KR20160110085A - Speech recognition system and method thereof - Google Patents

Abstract

The present disclosure may provide technology for effectively recognizing a speech by using a personalized wakeup keyword. For this, a device detects a wakeup keyword from the speech signal of a user received by using a personalized wakeup keyword model, transmits whether the wakeup keyword is detected or a received speech signal to a speech recognition server. A speech recognition server sets a speech recognition model according to whether the wakeup keyword is detected and can perform a recognition process on the speech signal of the user. So, an accurate speech recognition function can be provided.

Description

[0001] SPEECH RECOGNITION SYSTEM AND METHOD THEREOF [0002]

The present disclosure relates to a speech recognition system and method, and more particularly, to a speech recognition system and method based on a wake up keyword.

As the devices become smart, the device is equipped with a voice recognition function capable of executing the function of the device by using the user's voice signal.

In order to use the voice recognition function mounted on the device, the voice recognition function of the device must be woken up. The existing speech recognition function is being woken up using the fixed wakeup keyword. Thus, when there are a plurality of devices equipped with the same voice recognition function in the same place, the voice recognition function of the unwanted device can be woken up.

In addition, the existing speech recognition function is divided into a wakeup keyword and a voice command. Accordingly, after inputting the wake-up keyword, the user must input a voice command when the voice recognition function of the device wakes up. If the user continuously inputs the wakeup keyword and the voice command, the conventional voice recognition function may not be woken up or wake up, but a recognition error may occur for the inputted voice command.

Accordingly, there is a need for a technique that can more accurately and accurately recognize a voice command while waking up the voice recognition function of the device.

The above-described background information is information that the inventor holds for the purpose of deriving the present disclosure or obtained in the process of deriving the present disclosure and is not necessarily a known technology disclosed to the general public prior to the filing of the present disclosure.

Embodiments of the present disclosure are intended to provide a more convenient and accurate speech recognition function by continuously recognizing personalized wakeup keywords and voice commands.

Embodiments of the present disclosure are also intended to provide a speech recognition function that is more effectively woken up using personalized wakeup keywords.

Embodiments of the present disclosure are also intended to provide a speech recognition function that is more effectively woken up using personalized wakeup keywords according to device-based environmental information.

As a technical means for achieving the above-mentioned technical object, a first aspect of the present disclosure relates to a device, comprising: an audio input for receiving a user's voice signal; A memory for storing a wake-up keyword model; A communication unit capable of communicating with a speech recognition server; And a controller for detecting a wake-up keyword from the user's voice signal using the wake-up keyword model as the voice signal of the user is received through the audio input unit, A processor for transmitting a signal to the speech recognition server via the communication unit, receiving a speech recognition result from the speech recognition server via the communication unit, and controlling the device in accordance with the speech recognition result .

As a technical means for achieving the above technical object, a second aspect of the present disclosure is a speech recognition server comprising: a communication unit capable of communicating with at least one device; A memory for storing a wake-up keyword model and a speech recognition model; And a speech recognition model in which the wake-up keyword model is combined with a detection signal for a wake-up keyword and a voice signal of a user from a device of one of the at least one device through the communication unit, Recognizing the voice signal of the user using the recognition model, removing the wake-up keyword from the recognition result of the user's voice signal, transmitting the recognition result from which the wake-up keyword is removed to the device through the communication unit A voice recognition server, and a voice recognition server.

As a technical means for achieving the above-mentioned technical object, a third aspect of the present disclosure relates to a device for detecting a wake-up keyword from a voice signal of a user; And a voice recognition model in which a wake-up keyword model is combined with a detection signal for the wake-up keyword and a voice signal of the user are received from the device, And a voice recognition server for recognizing the voice recognition result and transmitting the recognition result to the device.

 As a technical means for achieving the above-mentioned technical problem, a fourth aspect of the present disclosure is a method for detecting a wake-up keyword from a voice signal of a user using a wake-up keyword model as a voice signal of a user is received, Transmitting a detection signal for the wake-up keyword and the voice signal of the user to a voice recognition server; Receiving a recognition result of the user's voice signal from the voice recognition server; And controlling the device in accordance with the recognition result.

 As a technical means for achieving the above-mentioned technical object, a fifth aspect of the present disclosure is a method comprising: receiving a detection signal and a voice signal of a user for a wake-up keyword from a device; Setting a speech recognition model according to a detection signal of the wakeup keyword; Recognizing the user's speech signal using the speech recognition model; Removing the wake-up keyword from recognition results of the user's voice signal; And transmitting the recognition result of the user's voice signal from which the wake-up keyword is removed to the device.

As a technical means to achieve the above-mentioned technical problem, a sixth aspect of the present disclosure is a method for detecting a wake-up keyword from a user's voice signal using the wake-up keyword model as a user's voice signal is received via the device, ; Transmitting a detection signal for the wake-up keyword and the user's voice signal from the device to the voice recognition server; Setting a speech recognition model in the speech recognition server in accordance with a detection signal of the wakeup keyword; Recognizing the speech signal of the user using the speech recognition model in the speech recognition server; Removing, at the speech recognition server, the wakeup keyword from recognition results of the user's speech signal; Transmitting a recognition result of the user's voice signal from which the wake-up keyword is removed to the device from the voice recognition server; And controlling, in the device, a device according to the received recognition result.

A seventh aspect of the present disclosure can provide a computer-readable recording medium having recorded thereon a program for causing a computer to execute the method of the fifth aspect.

The eighth aspect of the present disclosure can provide a computer-readable recording medium on which a program for causing a computer to execute the method of the sixth aspect is recorded.

1 is a diagram for explaining a speech recognition system according to some embodiments.
2 is a flowchart illustrating an operation of a device included in the speech recognition system and a speech recognition method based on the speech recognition server according to some embodiments.
3 is a flowchart of an operation of a process of registering a wake-up keyword model in a speech recognition method according to some embodiments.
Fig. 4 is an operational flowchart of another process for registering a wake-up keyword in the speech recognition method according to some embodiments.
5A and 5B are examples in which a candidate wake-up keyword model is displayed on a display of a device included in the speech recognition system according to some embodiments.
6 and 7 are operational flowcharts of a device included in the speech recognition system and a speech recognition method based on the speech recognition server according to some embodiments.
8 is a flowchart illustrating an operation of a speech recognition method by a device according to some embodiments.
9 and 10 are block diagrams of a device included in a speech recognition system according to some embodiments.
11 is a configuration diagram of a speech recognition server included in a speech recognition system according to some embodiments.
12 is a configuration diagram of a speech recognition system according to some other embodiments.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. However, the present disclosure may be embodied in many different forms and is not limited to the embodiments described herein. In order that the present disclosure may be more fully understood, the same reference numbers are used throughout the specification to refer to the same or like parts.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "electrically connected" with another part in between . Also, when an element is referred to as "comprising ", it means that it can include other elements as well, without departing from the other elements unless specifically stated otherwise.

In the present disclosure, the wakeup keyword refers to information capable of waking up the speech recognition function. In this disclosure, the wakeup keyword can be said to be a wakeup word. In this disclosure, the wake-up keyword may be based on the user's voice signal, but in this disclosure the wake-up keyword is not limited thereto. For example, in the present disclosure, the wakeup keyword may include a gesture-based sound (or audio signal) of the user.

A gesture-based sound of a user may include, for example, a sound that a user hits a finger. A gesture-based sound of a user may include, for example, a user's tongue-tearing sound. The user's gesture-based sound may include, for example, the user's laughing sound. The gesture-based sound of the user may include, for example, the sound of the user's lips being pounded. The user's gesture-based sound may, for example, include the user's whistling sound. In this disclosure, the gesture-based sound of the user is not limited to the above-mentioned one.

In the present disclosure, the wake-up keyword may be referred to as a wake-up signal if it includes the gesture-based sound of the user described above.

In this disclosure, the wake-up keyword model refers to a wake-up keyword pre-registered with a device or / and a speech recognition server to detect or recognize a wake-up keyword. The wakeup keyword model may include a personalized acoustic model and / or a language model, but in the present disclosure, the wakeup keyword model is not limited thereto. The acoustic model is a model of the signal characteristics of the user's voice (or the sound of the user's gesture). The language model is a model of the linguistic order of words or syllables corresponding to the recognized vocabulary.

Since the wake-up keyword model registered in the device of the present disclosure is used for detecting the wake-up keyword, it can be said that it is a model for detecting the wake-up keyword. Since the wake-up keyword model registered in the voice recognition server of the present disclosure is used to recognize the wake-up keyword, it can be said that it is a model for recognizing the wake-up keyword.

The model for detecting the wake-up keyword and the model for recognizing the wake-up keyword may be the same, but there may be a difference. For example, when the model for detecting the wake-up keyword includes the acoustic model corresponding to the personalized wakeup keyword Hi, the model for recognizing the wakeup keyword includes, for example, an acoustic model corresponding to the personalized wakeup keyword Hi, (E.g., < RTI ID = 0.0 >!) ≪ / RTI > In the present disclosure, the model for wake-up keyword detection and the model for wakeup keyword recognition are not limited to the above-described ones.

In the following description, a model for detecting a wake-up keyword and a model for recognizing a wake-up keyword are not distinguished from each other, and will be referred to as a wake-up keyword model. However, the wake-up keyword model registered in the device is understood as a model for wake-up keyword detection, and the wake-up keyword model registered in the speech recognition server can be understood as a model for wake-up keyword recognition.

The wakeup keyword model may be generated by the device or the speech recognition server. The device or the speech recognition server can send and receive data to share the generated wakeup keyword model with each other.

In the present disclosure, the speech recognition function refers to converting a user's voice signal into a character string (or text). The user's voice signal may include voice commands. The voice command can execute a specific function of the device.

The specific functionality of the device in this disclosure may include, but is not limited to, for example, executing an application configured in the device.

For example, when the device is a smart phone, running the application may include dialing, navigating, searching the internet, or setting an alarm. For example, in the case where the device is a smart device, executing an application may include searching for a program, searching for a channel, and the like. If the device is a smart oven, running the application may include searching for cooking methods, and the like. If the device is a smart refrigerator, running the application may include checking the refrigeration condition, or checking the refrigeration condition, and the like. When the device is a smart car, running the application may include autostarting, autonomous driving, automatic parking, and the like. Implementation of an application in this disclosure is not limited to the one just described.

Voice instructions in this disclosure may have a word form. In the present disclosure, voice commands may have the form of a sentence. Voice instructions in this disclosure may have a sphere shape. In this disclosure, the speech recognition model may include a personalized acoustic model and / or a language model.

The present disclosure will be described in detail below with reference to the accompanying drawings.

1 is a diagram for explaining a speech recognition system 10 according to some embodiments. The speech recognition system 10 includes a device 100 and a speech recognition server 110.

The device 100 may receive a voice signal from the user 101. [ The device 100 may detect the wakeup keyword from the voice signal of the user 101 that is received using the wakeup keyword model. The device 100 may generate a wake-up keyword model and register it in the device 100. [ The device 100 may transmit the generated wakeup keyword model to the speech recognition server 110. [ The device 100 may receive and use a wake-up keyword model from the speech recognition server 110.

The device 100 may send a detection signal for the wakeup keyword and a voice signal of the user 101 to the voice recognition server 110. [

The detection / non-detection signal for the wake-up keyword is a signal indicating whether or not the wake-up keyword has been detected from the voice signal of the user 101 to be received. The device 100 may represent the detection signal for the wakeup keyword in binary data. When the wake-up keyword is detected from the voice signal of the received user 101, the device 100 can express the detection signal for the wake-up keyword by, for example, '0'. If the wakeup keyword is not detected from the voice signal of the received user 101, the device 100 can express the detection signal for the wakeup keyword as '1', for example.

The voice signal of the user 101 received by the device 100 may include a wakeup keyword and a voice command. The voice signal of the user 101 received by the device 100 may not include the wakeup keyword.

The speech recognition server 110 may receive a detection signal for the wakeup keyword from the device 100 and a voice signal of the user 101. [ The voice signal of the user 101 received from the device 100 is the same as the voice signal of the user 101 received by the device 100. [

The speech recognition server 110 can set the speech recognition model according to the detection signal of the received wake-up keyword. If the detection signal for the wake-up keyword indicates that the voice signal of the user 101 includes the wake-up keyword, the voice recognition server 110 uses the voice recognition model combined with the wake- The speech recognition model can be set to recognize the speech signal.

A wakeup keyword model coupled to the speech recognition model at the speech recognition server 110 may follow the wakeup keyword detected at the device 100. [ For example, when the wakeup keyword detected in the device 100 is " Hi ", the speech recognition server 110 determines that the speech recognition model (e.g., play the music) Quot; can be used to set the speech recognition model to recognize the speech signal of the user 101. [ When combining the wakeup keyword model and the speech recognition model, the speech recognition server 110 may consider the silence duration between the wakeup keyword model and the speech recognition model.

As described above, the speech recognition server 110 continuously performs the recognition processing on the voice command and the wake-up keyword model included in the voice signal of the user, thereby stably obtaining the voice signal of the user and improving the voice recognition performance .

If the detection signal for the wake-up keyword indicates that the voice signal of the user 101 does not include the wake-up keyword, the voice recognition server 110 uses the voice recognition model that does not incorporate the wake- 101 can recognize the speech signal of the speech recognition model.

In this manner, the speech recognition server 110 can dynamically reconfigure (or switch) the speech recognition model used for recognizing speech according to the detection signal for the wakeup keyword. Accordingly, it can be said that setting the speech recognition model according to the detection / non-detection signal of the wake-up keyword performed in the speech recognition server 110 determines the configuration of the speech recognition model according to whether or not the wake-up keyword is detected.

Setting the speech recognition model in the speech recognition server 110 may include loading the speech recognition model. Accordingly, the detection signal for the wakeup keyword may be interpreted as including a voice recognition model loading request signal, a voice recognition model setting request signal, or a voice recognition model loading trigger signal. In this disclosure, the expression for the detection signal for the wake-up keyword is not limited to those described above.

The speech recognition server 110 may generate a speech recognition model for recognizing speech commands. The speech recognition model may include an acoustic model and a language model. The acoustic model refers to modeling the signal characteristics of speech. The language model is a model of a linguistic order relation such as a word or syllable corresponding to a recognition vocabulary.

The speech recognition server 110 can detect only the speech portion from the speech signal of the user 101 that is received. The speech recognition server 110 may extract the speech feature from the detected speech portion. The speech recognition server 110 may perform a speech recognition process on the received speech signal of the user 101 using the extracted speech features, features of the pre-registered acoustic model, and language model. The speech recognition server 110 can perform speech recognition processing by comparing features of the extracted speech features and pre-registered acoustic models. The speech recognition processing of the received voice signal of the user 101 performed by the voice recognition server 110 is not limited to the above-described one.

The speech recognition server 110 may remove the wakeup keyword from the speech recognition result that has performed the speech recognition processing. The voice recognition server 110 can transmit the voice recognition result to the device 100 from which the wake-up keyword has been removed.

The speech recognition server 110 may generate a wake-up keyword model. The speech recognition server 110 can register the generated wake-up keyword model in the speech recognition server 110 and transmit it to the device 100. [ Accordingly, the device 100 and the speech recognition server 110 can share the wake-up keyword model.

The device 100 may control the function of the device 100 according to the voice recognition result received from the voice recognition server 110. [

The device 100 or the speech recognition server 110 can assign identification information for each wake-up keyword model when there are a plurality of wake-up keyword models generated by the device 100 or the speech recognition server 110 have. When the identification information is assigned to each wakeup keyword model, the detection signal for the wakeup keyword transmitted from the device 100 to the voice recognition server 110 may include identification information for the detected wakeup keyword have.

In the case where the device 100 is a portable device, the device 100 may be a smart phone, a notebook, a smart board, a tablet PC, a handheld ) Device, a handheld computer, a media player, an electronic book device, and a device such as a PDA (Personal Digital Assistant), but the device 100 in this disclosure is not limited to the one just described.

In the case where the device 100 is a wearable device, the device 100 may be a smart phone, a smart watch, a smart band (e.g., a smart waistband and a smart hairband), various smart ornaments Smart smart fingers, smart smart pins, smart clips, and smart necklaces), various smart body protectors (eg, smart knee protectors, and smart elbow protectors). The device 100 may include at least one of a smart shoe, a smart glove, a smart garment, a smart hat, a smart prosthesis, or a smart shoe, but the device 100 in this disclosure is not limited to the one just described.

The device 100 may include devices such as M2M (Machine to Machine) or devices based on Internet of Things (IoT) networks (e.g., smart appliances, smart sensors), automobiles, and automotive navigation devices The device 100 in this disclosure is not limited to the one just described.

The device 100 and the voice recognition server 110 may be connected based on a wired or wireless network. The device 100 and the voice recognition server 110 may be connected to a short-range wireless network and / or a long-range wireless network.

2 is a flowchart illustrating an operation of a speech recognition method performed on the basis of the device 100 and the speech recognition server 110 included in the speech recognition system 10 according to some embodiments. Fig. 2 shows a case where voice recognition is performed based on the voice signal of the user 101. Fig.

Referring to FIG. 2, in step S201, the device 100 registers a wake-up keyword model. 3 is a flowchart of an operation of a process of registering a wake-up keyword model in a speech recognition method according to some embodiments.

Referring to FIG. 3, in step S301, the device 100 receives a voice signal of the user 101. FIG. The voice signal of the user received in step S301 is for registering the wake-up keyword model. In step S301, the device 100 may receive a sound (or an audio signal) based on the specific gesture of the user 101 described above instead of the voice signal of the user 101. [

In step S302, the device 100 can recognize the user's voice signal using the voice recognition model. The speech recognition model may include an acoustic model and / or a language model based on ASR (Automatic Speech Recognition), but the speech recognition model is not limited to this in the present disclosure.

If it is determined in step S303 that the voice signal of the received user 101 is valid as a wake-up keyword model, the device 100 generates and registers a wake-up keyword model in step S304. Registering the wakeup keyword model in the device 100 may mean storing the wakeup keyword model in the device 100. [

In step S303, the device 100 can determine whether the voice signal of the received user 101 is valid as a wake-up keyword model based on the voice matching rate for the voice signal of the user 101. [

For example, when the device 100 recognizes the voice signal of the user 101 a plurality of times, compares the recognized results, and if the consistent result exceeds a predetermined number of times according to the comparison result, It can be determined that the voice signal of the wakeup keyword is valid as the wake-up keyword model.

If it is determined in step S303 that the voice signal of the received user 101 is valid, the device 100 registers the wake-up keyword model determined to be valid in the device 100 in step S304.

In step S303, the device 100 recognizes the voice signal of the user 101 a plurality of times, and as a result of comparing the recognized results, if the coherent result is less than a predetermined number of times, It can be determined that it is not valid as a wake-up keyword model.

If it is determined in step S303 that the voice signal of the received user 101 is not valid, the device 100 does not register the voice signal of the received user 101 as a wake-up keyword model.

If the voice signal of the received user 101 is determined to be invalid as a wake-up keyword model, the device 100 can output a notification message. The notification message can have various forms and contents. For example, the notification message may include a message that the voice signal of the currently input user 101 is not registered as a wake-up keyword model. The notification message may include a message that guides the user 101 to input a voice signal that can be registered as a wake-up keyword model.

Fig. 4 is an operational flowchart of another process for registering a wake-up keyword in the speech recognition method according to some embodiments.

In step S401, the device 100 requests a candidate wake-up keyword model stored in the device 100. [ The candidate wakeup keyword model request may be based on the voice signal of the user 101, but the present disclosure is not limited thereto. For example, the device 100 may receive a user input requesting a candidate wake-up keyword model according to a specific button control (or dedicated button) or touch-based input of the device 100.

In step S402, the device 100 outputs a candidate wake-up keyword model. The device 100 may output a candidate wake-up keyword model through the display of the device 100. [

Figs. 5 (a) and 5 (b) are examples of displaying a candidate wake-up keyword model on the display of the device 100 included in the speech recognition system 10 according to some embodiments.

5 (a) is an example of a candidate wake-up keyword model list being displayed on the display of the device 100. Fig. Referring to Fig. 5 (a), a candidate wake-up keyword model is provided in text form.

If a touch-based input to the first candidate wake-up keyword model is received based on the candidate wake-up keyword model list shown in Fig. 5 (a), the device 100 displays the speech waveform for the selected candidate wake- the audio signal corresponding to the candidate wake-up keyword model can be output while being displayed as shown in (b) of FIG. Accordingly, the user can confirm the wake-up keyword model to be selected before selecting the wake-up keyword model.

In step S402, the device 100 may output a candidate wake-up keyword model through the audio output portion (e.g., a speaker) of the device 100. [

When a selection signal for one candidate wakeup keyword model is received based on the candidate wakeup keyword model in step S403, the device 100 registers the selected candidate wakeup keyword model in step S404. In step S404, the device 100 requests the input of the voice signal of the user 101 corresponding to the selected candidate wake-up keyword model, generates the voice signal of the received user 101 as a wake-up keyword model, You can register.

In step S201, the device 100 can receive the wake-up keyword model from the voice recognition server 110 and register it. In step S201, the device 100 establishes a communication channel with the voice recognition server 110, transmits a voice signal of the user 101 received via the set communication channel to the voice recognition server 110, You may request registration. Accordingly, the device 100 can receive the wake-up keyword model generated by the speech recognition server 110. [

On the other hand, in step S202 of FIG. 2, the speech recognition server 110 registers a wake-up keyword model. In step S202, the voice recognition server 110 may register the wake-up keyword model received from the device 100, but the method of registering the wake-up keyword model in the voice recognition server 110 in this disclosure is not limited to the above- It does not.

For example, the speech recognition server 110 may request and receive a wake-up keyword model from the device 100. For this purpose, the voice recognition server 110 may monitor the device 100. [ The voice recognition server 110 may periodically monitor the device 100. [ The voice recognition server 110 may monitor the device 100 when the access to the device 100 is recognized. The voice recognition server 110 may request the device 100 for a wake-up keyword model when it is recognized that the device 100 is connected to the voice recognition server 110. [

In step S202, when the voice recognition server 110 registers the wake-up keyword model, the voice recognition server 110 may add a tag indicating the wake-up keyword to the wake-up keyword. The tag may be represented, for example, by a special symbol (e.g., an exclamation point (!)), But the representation for the tag in this disclosure is not limited to the one just described.

In step S202, the wake-up keyword model registered in the voice recognition server 110 may be synchronized with the wake-up keyword model registered in the device 100. [ When the wake-up keyword model registered in the device 100 is updated, the wake-up keyword model registered in the speech recognition server 110 can be updated.

In step S202, the speech recognition server 110 receives the voice signal of the user 101 from the device 100 and generates and registers a wake-up keyword model. The speech recognition server 110 may generate a wake-up keyword model as described above with reference to FIG. 3 or FIG. The speech recognition server 110 may receive the speech signal of the user 101 to generate a wake-up keyword model from the device 100 prior to step S201.

In step S203, the device 100 can receive the voice signal of the user 101. [ In step S204, the device 100 can detect the wake-up keyword from the voice signal of the received user 101 using the registered wake-up keyword model. The device 100 can detect the wake-up keyword by comparing the signal characteristics between the registered wake-up keyword model and the voice signal of the user 101 to be received.

In step S205, the device 100 may transmit the detection signal for the wake-up keyword and the voice signal of the user 101 to the voice recognition server 110. [

In step S206, the speech recognition server 110 can set the speech recognition model according to the detection whether or not the received wake-up keyword is detected. The speech recognition model setting may be as described in FIG. That is, when the detection signal for the wake-up keyword indicates that the wake-up keyword is detected, the voice recognition server 110 can set the voice recognition model combined with the wake-up keyword model. The speech recognition server 110 can set the speech recognition model to which the wake-up keyword model is not combined when the detection signal for the wake-up keyword indicates that the wake-up keyword is not detected.

In step S207, the voice recognition server 110 can recognize the voice signal of the user 101 that is received using the voice recognition model set. In step S208, the speech recognition server 110 can remove the wakeup keyword from the speech recognition result. The speech recognition server 110 can remove the wakeup keyword from the speech recognition result by using the tag added to the wakeup keyword when registering the wakeup keyword model.

In step S209, the speech recognition server 110 can transmit the speech recognition result from which the wake-up keyword is removed to the device 100. [ In step S210, the device 100 can control the device according to the received speech recognition result.

6 is a flowchart illustrating an operation of a voice recognition method performed on the basis of the device 100 and the voice recognition server 110 included in the voice recognition system 10 according to some embodiments. Fig. 6 is an example of performing speech recognition using a wake-up keyword model according to environment information based on the device 100. Fig.

In step S601, the device 100 can register a plurality of wakeup keyword models based on the environment information. The environment information may include location information. The location information may include physical location information and logical location information. Physical location information refers to information expressed in latitude and longitude. Logical location information refers to information represented by semantic information such as home, office, or cafe. Environmental information may include weather information. The environmental information may include time information. Environmental information may include schedule information. The environmental information may include location, time, weather, and / or schedule information. In the present disclosure, the environmental information is not limited to the above-described one, and may include condition information or situation information that directly or indirectly affects the user 101. [

For example, the device 100 can register the wake-up keyword model differently when the position of the device 100 is the home and the wake-up keyword model when the position of the device 100 is the office. The device 100 may register the wake-up keyword model differently when the time detected by the device 100 is 6 am and the wake-up keyword model when the time detected by the device 100 is 6 pm. The device 100 may register the wakeup keyword model differently when the weather detected by the device 100 is clear and the wakeup keyword model differently when the weather detected by the device 100 is rainy. The device 100 may register another wakeup keyword model according to the schedule of the user 101 detected by the device 100. [

In step S601, the device 100 can receive and register a plurality of wakeup keyword models based on environmental information from the speech recognition server 110 as in step S201 described above.

In step S602, the speech recognition server 110 can register a plurality of wake-up keyword models based on the environment information.

A plurality of wakeup keyword models registered in the speech recognition server 110 can be synchronized in real time with a plurality of wakeup keyword models registered in the device 100. [ Therefore, each time a plurality of wake-up keyword models registered in the device 100 are updated, a plurality of wake-up keyword models registered in the speech recognition server 110 can be updated.

In step S602, the voice recognition server 110 can register a plurality of wake-up keyword models received from the device 100. [ In step S602, the speech recognition server 110 may request a plurality of wake-up keyword models from the device 100 to receive a plurality of wake-up keyword models from the device 100. [

In step S602, the voice recognition server 110 sets a communication channel between the device 100 and the voice recognition server 110 as in the above-described step S202, A plurality of wakeup keyword models based on the environmental information described above can be generated and registered based on the voice signal of the user 101. [ The speech recognition server 110 can provide the device 100 with a plurality of wakeup keyword models thus registered.

In step 603, the device 100 may receive the voice signal of the user 101. [ In step S604, the device 100 can detect environment information based on the device 100. [ The device 100 can detect environmental information based on the device 100 using the sensors included in the device 100 or an application set in the device 100. [

For example, the device 100 may detect position information using a position sensor (e.g., a Global Positioning System (GPS) sensor) included in the device 100. The device 100 can detect the time information using the timer application set in the device 100. [ The device 100 may detect weather information using the weather application set in the device 100. [ The device 100 can detect the schedule of the user 101 using the fixed application set in the device 100. [

In step S605, the device 100 can detect the wakeup keyword from the voice signal of the received user 101 using the wake-up keyword model corresponding to the detected environment information among the plurality of registered wake-up keyword models .

For example, when the wake-up keyword model at the home is " Hi " and the wake-up keyword model at the office is " Good ", the device 100 If the location is an office, the device 100 can detect the wakeup keyword from the voice signal of the user 101 received using " Good. &Quot;

In step S606, the device 100 may transmit the detected environment information, a detection whether or not to detect the wakeup keyword, and a voice signal of the user 101 to be received to the voice recognition server 110. [

In step S607, the voice recognition server 110 determines a wake-up keyword model in accordance with the detection signal for the wake-up keyword and the environment information based on the received device 100, and the determined wake- You can set the model.

In step S608, the speech recognition server 110 can recognize the received speech signal using the set speech recognition model. In step S609, the speech recognition server 110 can remove the wakeup keyword from the speech recognition result. The speech recognition server 110 can remove the wakeup keyword contained in the speech recognition result by using the tag added to the wakeup keyword when registering the wakeup keyword model.

In step S610, the voice recognition server 110 may transmit the voice recognition result to the device 100 from which the wake-up keyword has been removed. In step S611, the device 100 can control the device 100 according to the received voice recognition result.

7 is a flowchart illustrating an operation of a voice recognition method performed on the basis of the device 100 and the voice recognition server 110 included in the voice recognition system 10 according to some embodiments. 7 shows an example in which speech recognition is performed by setting a speech recognition model in accordance with identification information of the user 101, environment information based on the device 100, and a wakeup keyword detection signal.

In step S701, the device 100 can register a plurality of wake-up keyword models based on the environment information. The environment information may be as described in step S601 of Fig. 6, but is not limited thereto. In step S701, the device 100 can register a plurality of wake-up keyword models received from the voice recognition server 110. [

In step S702, the voice recognition server 110 can register a plurality of wake-up keyword models based on the environment information and the identification information of the user 101. [ For example, the voice recognition server 110 can register a plurality of wake-up keyword models based on the environment information with respect to the identification information A of the user 101. [ The speech recognition server 110 can register a plurality of wakeup keyword models based on the environment information with respect to the identification information B of the user 101. [

The plurality of wake-up keyword models registered in the voice recognition server 110 can be synchronized on a user-by-user basis. For example, when a plurality of wake-up keyword models of the user A are updated, a plurality of wake-up keyword models of the user A among the plurality of wakeup keyword models registered in the speech recognition server 110 can be updated.

In step S702, the voice recognition server 110 can register the above-described wake-up keyword model based on the voice signal of the user 101 received from the device 100. [ In such a case, the voice recognition server 110 may provide a wake-up keyword model registered with the device 100. [

In step 703, the device 100 may receive the voice signal of the user 101. [ In step S704, the device 100 can detect the environment information based on the device 100. [ In step S705, the device 100 can acquire the identification information of the user 101 based on the voice signal of the user 101 to be received. The identification information of the user 101 may include a nickname, a sex, a name, etc. of the user 101, but the identification information of the user 101 in this disclosure is not limited to the above.

Step S705 may be configured to acquire the identification information of the user 101 using the fingerprint recognition of the user 101 or the iris recognition technique of the user 101. [

In step S706, the device 100 can detect the wakeup keyword from the voice signal of the received user 101 using the wakeup keyword model corresponding to the detected environment information among the plurality of registered wakeup keyword models .

In step S707, the device 100 transmits the detected environment information, the identification information of the user 101, the detection / non-detection signal for the wakeup keyword, and the voice signal of the received user 101 to the voice recognition server 110 .

In step S708, the voice recognition server 110 determines a wake-up keyword model in accordance with the detection signal for the wake-up keyword, the environment information based on the received device 100, and the identification information of the user 101, You can set up a speech recognition model that combines a wakeup keyword model.

In step S709, the speech recognition server 110 can recognize the received speech signal using the set speech recognition model. In step S710, the speech recognition server 110 may remove the wakeup keyword from the speech recognition result. The speech recognition server 110 can remove the wakeup keyword contained in the speech recognition result by using the tag added to the wakeup keyword when registering the wakeup keyword model.

In step S711, the voice recognition server 110 can transmit the voice recognition result from which the wake-up keyword is removed to the device 100. [ In step S712, the device 100 can control the device 100 according to the received speech recognition result.

FIG. 8 is a flow chart of the operation of the speech recognition method by the device 100 according to some embodiments. FIG. 8 shows a case where the device 100 performs voice recognition regardless of the voice recognition server 110. FIG.

In step S801, the device 100 can register the wake-up keyword model. When registering, the device 100 may add a tag capable of identifying the wake-up keyword to the wake-up keyword. In step S801, the device 100 can receive and register the wake-up keyword model from the voice recognition server 110. [

In step S802, the device 100 can receive the voice signal of the user 101. [ In step S803, the device 100 can detect the wakeup keyword from the voice signal of the user 101 using the wake-up keyword model.

If it is determined in step S804 that the wakeup keyword has been detected, the process proceeds to step S805 where the device 100 can set the voice recognition model in which the wakeup keyword model is combined. In step S806, the device 100 can perform speech recognition processing on the received speech signal of the user 101 using the speech recognition model.

In step S807, the device 100 may remove the wakeup keyword from the speech recognition result. The device 100 may remove the wake-up keyword from the speech recognition result using a tag that identifies the wake-up keyword. In step S808, the device 100 can control the device 100 using the voice recognition result from which the wake-up keyword has been removed.

If it is determined in step S804 that the wakeup keyword is not detected, the process advances to step S809, and the device 100 can set a speech recognition model to which the wakeup keyword model is not combined. In step S810, the device 100 can perform recognition processing on the voice signal of the user 101 using the voice recognition model. In step S811, the device 100 can control the device 100 using the voice recognition result.

Fig. 8 can be modified to recognize a voice signal by registering a plurality of wake-up keyword models based on environmental information as in Fig. 6 described above.

2, 6, 7, or 8 described above can be modified to recognize a voice signal by registering a plurality of wake-up keyword models regardless of environmental information. A plurality of wake-up keyword models may be set for each user. When a plurality of wakeup keyword models are registered, each wakeup keyword model may include identification information capable of identifying wakeup keywords.

9 is a functional block diagram of a device 100 according to some embodiments.

9, the device 100 includes an audio input 910, a communication unit 920, a processor 930, a display 940, a user input 950, and a memory 960.

The audio input unit 910 can receive the audio signal of the user 101. [ The audio input unit 910 can receive a sound (audio signal) based on the specific gesture of the user 101 described above.

The audio input unit 910 can receive an audio signal input from the outside of the device 100. The audio input unit 910 converts the received audio signal into an electrical audio signal and transmits the electrical audio signal to the processor 930. The audio input unit 910 may be configured to perform an operation based on various noise reduction algorithms for eliminating noise generated in receiving an external sound signal. The audio input unit 910 may be a microphone.

The communication unit 920 may be configured to be connected between the voice recognition server 110 and the device 100 in a wired or wireless manner. The communication unit 920 may be configured as the communication unit 1040 described later with reference to FIG.

The processor 930 may be referred to as a controller that controls the operation of the device 100. [ The processor 930 may control an audio input 910, a communication unit 920, a display 940, a user input 950, and a memory 960. When the voice signal of the user 101 is received through the audio input unit 910, the processor 930 can perform speech recognition processing using the wake-up keyword model in real time.

Processor 930 may register the wakeup keyword model in memory 960. [ The processor 930 can register the wake-up keyword model received from the speech recognition server 110 via the communication unit 920 in the memory 960. [ The processor 930 may request a wake-up keyword model based on the voice signal of the user 101 while transmitting the voice signal of the user 101 received through the audio input unit 910 to the voice recognition server 110. [

The processor 930 may transmit the wake-up keyword model registered in the memory 960 to the voice recognition server 110 via the communication unit 920. [ The processor 930 can transmit the registered wakeup keyword model to the voice recognition server 110 when the wakeup keyword model request signal is received from the voice recognition server 110 through the communication unit 920. [ The processor 930 may transmit the wake-up keyword model registered with the voice recognition server 110 at the same time that the wake-up keyword model is registered in the memory 960. [

The processor 930 receives a wakeup keyword from the voice signal of the user 101 received using the wakeup keyword model registered in the memory 960 as the voice signal of the user 101 is received through the audio input 910, Can be detected. The processor 930 may transmit the detection signal for the wakeup keyword and the voice signal of the user 101 to the voice recognition server 110 through the communication unit 920. [

The processor 930 may receive the voice recognition result from the voice recognition server 110 through the communication unit 920. [ The processor 930 may control the device 100 according to the received speech recognition result.

When an audio signal for registering the wake-up keyword model is received through the audio input unit 910, the processor 930 determines whether or not the audio signal is an audio signal based on the matching rate for the audio signal, as in step S303 of FIG. It can be determined whether or not it can be used as a wake-up keyword model.

The processor 930 may register the candidate wakeup keyword model selected from the candidate wakeup keyword keywords stored in the memory 960 in the memory 960 in accordance with the user input received through the user input unit 950. [

The processor 930 may be divided into a main processor and a sub-processor depending on the configuration of the device 100. [ The sub-processor may be configured as a low-power processor.

Display 940 may be controlled by processor 930 to display the candidate wakeup keyword requested by user 101. [ The display 940 may be a liquid crystal display, a thin film transistor-liquid crystal display, an organic light-emitting diode, a flexible display, a three-dimensional display display, or an electrophoretic display (EPD). The display 940 may include, for example, a touch screen, but this disclosure does not limit the construction of the display 940 just as described above.

The user input 950 may receive user input to the device 100. The user input 950 receives a user input indicating a wakeup keyword registration request, a user input selecting one of the candidate wakeup keywords, and / or a user input indicating a registration for a selected candidate wakeup keyword . The user input received via the user input 950 in this disclosure is not limited to the one just described. The user input 950 transmits the received user input to the processor 930.

The memory 960 may store a wakeup keyword model. The memory 960 may store programs for processing and control of the processor 930. The program stored in the memory 960 may include an OS (Operating System) program and various application programs. Various application programs may include speech recognition applications, camera applications, etc., in accordance with embodiments of the present disclosure. The memory 960 may store information managed by the application program (e.g., history information of the user 101's wakeup keyword, schedule information of the user 101, or user profile information).

Programs stored in the memory 960 may be classified into a plurality of modules according to their functions. The plurality of modules may include, for example, a mobile communication module, a Wi-Fi module, a Bluetooth module, a DMB module, a camera module, a sensor module, a GPS module, a video playback module, an audio playback module, , ≪ / RTI > and / or application modules.

The memory 960 may be a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory), a RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory) Disk, or optical disk type storage media.

10 is a block diagram of a device 100 in accordance with another embodiment of the present invention.

10, the device 100 includes a sensor unit 1010, a user interface unit 1020, a memory 1030, a communication unit 1040, an image processing unit 1050, an audio output unit 1060, a camera 1070 ), And a processor 1090.

The device 100 may include a battery. The battery may be embedded in the device 100 in a form that is embedded in the device 100 or removable. The battery can supply power to all components contained in the device 100. The device 100 may receive power from an external power supply (not shown) through the communication unit 1040. [ The device 100 may further include a connector connectable to an external power supply.

The processor 1090 shown in Fig. 10, the display 1021 and the user input unit 1022, the memory 1030, the audio input unit 1070, and the communication unit 1040 included in the user interface unit 1020 are shown in Fig. May be referred to as similar or identical components to the illustrated processor 930, audio input 910, communication unit 920, display 940, user input 950, and memory 960.

Programs stored in the memory 1030 can be classified into a plurality of modules according to their functions. For example, the programs stored in the memory 1030 may be classified into the UI module 1031, the notification module 1032, and the application module 1033, but the present disclosure is not limited thereto. For example, programs stored in the memory 1030 may be classified into a plurality of modules as mentioned in the memory 960 of FIG.

The UI module 1031 includes GUI information for registering a wake-up keyword for speech recognition, GUI information indicating a speech recognition result (e.g., text information), GUI information indicating a speech recognition waveform, etc. To the processor 1090. The processor 1090 may display on the display 10210 a screen based on GUI information received from the UI module 1032. [ The UI module 1031 can provide the processor 1090 with a UI and / or GUI specialized for each application installed in the device 100. [

The notification module 1032 can generate a notification based on the voice recognition of the device 100, a notification based on the registration of the wakeup keyword, a notification based on the input of the false wakeup keyword, or a notification based on the wakeup keyword recognition, 1032 are not limited to the above-mentioned ones.

The notification module 1032 can output a notification signal in the form of a video signal through the display 1021 and output a notification signal in the form of an audio signal through the audio output unit 1060. However,

The application module 1033 may include various applications, including speech recognition applications, as described in the embodiments of this disclosure.

The communication unit 1040 is a unit for communication between the device 100 and at least one external device (e.g., voice recognition server 110, smart TV, smart clock, smart mirror, and / or IoT network- And may include the above components. For example, the communication unit 1040 may include at least one of the local communicator 1041, the mobile communication device 1042, and the broadcast receiver 1043, but the components included in the communication unit 1040 are not limited thereto .

A short-range wireless communicator 1041 may be a Bluetooth communication module, a Bluetooth low energy (BLE) communication module, a near field communication unit (RFID) module, a WLAN communication module, a Zigbee ) Communication module, an Ant + communication module, a WFD (Wi-Fi Direct) communication module, a beacon communication module, or an UWB (ultra wideband) communication module. For example, the local communicator 7451 may include an infrared (IRDA) communication module.

The mobile communication device 1042 can transmit and receive a radio signal with at least one of a base station, an external device, and a server on a mobile communication network. Here, the wireless signal may include various types of data depending on a voice call signal, a video call signal, or a text / multimedia message transmission / reception.

The broadcast receiver 1043 can receive broadcast signals and / or broadcast-related information from outside via a broadcast channel. The broadcast channel may include, but is not limited to, at least one of a satellite channel, a terrestrial channel, and a radio channel.

The communication unit 1040 can transmit at least one information generated by the device 100 to at least one external device or receive information transmitted from at least one external device according to a preferred embodiment.

The sensor unit 1010 includes a proximity sensor 1011 that detects whether the user 101 is approaching the device 100 or a biosensor (or a health sensor (E.g., a heartbeat sensor, a blood flow sensor, a diabetes sensor, a blood pressure sensor, and / or a stress sensor) 1012, an illuminance sensor 1013 (or an optical sensor, LED sensor) A mood sensor 1014 for sensing the mood of the user 101 of the device 100, a motion detection sensor 1015 for sensing activity, a position for detecting the position of the device 100, A gyroscope sensor 1017 for measuring the azimuth angle of the device 100, a tilt and an acceleration of the device 100 about the ground surface, etc. An accelerometer sensor 1018 that measures the acceleration of the device 100, The north, south, east, and west, and the like geomagnetic sensor 1019 for sensing the orientation, but the sensor included in the sensor unit 1010 in the present disclosure is not limited just above.

For example, the sensor unit 1010 may be a temperature sensor, a gravity sensor, an altitude sensor, a chemical sensor (e.g., Odorant sensor), an air pressure sensor, a fine dust measuring sensor, (E.g., a network sensor based on WiFi, Bluetooth, 3G, Long Term Evolution (LTE), or / and Near Field Communication (NFC), etc.) But is not limited to.

The sensor unit 1010 may include a pressure sensor (e.g., a touch sensor, a piezoelectric sensor, a physical button, etc.), a status sensor (e.g., an earphone terminal, a Digital Multimedia Broadcasting (DMB) A terminal for recognizing whether or not the charging progress is recognized, a terminal for recognizing whether or not a PC (Personal Computer) is connected, a terminal for recognizing whether a dock is connected), and / or a time sensor. It is not limited.

The sensors included in the sensor portion 1010 may include fewer sensors than those shown in FIG. For example, the sensor unit 1010 may include only the position sensor 1016. [ In the case where only the position sensor 1016 is included in the sensor unit 1010, the sensor unit 1010 may be represented by a GPS receiver.

The result (or sensing value) sensed by the sensor unit 1010 is transmitted to the processor 1090. When the sensing value received from the sensor unit 1010 is a value indicating the position, the processor 1090 can determine whether the current position of the device 100 is the home or office based on the received sensing value.

The processor 1090 may operate as a controller that controls the overall operation of the device 100. [ For example, the processor 1090 may include a sensor unit 1010, a memory 1030, a user interface unit 1020, an image processing unit 1050, an audio output unit 1060, The audio input unit 1070, the camera 1080, and / or the communication unit 1040 and the like.

Processor 1090 may operate as processor 930 of FIG. For the operation of reading data from the memory 1030 in the processor 1090, the processor 1090 can perform operations to receive data from the external device via the communication unit 1040. [ For the operation of writing data from the processor 1090 to the memory 1030, the processor 1090 may perform an operation of transmitting the data to the external device through the communication unit 1040. [

The processor 1090 may perform at least one of the operations referred to in Figures 2, 3, 4, 6, 7, or 8 described above. The processor 1090 may be referred to as a controller that controls the above-described operations.

The image processing unit 1050 can process the image data received from the communication unit 1040 or stored in the memory 1030 so that the image data can be displayed on the display 1021.

The audio output unit 1060 can output the audio data received from the communication unit 1040 or stored in the memory 1030. The audio output unit 1060 can output a sound signal (e.g., a notification sound) related to a function performed by the device 100. [

The audio output unit 1060 may include, but is not limited to, a speaker, a buzzer, and the like.

11 is a functional block diagram of speech recognition server 110 according to some embodiments.

11, the voice recognition server 110 includes a communication unit 1110, a processor 1120, and a memory 1130, but the configuration of the voice recognition server 110 is not limited to that shown in FIG. That is, the components included in the speech recognition server 110 may be more or less than the components shown in FIG.

The communication unit 1110 may be configured as the communication unit 1040 shown in FIG. The communication unit 1110 can transmit and receive a voice recognition related signal to the device 100. [

The processor 1120 may perform the operations of the speech recognition server 110 mentioned in FIG. 2, FIG. 6, or FIG.

The memory 1130 stores the wakeup keyword model 1131 and the speech recognition model 1132 and controls the wakeup keyword model 1131 and the speech recognition model 1132 by the processor 1120 ). The speech recognition model 1132 can be referred to as a model for recognizing a voice command.

The wakeup keyword model 1131 and the speech recognition model 1132 stored in the memory 1130 can be updated according to the information received through the communication unit 1110. [ The wake-up keyword model 1131 and the speech recognition model 1132 stored in the memory 1130 can be updated by an operator's information input. For this purpose, the voice recognition server 110 may further include a component for allowing an operator to input information.

12 is a block diagram of a speech recognition system 1200 according to some other embodiments. 12 is a case where the voice recognition server 110 recognizes a voice signal of the user 101 received from a plurality of devices 100, 1210, 1220, 1230, 1240,

The plurality of devices may include a wearable glass 1210, a smart watch 1220, an IoT device 1230, an IoT sensor 1240, and / or a smart TV 1250.

The plurality of devices 100, 1210, 1220, 1230, 1240, and 1250 described above may be the same as the user, but may be different from each other. When the users of the plurality of devices 100, 1210, 1220, 1230, 1240 and 1250 are the same, the voice recognition server 110 can perform a voice recognition function by registering a wake-up keyword model for each device. When the users of the plurality of devices 100, 1210, 1220, 1230, 1240 and 1250 are different from each other, the voice recognition server 110 uses the device identification information and the device- So that the voice recognition function can be performed. Accordingly, the speech recognition system 1200 of the present disclosure can provide a more versatile and accurate speech recognition service. The speech recognition server 110 may provide the registered wakeup keyword model to the plurality of devices 100, 1210, 1220, 1230, 1240 and 1250. [

In addition, the speech recognition server 110 recognizes the wake-up keyword and the voice command in order to recognize the wake-up keyword and the voice command in the vicinity of the plurality of devices 100, 1210, 1220, 1230, 1240, The noise level can be estimated or environment information can be recognized. The speech recognition server 110 provides the above-described estimated noise level or recognized environment information to the plurality of devices 100, 1210, 1220, 1230, 1240, and 1250 together with the speech recognition result, 1210, 1220, 1230, 1240, 1250, or may provide the user with estimated or perceived information.

The network 1260 may be configured as a wired or / and wireless network. 1260 and 1250 and the server 110 based on at least one of the communications mentioned in the communication unit 1040 shown in Fig. It can transmit and receive.

The above-described methods shown in Figs. 2, 3, 4, 6, 7, or 8 can be implemented by a computer program. For example, the operation of the device 100 in Fig. 2 described above may be performed by a voice recognition application installed in the device 100. [ The operation of the speech recognition server 110 shown in Fig. 2 may be performed by a speech recognition application installed in the speech recognition server 110. [ The computer program described above may be operated in an operating system environment installed in the device 100. [ The computer program described above may be operated in an operating system environment installed in the speech recognition server 110. [ The device 100 can write the above-described computer program onto a storage medium, read from the storage medium, and use the computer program. The speech recognition server 110 can write the above-described computer program onto a storage medium, read from the storage medium, and use the computer program.

An embodiment of the present disclosure may also be embodied in the form of a recording medium including instructions executable by a computer, such as program modules, being executed by a computer. Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. In addition, the computer-readable medium can include both computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Communication media typically includes any information delivery media, including computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave, or other transport mechanism.

It is to be understood that the foregoing description of the disclosure is for the purpose of illustration only and that those skilled in the art will readily understand that various changes in form and details may be made therein without departing from the spirit and scope of the disclosure will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

The scope of the present disclosure is defined by the appended claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

Claims (21)

In a device,
An audio input unit for receiving a user's voice signal;
A memory for storing a wake-up keyword model;
A communication unit capable of communicating with a speech recognition server; And
Detecting a wake-up keyword from the voice signal of the user using the wake-up keyword model as the voice signal of the user is received through the audio input unit,
A detection signal for the wake-up keyword and a voice signal of the user are transmitted to the voice recognition server through the communication unit,
Receives a speech recognition result from the speech recognition server through the communication unit, and
And a processor for controlling the device in accordance with the speech recognition result. 2. The apparatus of claim 1, wherein the device
Further comprising an environmental information sensing unit for sensing environmental information based on the device,
Wherein the wake-up keyword model includes a plurality of wake-up keyword models based on environment information,
The processor comprising:
A device for detecting a wake-up keyword from a voice signal of a user using a wake-up keyword model corresponding to environment information based on the device detected through the environment information sensing unit among the plurality of wake-up keyword models. 3. The device of claim 2, wherein the environment information includes location information of the device. 3. The apparatus of claim 2,
Acquiring identification information of the user based on a user's voice signal received through the audio input unit,
And transmits the obtained identification information of the user to the voice recognition server through the communication unit. The method of claim 1, further comprising: when an audio signal for registering the wake-up keyword model is received through the audio input unit,
The processor comprising:
And determines whether the audio signal is valid as the wake-up keyword model based on a matching rate for the audio signal. 2. The apparatus of claim 1, wherein the device
Further comprising a user input for receiving user input,
The processor comprising:
And registers the candidate wake-up keyword model selected from the candidate wake-up keyword models stored in the memory in accordance with a user input received through the user input unit. 2. The apparatus of claim 1,
Up keyword model from the speech recognition server via the communication unit,
And registers the received wake-up keyword model in the memory. 2. The apparatus of claim 1,
Up keyword model based on a voice signal of a user received through the audio input unit,
And registers the generated wake-up keyword model in the memory. A voice recognition server comprising:
A communication unit capable of communicating with at least one device;
A memory for storing a wake-up keyword model and a speech recognition model; And
A speech recognition model in which the wake-up keyword model is combined with a detection signal for a wake-up keyword and a voice signal of a user are received from one of the at least one device through the communication unit,
Recognizes the user's voice signal using the set speech recognition model,
Removing the wake-up keyword from a voice recognition result of the voice signal of the user,
And a processor for transmitting the voice recognition result from which the wake-up keyword is removed to the device via the communication unit. The method of claim 9, wherein the wake-up keyword model includes a plurality of wake-up keyword models based on environmental information,
The processor comprising:
And a voice recognition model in which a wake-up keyword model corresponding to environment information based on the device is combined among the plurality of wake-up keyword models as the environment information based on the device is received through the communication unit. 10. The apparatus of claim 9,
And a voice recognition model in which a wake-up keyword model associated with the identification information of the user and the environment information based on the device is combined is received as the identification information of the user is received through the communication unit. In a speech recognition system,
A device for detecting a wakeup keyword from a voice signal of a user;
Setting a voice recognition model in which a wake-up keyword model is combined with a detection signal for the wake-up keyword and a voice signal of the user from the device,
And a speech recognition server for recognizing the user's speech signal using the set speech recognition model and transmitting the speech recognition result to the device. Detecting a wake-up keyword from the user's voice signal using a wake-up keyword model as the user's voice signal is received;
Transmitting a detection signal for the wake-up keyword and the voice signal of the user to a voice recognition server;
Receiving a recognition result of the user's voice signal from the voice recognition server; And
And controlling the device in accordance with recognition results of the user's voice signal. 14. The method of claim 13, wherein the wake-up keyword model includes a plurality of wake-up keyword models based on environmental information,
The method comprises:
Detecting environmental information based on the device; And
And transmitting environment information based on the detected device to the voice recognition server,
Wherein detecting the wakeup keyword comprises:
Detecting a wake-up keyword from the user's voice signal using a wake-up keyword model corresponding to environment information based on the detected device among the plurality of wake-up keyword models. 15. The method of claim 14,
Obtaining identification information of the user based on the voice signal of the user; And
And transmitting the obtained identification information of the user to the voice recognition server. Receiving a detection signal and a voice signal of a user for a wakeup keyword from the device;
Setting a speech recognition model according to a detection signal of the wakeup keyword;
Recognizing the user's speech signal using the speech recognition model;
Removing the wake-up keyword from recognition results of the user's voice signal; And
And transmitting to the device a recognition result of the user's voice signal from which the wake-up keyword has been removed. 17. The method of claim 16,
The method may further comprise receiving device-based environment information from the device,
Wherein the step of setting the speech recognition model comprises:
Setting a voice recognition model in which a wake-up keyword model corresponding to environment information based on the device is combined among a plurality of wake-up keyword models based on environment information. 17. The method of claim 16,
Further comprising receiving user identification information from the device,
Wherein the step of setting the speech recognition model comprises:
And setting a speech recognition model in which environmental information based on the device and a wake-up keyword model related to the identification information of the user are combined. Registering a wake-up keyword model in a device and a voice recognition server;
Detecting the wake-up keyword from the user's voice signal using the wake-up keyword model as the user's voice signal is received through the device;
Transmitting a detection signal for the wake-up keyword and the user's voice signal from the device to the voice recognition server;
Setting a speech recognition model in the speech recognition server in accordance with a detection signal of the wakeup keyword;
Recognizing the speech signal of the user using the speech recognition model in the speech recognition server;
Removing, at the speech recognition server, the wakeup keyword from recognition results of the user's speech signal;
Transmitting a recognition result of the user's voice signal from which the wake-up keyword is removed to the device from the voice recognition server; And
And in the device, controlling the device in accordance with the received speech recognition result. A computer-readable recording medium storing a program for causing a computer to execute the method according to any one of claims 13 to 15. A computer-readable recording medium having recorded thereon a program for causing a computer to execute the method according to any one of claims 16 to 18.

Images