content of the speech from the beginning of the speech to the end of the speech. Vs(Language)Content=BOS(t)∼EOS(t)

[0053] The speech recognition device may remove the user prompt from the microphone detection signal (En + Mp + Vs) in step S303.

[0054] The speech recognition device may recognize a speech based on the value (En + Vs) from which the user prompt is removed in step S304.

[0055] The speech recognition device may transmit a response regarding the speech to the terminal or the server based on the result of recognizing the speech in step S305.

[0056] Fig. 4 is a flowchart of a speech recognition device according to another embodiment of the invention.

[0057] After step S305 of Fig. 3, the speech recognition device may extract a domain value (Rdn) and a speech recognition result value from the speech recognition result in step S401.

[0058] The speech recognition device may acquire a result value (Rt) output as text from the speech recognition result value in step S402.

[0059] The speech recognition device may extract a keyword from the result value (Rt) output as text in step S403 and transmit the extracted keyword to the server.

[0060] The speech recognition device may detect a sound source Rdn(x) corresponding to the keyword extracted from the server in step S404. A recommended sound source corresponding to the keyword extracted from the server may be stored in the database of Table 1 as Rdn(x).

[0061] The speech recognition device may combine the sound source (Rdn(x)) corresponding to the keyword extracted from the server, the user prompt (Mp), and the result value (Rt) output as text, and output them in step S405. Media sources set for each domain may be selected based on a domain value, as shown in the database of Table 1.

[0062] The speech recognition device may output the mixed result value in step S406.

[0063] The operation of Fig. 4 is further described as follows.

[0064] In the case of a domain with (Rt), a method may include selecting a background sound for each domain from the database of Table 1.

[0065] The method for selecting a background sound includes a method for selecting from a standard sound source or a method for sending a user-created / downloaded / purchased sound source to the server to be downloaded and used in the vehicle.

[0066] As mentioned above, in addition to the method for the user to select the background sound, a method may include providing a random option to store multiple sound sources for each domain and use them randomly. Alternatively, in the case of a weather domain, a method may include extracting the weather type from text and extracting a background sound source appropriate for the weather. In the case of the weather domain, a value for the weather type (this value may also be used to display a weather icon in the vehicle) may be used. If the weather type is "a little rain," background music matching the weather result may be searched for and downloaded from a streaming server for immediate playback.For example, upon receiving an utterance such as "Inform me about the results of the baseball game," a keyword "baseball" can be extracted and sent to a music server. A sound source matching "baseball" can be searched for and sent to the vehicle. A background sound source can be stored so that the corresponding sound source can be mixed with the baseball scores and played back.

[0067] By incorporating user tastes and preferences, embodiments of the invention provide users with a perception that speech recognition is a technology that can be conveniently used. With increasing usability, various speech patterns can be collected and reflected in the recognition performance, thereby achieving an efficient development cycle structure.

[0068] In the drawings described above, although the processes are described as being performed sequentially, they merely illustrate the technical spirit of some embodiments of the invention. In other words, one skilled in the art to which various embodiments of the invention relate can variously modify and apply the invention by changing and performing the processes described in the flowchart or by performing one or more of the processes in parallel without deviating from an intrinsic characteristic of some embodiments of the invention. Thus, the drawings are not limited to a time-series sequence.

[0069] The Fig. 3 and Fig.The processes shown in Figure 4 may be implemented as computer-readable code on a non-transitory computer-readable recording medium. The computer-readable recording medium may include any type of recording device that stores data readable by a computer system. That is, such a computer-readable recording medium includes non-transitory media such as ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices. In addition, computer-readable recording media may be distributed among a computer system connected by a network, and computer-readable code may be stored and executed in a distributed manner.

[0070] Additionally, the components of the present invention may utilize integrated circuit structures such as memory, a processor, logic circuitry, a lookup table, and the like. These integrated circuit structures perform each function described herein under the control of one or more microprocessors or other control devices. Additionally, the components of the present invention may be specifically implemented by a program or portion of code that includes one or more computer-executable instructions for performing specific logical functions and is executed by one or more microprocessors, processors, or other control devices. Additionally, the components of the present invention may include or be implemented by a central processing unit (CPU), a microprocessor, and the like that perform the respective functions thereof.Additionally, the components of the present invention may store instructions executed by one or more processors in one or more memories.

Claims

[1] Device for speech recognition, the device comprising: a user prompt playback unit configured to change a user prompt inducing an utterance for speech recognition into a sound source and play the sound source; a microphone detection signal extraction unit configured to extract a microphone detection signal when a user's speech is received as a speech signal by a microphone; a user prompt removing unit configured to remove the user prompt from the microphone detection signal; a speech recognition unit configured to recognize a speech based on a value from which the user prompt is removed; and a response output unit configured to output a response related to the speech based on a result of recognizing the speech. [2] The device according to claim 1, wherein the microphone detection signal includes a user prompt (Mp), a sound introduced into a vehicle (En) and contents of speech (Vs). [3] Device according to claim 2, wherein the noise (En) introduced into the vehicle includes wind noise and / or road noise. [4] The device according to claim 2, wherein the contents of speech (Vs) include the contents of the speech from the beginning of the speech to the end of the speech. [5] Device according to claim 1, further comprising: a selection domain extraction unit configured to extract a domain value (Rdn) and a speech recognition result value from the result of recognizing the speech; a text-to-speech (TTS) result acquisition unit configured to acquire a result value (Rt) output as text from the speech recognition result value; a keyword extraction unit configured to extract a keyword from the result value (Rt) and transmit the extracted keyword to a server; a sound source acquisition unit configured to acquire a sound source (Rdn(x)) corresponding to the keyword extracted from the server; a mixing unit configured to mix a sound source (Rdn(x)) corresponding to the keyword extracted from the server, the user prompt (Mp), and the result value (Rt); and a result output unit configured to output a result of the mixing unit. [6] The apparatus according to claim 5, wherein the sound source detecting unit is further configured to detect a background sound for each domain from a database when a domain has the result value (Rt) output as text from the result of recognizing the speech. [7] The device according to claim 6, wherein the sound source corresponding to the domain is pre-stored in the database. [8] A method for speech recognition, the method comprising: Changing a user prompt that induces a speech recognition utterance into a sound source and playing the sound source; Extracting a microphone detection signal when a user's speech is received as a speech signal by a microphone; Removing the user prompt from the microphone detection signal; Detecting a language based on a value from which the user prompt is removed; and Outputting a response related to the language based on a result of recognizing the language. [9] The method of claim 8, wherein the microphone detection signal includes the user prompt (Mp), noise introduced into a vehicle (En) and contents of speech (Vs). [10] A method according to claim 9, wherein the noise (En) introduced into the vehicle includes wind noise and / or road noise. [11] The method according to claim 9, wherein the contents of speech (Vs) include the content of the speech from the beginning of the speech to the end of the speech. [12] The method of claim 8, further comprising: Extracting a domain value (Rdn) and a speech recognition result value from the result of recognizing the speech; capturing a result value (Rt) output as text from the speech recognition result value; Extracting a keyword from the result value (Rt) output as text and transmitting the extracted keyword to a server; Detecting a sound source (Rdn(x)) corresponding to the keyword extracted from the server; Mixing the sound source Rdn(x) corresponding to the keyword extracted from the server, the user prompt (Mp) and the result value (Rt) output as text; and Output a mixed result. [13] The method of claim 12, wherein detecting the sound source Rdn(x) further comprises detecting a background sound for each domain from a database when a domain has the result value (Rt) output as text from the result of recognizing the speech. [14] The method of claim 13, wherein the sound source corresponding to the domain is pre-stored in the database. [15] A non-transitory computer-readable recording medium storing a computer program comprising computer-executable instructions configured to be executable by a speech recognition device comprising a processor to cause the processor to perform: a function that includes changing a user prompt that induces an utterance for speech recognition into a sound source and playing the sound source; a function comprising extracting a microphone detection signal when a user's speech is received as a speech signal by a microphone; a function comprising removing the user prompt from the microphone detection signal; a function that includes recognizing a language based on a value from which the user prompt is removed; and a function that includes outputting a response related to the speech based on a result of recognizing the speech. [16] The computer-readable recording medium according to claim 15, wherein the microphone detection signal includes the user prompt (Mp), sound introduced into a vehicle (En), and contents of speech (Vs). [17] A computer-readable recording medium according to claim 16, wherein the noise (En) introduced into the vehicle includes wind noise and / or road noise. [18] The computer-readable recording medium according to claim 16, wherein the contents of speech (Vs) include the contents of the speech from the beginning of the speech to the end of the speech. [19] The computer-readable recording medium of claim 15, further comprising computer-executable instructions to cause the processor to perform: a function for extracting a domain value (Rdn) and a speech recognition result value from the result of recognizing the speech; a function for capturing a result value (Rt) output as text from the speech recognition result value; a function for extracting a keyword from the result value (Rt) and transmitting the extracted keyword to a server; a function to capture a sound source (Rdn(x)) corresponding to the keyword extracted from the server; a function for mixing the sound source (Rdn(x)) corresponding to the keyword extracted from the server, the user prompt (Mp) and the result value (Rt); and a function to return a result of the mixing function. [20] The computer-readable recording medium of claim 19, wherein the function of detecting a sound source (Rdn(x)) further comprises computer-executable instructions for causing the processor to directly detect a background sound for each domain from a database when a domain has the result value (Rt) output as text from the result of recognizing the speech.

Images