KR20190062369A - Speech-controlled apparatus for preventing false detections of keyword and method of operating the same - Google Patents

Abstract

Provided are a voice control device capable of preventing keyword misrecognition and a method for operating the same. According to the present invention, the method comprises the steps of: receiving an audio signal corresponding to a surrounding sound and generating audio stream data; detecting a candidate keyword corresponding to a predefined keyword from the audio stream data and determining a first section corresponding to first audio data in which the candidate keyword is detected from the audio stream data; extracting a first speaker feature vector for the first audio data; extracting a second speaker feature vector for second audio data corresponding to a second section preceding the first section from the audio stream data; and determining whether the keyword is included in the first audio data on the basis of the similarity between the first and second speaker feature vectors and determining whether to perform wake-up.

Description

TECHNICAL FIELD [0001] The present invention relates to a speech control apparatus and a speech control apparatus,

BACKGROUND OF THE INVENTION 1. Field of the Invention [0002] The present invention relates to a voice control apparatus, and more particularly, to a voice control apparatus and an operation method thereof that can prevent a keyword from being erroneously recognized.

Background of the Invention [0002] As performance of computing devices such as portable communication devices, desktops, tablets, and entertainment systems have been enhanced, electronic devices in which speech recognition functions are incorporated and are controlled by voice are being introduced to improve operability. The voice recognition function has an advantage that the device can be easily controlled by recognizing the voice of the user without touching the touch module or by operating a separate button.

According to the voice recognition function, for example, a portable communication device such as a smart phone can perform a calling function or a text message without any operation of pressing a separate button, and can easily perform various functions such as route search, Can be set. However, such a voice control apparatus may misunderstand the voice of the user and cause unintended operation.

One embodiment can provide a voice control apparatus and an operation method thereof that can prevent a keyword from being misrecognized.

According to a first aspect of the present invention, there is provided an audio processing apparatus comprising: an audio processing unit for receiving audio signals corresponding to ambient sounds and generating audio stream data; A keyword detection unit detecting a candidate keyword corresponding to a predefined keyword from the audio stream data and determining a first section corresponding to the first audio data in which the candidate keyword is detected in the audio stream data; Extracting a first speaker feature vector for the first audio data and extracting a second speaker feature vector for second audio data corresponding to a second section preceding the first section in the audio stream data A vector extractor; And a wake-up determination unit for determining whether the keyword is included in the first audio data based on the similarity between the first speaker feature vector and the second speaker feature vector.

Also, a second aspect of the present disclosure is a method comprising: receiving an audio signal corresponding to ambient sound to generate audio stream data; Detecting a candidate keyword corresponding to a predefined keyword from the audio stream data and determining a first section corresponding to the first audio data in which the candidate keyword is detected in the audio stream data; Extracting a first speaker feature vector for the first audio data; Extracting a second speaker feature vector for second audio data corresponding to a second section preceding the first section in the audio stream data; And determining whether or not the keyword is included in the first audio data based on the degree of similarity between the first speaker feature vector and the second speaker feature vector and determining whether or not the keyword is included in the first audio data. A method of operation can be provided.

In addition, the third aspect of the present disclosure may provide a computer-readable recording medium having recorded thereon one or more programs including instructions for causing a processor of a voice control apparatus to execute a method of operation according to the second aspect.

According to various embodiments of the present disclosure, malfunction of the voice control device can be prevented since the possibility of misrecognizing the keyword is reduced.

1 is a diagram illustrating an example of a network environment according to an embodiment.
2 is a block diagram for explaining an internal configuration of an electronic device and a server according to an embodiment.
3 is a diagram illustrating an example of functional blocks that a processor of a voice control apparatus according to an embodiment may include.
4 is a flowchart illustrating an example of an operation method that can be performed by the voice control apparatus according to an embodiment.
5 is a flowchart showing an example of an operation method that can be performed by the voice control apparatus according to another embodiment.
6A shows an example in which a single command keyword is uttered when the voice control apparatus according to an embodiment executes the operation method of FIG.
6B shows an example in which a general conversation voice is uttered when the voice control apparatus according to an embodiment executes the operation method of Fig.
7 is a flowchart showing an example of an operation method that the voice control apparatus can perform according to another embodiment.
8A shows an example in which a wake-up keyword and a natural language voice command are uttered when the voice control apparatus according to an embodiment executes the operation method of Fig.
FIG. 8B shows an example in which a general conversation voice is uttered when the voice control apparatus according to an embodiment executes the operation method of FIG. 7. FIG.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, which will be readily apparent to those skilled in the art. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

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.

The phrases " in some embodiments " or " in one embodiment " appearing in various places in this specification are not necessarily all referring to the same embodiment.

Some embodiments may be represented by functional block configurations and various processing steps. Some or all of these functional blocks may be implemented with various numbers of hardware and / or software configurations that perform particular functions. For example, the functional blocks of the present disclosure may be implemented by one or more microprocessors, or by circuit configurations for a given function. Also, for example, the functional blocks of the present disclosure may be implemented in various programming or scripting languages. The functional blocks may be implemented with algorithms running on one or more processors. In addition, the present disclosure may employ conventional techniques for electronic configuration, signal processing, and / or data processing, and the like. The terms " module " and " configuration " and the like are used extensively and are not limited to mechanical and physical configurations.

Also, the connection lines or connection members between the components shown in the figures are merely illustrative of functional connections and / or physical or circuit connections. In practical devices, connections between components can be represented by various functional connections, physical connections, or circuit connections that can be replaced or added.

In this disclosure, the keyword refers to voice information capable of waking up a specific function of the voice control device. The keyword is based on the user's voice signal, and may be a single command keyword or a wake-up keyword. The wakeup keyword is a voice-based keyword capable of switching the voice control apparatus in the sleep mode state to the wakeup mode, and may be a voice keyword such as " Clover ", " After the user has uttered the wake-up keyword, the user can utter a command for instructing the function or operation desired to be performed by the voice control apparatus in a natural language form. In this case, the voice control apparatus can voice-recognize a voice command in a natural language form and perform a function or an operation corresponding to the voice-recognized result. The single command keyword can be a voice keyword that can directly control the operation of the voice control device such as " stop " when the music is being played back. The wakeup keyword referred to in this disclosure may be referred to as a wakeup word, a hot word, a trigger word, and the like.

In the present disclosure, the candidate keyword includes words whose pronunciation is similar to that of the keyword. For example, when the keyword is "Clover", the candidate keywords may be "Clover", "Global", "Club", and the like. The candidate keyword may be defined as a keyword detected by the keyword detection unit of the voice control apparatus as a keyword in the audio data. The candidate keyword may be the same as the keyword, but it may be another word having a similar pronunciation to the keyword. Generally speaking, even if the user utters a sentence including a term corresponding to a candidate keyword, the voice control apparatus can wake up the word by misrecognizing the keyword. The voice control apparatus according to the present disclosure can also prevent the candidate keyword from being woken up by the candidate keyword even when the candidate keyword is detected in the voice signal.

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 perform a specific function of the voice control device.

In this disclosure, the voice control device refers to an electronic device equipped with a voice control function. The electronic device equipped with the voice control function may be an independent electronic device such as a smart speaker or an artificial intelligent speaker. The electronic device equipped with the voice control function may be a computing device equipped with a voice control function, for example, a desktop, a notebook, and the like, as well as a portable computer device such as a smart phone. In this case, the computing device may be provided with a program or application for executing the voice control function. The electronic device equipped with the voice control function may be an electronic product mainly performing a specific function, for example, a smart television, a smart refrigerator, a smart air conditioner, a smart navigation, or an infotainment system of an automobile. In addition, object Internet devices that can be controlled by voice may also be applicable.

Certain functions of the voice control device in this disclosure may include, but are not limited to, running an application installed in, for example, a voice control device. For example, if the voice control device is a smart speaker, particular functions of the voice control device may include music playback, internet shopping, voice information provision, control of electronic or mechanical devices connected to the smart speaker, and the like. For example, in the case where the voice control device is a smartphone, running the application may include dialing, navigating, searching the internet, or setting an alarm. For example, in the case where the voice control apparatus is a smart television, executing the application may include program search, channel search, or the like. In the case where the voice control device is a smart oven, executing the application may include searching for cooking methods and the like. In the case where the voice control device is a smart refrigerator, executing the application may include refrigeration and freezing status check, or temperature setting, and the like. In the case where the voice control device is a smart car, executing the application may include automatic starting, autonomous driving, automatic parking, and the like. Implementation of an application in this disclosure is not limited to the one just described.

In the present disclosure, the keyword may have a word form or a sphere form. In the present disclosure, voice commands that are uttered after the wakeup keyword may have the form of a sentence form, a word form, or a sphere form in natural language form.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram illustrating an example of a network environment according to an embodiment.

The network environment shown in FIG. 1 is illustratively illustrated as including a plurality of electronic devices 100a-100f, a server 200, and a network 300. [

The electronic devices 100a-100f are exemplary electronic devices that can be controlled by voice. Each of the electronic devices 100a-100f may perform a specific function in addition to the voice recognition function. Examples of the electronic devices 100a to 100f include a smart or artificial intelligent speaker, a smart phone, a mobile phone, a navigation device, a computer, a notebook, a terminal for digital broadcasting, a personal digital assistant (PDA) ), Tablet PCs, and smart electronic products. The electronic devices 100a-100f may communicate with the server 200 and / or other electronic devices 100a-100f via the network 300 using a wireless or wired communication scheme. However, the present invention is not limited thereto, and each of the electronic devices 100a-100f may operate independently without being connected to the network 300. [ The electronic devices 100a-100f may be referred to as an electronic device 100. [

The communication method of the network 300 is not limited and can be applied not only to a communication method using a communication network (for example, a mobile communication network, a wired Internet, a wireless Internet, a broadcasting network) -100f) may be included. For example, the network 300 may be a personal area network (LAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN) , A network such as the Internet, and the like. The network 300 may also include any one or more of a network topology including a bus network, a star network, a ring network, a mesh network, a star-bus network, a tree or a hierarchical network, It is not limited.

The server 200 may be implemented as a computer device or a plurality of computer devices that communicate with the electronic devices 100a-100f via the network 300 and perform a voice recognition function. The server 200 can be distributed in a cloud form and can provide commands, codes, files, contents, and the like.

For example, the server 200 receives an audio file provided from the electronic devices 100a to 100f, converts the audio signal in the audio file into a string (or text), and transmits the converted string (or text) 100a-100f. ≪ / RTI > In addition, the server 200 may provide a file for installing an application for performing a voice control function to the electronic devices 100a-100f connected through the network 300. [ For example, the second electronic device 100b can install an application using a file provided from the server 200. [ The second electronic device 100b accesses the server 200 under the control of an installed operating system (OS) and / or at least one program (for example, an installed voice control application) Recognition service can be provided.

2 is a block diagram for explaining an internal configuration of an electronic device and a server according to an embodiment.

The electronic apparatus 100 is one of the electronic apparatuses 100a to 100f in Fig. 1, and the electronic apparatuses 100a to 100f can have at least the internal configuration shown in Fig. Although the electronic device 100 is shown connected to the server 200 performing the voice recognition function through the network 300, this is exemplary, and the electronic device 100 may perform the voice recognition function independently have. The electronic device 100 is an electronic device that can be controlled by voice, and may be referred to as a voice control device 100. The voice control apparatus 100 may be included in or connected to a smart or artificial intelligent speaker, a computing device, a portable computing device, a smart home appliance, or the like, and may be wired and / or wirelessly connected thereto.

The electronic device 100 and the server 200 may include memories 110 and 210, processors 120 and 220, communication modules 130 and 230, and input / output interfaces 140 and 240. The memories 110 and 210 are computer readable recording media and may include a permanent mass storage device such as a random access memory (RAM), a read only memory (ROM), and a disk drive. The memory 110 and 210 may store an operating system and at least one program code (for example, a voice control application installed in the electronic device 100 and a code for a voice recognition application). These software components may be loaded into the memories 110 and 210 via communication modules 130 and 230 rather than a computer readable recording medium. For example, at least one program may be loaded into the memory 110, 210 based on a program installed by the developers or the file distribution system that distributes the installation files of the application via the network 300 .

Processors 120 and 220 may be configured to process instructions of a computer program by performing basic arithmetic, logic, and I / O operations. An instruction may be provided to the processor 120, 220 by the memory 110, 210 or the communication module 130, 230. For example, the processor 120, 220 may be configured to execute a command received in accordance with a program code stored in a recording device, such as the memory 110, 210.

The communication modules 130 and 230 can provide a function for the electronic device 100 and the server 200 to communicate with each other through the network 300 and provide a function for communicating with other electronic devices 100b and 100f . For example, a request (e.g., a voice recognition service request) generated by the processor 120 of the electronic device 100 in accordance with a program code stored in a recording device such as the memory 110 is transmitted to the network 110 under the control of the communication module 130. [ And may be transmitted to the server 200 through the Internet 300. Conversely, a character string (text) as a result of speech recognition provided under the control of the processor 220 of the server 200 may be transmitted to the communication module 130 of the electronic device 100 via the communication module 230 and the network 300 To the electronic device 100 via the Internet. For example, the voice recognition result of the server 200 received through the communication module 130 may be transmitted to the processor 120 or the memory 110. The server 200 can transmit control signals, commands, contents, files, and the like to the electronic device 100. The control signals and commands received through the communication module 130 can be transmitted to the processor 120 or the memory 110, And contents, files, and the like may be stored as a separate storage medium that the electronic device 100 may further include.

The input / output interfaces 140 and 240 may be means for interfacing with the input / output devices 150. For example, the input device may include not only the microphone 151 but also a device such as a keyboard or a mouse, and the output device may include not only the speaker 152 but also a status indicating LED (Light Emitting Diode) And a display for displaying a communication session. As another example, the input / output devices 150 may include a device with integrated functions for input and output, such as a touch screen.

The microphone 151 can convert the ambient sound into an electrical audio signal. The microphone 151 may be mounted on an external device (for example, a smart clock) communicably connected without being directly mounted in the electronic device 100, and the generated external signal may be transmitted to the electronic device 100 through communication . 2, the microphone 151 is shown as being included in the interior of the electronic device 100, but according to another embodiment, the microphone 151 is included in a separate device, And may be implemented in a form connected to wireless communication.

In other embodiments, the electronic device 100 and the server 200 may include more components than the components of FIG. For example, the electronic device 100 may be configured to include at least some of the input / output devices 150 described above, or may include other components such as a transceiver, Global Positioning System (GPS) module, camera, Elements.

FIG. 3 is a diagram illustrating an example of functional blocks that a processor of a voice control apparatus according to an embodiment may include; FIG. 4 illustrates an example of an operation method that a voice control apparatus can perform according to an embodiment; It is a flow chart.

3, the processor 120 of the voice control apparatus 100 includes an audio processing unit 121, a keyword detection unit 122, a speaker feature vector extraction unit 123, a wakeup determination unit 124, A recognition unit 125 and a function unit 126. [ At least some of the processor 120 and the functional blocks 121-126 may control the voice control apparatus 100 to perform the steps S110 through S190 included in the operating method illustrated in FIG. For example, at least some of the functional blocks 121-126 of the processor 120 and the processor 120 may be associated with code of the operating system that the memory 110 of the voice control device 100 includes and at least one program code To execute the corresponding command.

Some or all of the functional blocks 121-126 shown in FIG. 3 may be implemented in hardware and / or software configurations that perform particular functions. The functions performed by the functional blocks 121-126 shown in FIG. 3 may be implemented by one or more microprocessors, or may be implemented by circuit configurations for the function. Some or all of the functional blocks 121-126 shown in FIG. 3 may be software modules configured in various programming or scripting languages that are executed in the processor 120. For example, the audio processing unit 121 and the keyword detection unit 122 are implemented by a digital signal processor (DSP), and the speaker feature vector extraction unit 123, the wakeup determination unit 124, and the voice recognition unit 125 Software module.

The audio processing unit 121 receives the audio signal corresponding to the ambient sound, and generates audio stream data. The audio processing unit 121 can receive an audio signal corresponding to the ambient sound from an input device such as the microphone 151. [ The microphone 151 is included in a peripheral device communicatively connected to the voice control apparatus 100 and the audio processing unit 121 can receive the audio signal generated in the microphone 151 by communication. The ambient sound includes not only the voice uttered by the user but also the background sound. Therefore, an audio signal includes not only a voice signal but also a background sound signal. The background sound signal may correspond to noise in keyword detection and speech recognition.

The audio processing unit 121 may generate audio stream data corresponding to the continuously received audio signals. The audio processing unit 121 may filter and digitize the audio signal to generate audio stream data. The audio processing unit 121 may filter the audio signal to remove the noise signal and amplify the audio signal as compared with the background sound signal. Also, the audio processing unit 121 may remove the echo of the audio signal from the audio signal.

The audio processing unit 121 can always operate to receive the audio signal even when the voice control apparatus 100 operates in the sleep mode. The audio processing unit 121 operates at a low operating frequency when the voice control apparatus 100 operates in the sleep mode and can operate at a high operating frequency when the voice control apparatus 100 operates in the normal mode.

The memory 110 may temporarily store the audio stream data generated by the audio processing unit 121. [ The audio processing unit 121 may buffer audio stream data using the memory 110. [ The memory 110 stores not only audio data including a keyword but also audio data before a keyword is detected. In order to store recent audio data in the memory 110, the audio data stored in the memory 110 the longest before can be deleted. If the sizes allocated to the memory 110 are the same, audio data of the same period can be always stored. It is preferable that the period corresponding to the audio data stored in the memory 110 is longer than the time period during which the keyword is uttered.

According to another embodiment of the present invention, the memory 110 may extract and store a speaker feature vector for an audio stream generated by the audio processing unit 121. [ At this time, the speaker feature vector may be extracted and stored for an audio stream of a specific length. As described above, the speaker feature vector that is stored the longest to store the speaker feature vector for the recently generated audio stream may be deleted.

The keyword detection unit 122 detects a candidate keyword corresponding to a predefined keyword from the audio stream data generated by the audio processing unit 121. [ The keyword detection unit 122 may detect a candidate keyword corresponding to a predefined keyword from the audio stream data temporarily stored in the memory 110. [ The plurality of predefined keywords may be stored in the keyword storage 110a. The keyword storage 110a may be included in the memory 110. [

The candidate keyword means that the keyword is detected as a keyword in the audio stream data by the keyword detection unit 122. [ The candidate keyword may be the same as the keyword, or it may be another word that is pronounced similar to the keyword. For example, when the keyword is " Clover ", the candidate keyword may be " global ". In other words, when the user utters a sentence including " global ", the keyword detection unit 122 can detect and detect " global " The thus detected " global " corresponds to a candidate keyword.

The keyword detection unit 122 compares the audio stream data with known keyword data, and calculates the possibility of including a voice corresponding to the keyword in the audio stream data. The keyword detection unit 122 may extract audio features such as filter bank energy or Mel-Frequency Cepstral coefficients from the audio stream data. The keyword detection unit 122 may process these audio features using classifying windows, for example, using a support vector machine or a neural network. Based on the processing of the audio features, the keyword detection unit 122 can calculate the possibility of including a keyword in the audio stream data. The keyword detecting unit 122 can detect the candidate keyword by determining that the keyword is included in the audio stream data when the probability is higher than a preset reference value.

The keyword detection unit 122 may generate the artificial neural network using the speech samples corresponding to the keyword data and may be trained to detect the keyword in the audio stream data using the generated neural network. The keyword detection unit 122 can calculate the probability of a phoneme constituting each keyword or the overall probability of a keyword for each frame in the audio stream data. The keyword detection unit 122 can output the probability of a probability sequence corresponding to each phoneme or the probability of the keyword itself from the audio stream data. Based on this sequence or probability, the keyword detection unit 122 can calculate the possibility that the keyword is included in the audio stream data, and can determine that the candidate keyword is detected when the probability is greater than a predetermined reference value. The above-described method is illustrative, and the operation of the keyword detection unit 122 can be implemented in various ways.

In addition, the keyword detection unit 122 may extract the audio features for each frame in the audio stream data, thereby calculating the likelihood that the audio data of the corresponding frame corresponds to the human voice and the possibility of corresponding to the background sound. The keyword detection unit 122 may compare the possibility of correspondence to the human voice and the possibility of corresponding to the background sound and determine that the audio data of the corresponding frame corresponds to the human voice. For example, when the possibility that the audio data of the frame corresponds to the human voice is higher than the possibility that it corresponds to the background sound, the keyword detection unit 122 determines that the audio data of the corresponding frame corresponds to the human voice can do.

The keyword detection unit 122 can specify a section in which the candidate keyword is detected in the audio stream data, and can determine the start point and the end point of the section in which the candidate keyword is detected. The section in which the candidate keyword is detected in the audio stream data may be referred to as a keyword detection section, a current section, or a first section. The audio data corresponding to the first section in the audio stream data is referred to as first audio data. The keyword detecting unit 122 can determine the end of the section in which the candidate keyword is detected as the end point. According to another example, the keyword detecting unit 122 waits until silence occurs for a preset time (for example, 0.5 second) after the candidate keyword is detected, and then waits until the silence period is included in the first period It is possible to determine the end point or determine the end point of the first section so that the silence period is not included.

The speaker characteristic vector extracting unit 123 reads second audio data corresponding to the second section from the memory 110 in the audio stream data temporarily stored in the memory 110. The second section may be the previous section of the first section and the end point of the second section may be the same as the beginning of the first section. The second section may be referred to as the previous section. The length of the second section may be variably set according to the keyword corresponding to the detected candidate keyword. According to another example, the length of the second section may be fixedly set. According to another example, the length of the second section may be adaptively variable such that keyword detection performance is optimized. For example, when the audio signal outputted by the microphone 151 is "four-leaf clover" and the candidate keyword is "clover", the second audio data can correspond to the voice "four-leaf".

The speaker feature vector extractor 123 extracts a first speaker feature vector of the first audio data corresponding to the first section and a second speaker feature vector of the second audio data corresponding to the second section. The speaker feature vector extracting unit 123 can extract a speaker feature vector robust to speaker recognition from the audio data. The speaker characteristic vector extracting unit 123 extracts a speaker characteristic vector by converting a time domain based speech signal into a signal in a frequency domain and modifying the frequency energy of the converted signal differently from each other have. For example, the speaker feature vector may be extracted based on Mel-Frequency Cepstral Coefficients or Filter Bank Energy, but is not limited thereto and may be derived from the audio data in a variety of ways, Can be extracted.

The speaker feature vector extractor 123 can generally operate in the sleep mode. The keyword detecting unit 122 can wake up the speaker feature vector extracting unit 123 upon detecting a candidate keyword in the audio stream data. The keyword detection unit 122 can transmit a wakeup signal to the speaker characteristic vector extraction unit 123 when it detects a candidate keyword in the audio stream data. The speaker feature vector extracting unit 123 may be woken up in response to a wake-up signal indicating that the candidate keyword is detected in the keyword detecting unit 122. [

According to one embodiment, the speaker feature vector extractor 123 extracts a frame feature vector for each frame of audio data, normalizes and averages the extracted frame feature vectors, and extracts a speaker feature vector representing the audio data have. L2-normalization may be used to normalize the extracted frame feature vectors. The averaging of the extracted frame feature vectors can be achieved by calculating an average of the normalized frame feature vectors generated by normalizing the extracted frame feature vectors for each of all the frames in the audio data.

For example, the speaker feature vector extractor 123 extracts a first frame feature vector for each frame of the first audio data, normalizes and averages the extracted first frame feature vectors, The first speaker feature vector can be extracted. The speaker feature vector extractor 123 extracts a second frame feature vector for each frame of the second audio data and normalizes and averages the extracted second frame feature vectors to obtain a second speaker representative of the second audio data, Feature vectors can be extracted.

According to another embodiment, the speaker feature vector extractor 123 may extract the frame feature vectors for some frames in the audio data, rather than extracting the frame feature vectors for all the frames in the audio data. The certain frame may be selected as a voice frame in a frame in which the audio data of the corresponding frame is likely to be voice data of the user. The selection of such a voice frame may be performed by the keyword detecting unit 122. [ The keyword detection unit 122 may calculate the first probability of a human voice and the second probability of a background sound for each frame of the audio stream data. The keyword detection unit 122 may determine a frame whose audio data of each frame is higher than a second predetermined probability that the first probability that the audio data is the human voice is the background sound as a voice frame. The keyword detection unit 122 may store a flag or bit indicating whether the frame is a voice frame in the memory 110 for storing each frame of the audio stream data.

The speaker characteristic vector extracting unit 123 reads out the first and second audio data from the memory 110, and reads the flag or bit together to know whether or not the corresponding frame is a voice frame.

The speaker feature vector extractor 123 extracts a frame feature vector for each of the frames determined as a voice frame among the frames in the audio data, normalizes and averages the extracted first frame feature vectors, The vector can be extracted. For example, the speaker feature vector extractor 123 extracts a first frame feature vector for each of frames determined as a voice frame among frames in the first audio data, and normalizes and averages the extracted first frame feature vectors So that the first speaker characteristic vector representing the first audio data can be extracted. In addition, the speaker feature vector extractor 123 extracts a second frame feature vector for each of the frames determined as a voice frame from the frames in the second audio data, normalizes and averages the extracted second frame feature vectors, 2 < / RTI > audio data representative of the second speaker feature vector.

The wakeup determination unit 124 determines whether or not the keyword is included in the first audio data based on the similarity between the first speaker feature vector extracted by the speaker feature vector extractor 123 and the second speaker feature vector, , It is determined whether or not the keyword is included in the audio signal of the first section. The wake-up determination unit 124 compares the similarity between the first speaker feature vector and the second speaker feature vector with a preset reference value, and when the similarity is equal to or less than the reference value, the wakeup determination unit 124 includes the keyword in the first audio data of the first section .

In a typical case where the voice control apparatus 100 misrecognizes a keyword, a word whose pronunciation is similar to the keyword is located in the middle of the voice. For example, in the case where the keyword is " Clover ", the voice control apparatus 100 can also wake up in response to the " clover " even when the user tells another person "how to find four- Can perform unintended operations. Even in the case of a television news announcement, the voice control device 100 may wake up in response to a " global " even if the announcer says " In order to prevent the occurrence of the keyword misrecognition, the voice control apparatus 100 may react only when a word having a pronunciation similar to the keyword is located at the front of the voice. Also, in an environment with a high ambient background noise or in a conversation environment of other people, even if the user speaks the voice corresponding to the keyword at the front, the background noise or the conversation of the other person causes the voice corresponding to the keyword It may not be detected that it has been voiced before. According to one embodiment, the voice control apparatus 100 extracts a first speaker feature vector of a section in which a candidate keyword is detected and a second speaker feature vector of a previous section, and outputs a first speaker feature vector and a second speaker feature vector If they are different from each other, it can be determined that the user uttered the voice corresponding to the keyword at the head.

For this determination, when the degree of similarity between the first and second speaker characteristic vectors is equal to or less than a preset reference value, the wake-up determination unit 124 may determine that the user uttered a voice corresponding to the keyword have. That is, the wake-up determination unit 124 can determine that the keyword is included in the first audio data of the first section, and can wake up some functions of the voice control apparatus 100. The high degree of similarity between the first speaker feature vector and the second speaker feature vector means that there is a high possibility that the person speaking the voice corresponding to the first audio data and the person speaking the voice corresponding to the second audio data are the same.

If the second audio data corresponds to silence, the speaker feature vector extracting unit 123 may extract a second speaker feature vector corresponding to the silence from the second audio data. Since the speaker feature vector extractor 123 extracts a first speaker feature vector corresponding to the user's voice from the first audio data, the similarity between the first speaker feature vector and the second speaker feature vector may be low.

The speech recognition unit 125 receives the third audio data corresponding to the third section from the audio stream data generated by the audio processing unit 121, and recognizes the third audio data. According to another example, the speech recognition section 125 may transmit the third audio data to the outside so that the third audio data is recognized as speech in the external (for example, the server 200), and receive the speech recognition result.

The function unit 126 may perform a function corresponding to the keyword. For example, when the voice control apparatus 100 is a smart speaker, the function unit 126 may include a music reproducing unit, an audio information providing unit, a peripheral device control unit, and the like, have. When the voice control apparatus 100 is a smart phone, the function unit 126 may include a telephone connection unit, a character transmission / reception unit, an Internet search unit, and the like, and may perform a function corresponding to the detected keyword. The function unit 126 may be configured variously according to the type of the voice control apparatus 100. The functional unit 126 is a functional block for performing various functions that the voice control apparatus 100 can perform.

The voice control apparatus 100 shown in Fig. 3 is shown as including voice recognition section 125, but this is exemplary, voice control apparatus 100 does not include voice recognition section 125, The server 200 shown in FIG. 2 may perform the voice recognition function instead. In this case, as shown in FIG. 1, the voice control apparatus 100 may be connected to the server 200 performing the voice recognition function through the network 300. The voice control apparatus 100 may provide a voice file including a voice signal required for voice recognition to the server 200. The server 200 performs voice recognition on the voice signal in the voice file, Can be generated. The server 200 can transmit the generated character string to the voice control apparatus 100 through the network 300. [ However, it is assumed below that the voice control apparatus 100 includes a voice recognition unit 125 that performs a voice recognition function.

The processor 120 may load the program code stored in the program file for the method of operation into the memory 110. [ For example, a program may be installed in the voice control apparatus 100 according to a program file. At this time, when the program installed in the voice control apparatus 100 is executed, the processor 120 may load the program code into the memory 110. [ At this time, the audio processing unit 121, the keyword detection unit 122, the speaker feature vector extraction unit 123, the wake-up determination unit 124, the voice recognition unit 125, and the functional unit 126 included in the processor 120, May be implemented to execute instructions in accordance with the corresponding one of the program codes loaded into memory 110 to execute steps S110 through S190 of FIG.

It is to be appreciated that the functional blocks 121-126 of the processor 120 then control the voice control device 100 as the processor 120 controls the other components of the voice control device 100. [ For example, the processor 120 controls the communication module 130 included in the voice control apparatus 100 to control the voice control apparatus 100 to communicate with the server 200, for example, .

In step S110, the processor 120, for example, the audio processing unit 121 receives an audio signal corresponding to the ambient sound. The audio processing unit 121 can continuously receive the audio signal corresponding to the ambient sound. The audio signal may be an electrical signal generated by an input device such as the microphone 151 in response to the ambient sound.

In step S120, the processor 120, for example, the audio processing unit 121 generates audio stream data based on the audio signal from the microphone 151. [ The audio stream data corresponds to an audio signal that is continuously received. The audio stream data may be data generated by filtering and digitizing the audio signal.

In step S130, the processor 120, for example, the audio processing unit 121 temporarily stores audio stream data generated in step S120 in the memory 110. [ The memory 110 has a limited size and a part of the audio stream data corresponding to the audio signal for the most recent period of time from the present can be temporarily stored in the memory 110. [ When the new audio stream data is generated, the oldest data among the audio stream data stored in the memory 110 is deleted, and new audio stream data can be stored in the space freed by the deletion in the memory 110.

In step S140, the processor 120, for example, the keyword detection unit 122 detects a candidate keyword corresponding to a predefined keyword from the audio stream data generated in step S120. The candidate keyword is a word having a pronunciation similar to a predefined keyword, and refers to a word detected as a keyword in the keyword detection unit 122 in step S140.

In step S150, the processor 120, for example, the keyword detection unit 122 identifies the keyword detection period in which the candidate keyword is detected in the audio stream data, and determines the start point and the end point of the keyword detection period. The keyword detection interval may be referred to as the current interval. The data corresponding to the current section in the audio stream data may be referred to as first audio data.

In step S160, the processor 120, for example, the speaker feature vector extracting unit 123 reads the second audio data corresponding to the previous section from the memory 110. [ The previous period may be immediately before the current period, and the end point of the previous period may be the same as the current period. The speaker feature vector extracting unit 123 may read the first audio data from the memory 110 together.

In step S170, the processor 120, for example, the speaker feature vector extractor 123 extracts a first speaker feature vector from the first audio data and a second speaker feature vector from the second audio data. The first speaker feature vector is an index for identifying a speaker of a voice corresponding to the first audio data and the second speaker feature vector is an indicator for identifying a speaker of a voice corresponding to the second audio data. The processor 120, for example, the wake-up determination unit 124 may determine whether or not the keyword is included in the first audio data based on the degree of similarity between the first speaker feature vector and the second speaker feature vector. The wake-up determination unit 124 may wake up some components of the voice control apparatus 100 when determining that the keyword is included in the first audio data.

In step S180, the processor 120, for example, the wakeup determination unit 124 compares the similarity between the first speaker feature vector and the second speaker feature vector with a preset reference value.

If the similarity between the first speaker feature vector and the second speaker feature vector is equal to or less than a preset reference value, the speaker of the first audio data in the current section and the speaker of the second audio data in the previous section are different from each other , It can be determined that the keyword is included in the first audio data. In this case, as in step S190, the processor 120, for example, the wakeup determination unit 124 may wake up some components of the voice control apparatus 100. [

However, when the degree of similarity between the first speaker feature vector and the second speaker feature vector is greater than a predetermined reference value, the wake-up determination unit 124 determines whether the speaker of the first audio data of the current section and the speaker of the second audio data of the previous section It is judged that the keyword is not included in the first audio data, so that the wake-up can not proceed. In this case, the flow advances to step S110 to receive an audio signal corresponding to the ambient sound. The reception of the audio signal in step S110 continues even when performing steps S120-S190.

A plurality of predefined keywords may be stored in the keyword storage 110a of FIG. These keywords may be wake-up keywords or single command keywords. The wakeup keyword is for waking up some functions of the voice control apparatus 100. [ Generally, the user utters the desired natural language voice command after uttering the wakeup keyword. The voice control apparatus 100 can voice-recognize the natural voice command and perform the operation and function corresponding to the natural voice command.

The single command keyword is for the voice control apparatus 100 to directly perform a specific operation or function, and may be a predefined simple word such as " playback ", " When the single command keyword is received, the voice control apparatus 100 can wake up the function corresponding to the single command keyword and perform the corresponding function.

Hereinafter, the case where a candidate keyword corresponding to a single instruction keyword is detected from audio stream data and the case where a candidate keyword corresponding to a wakeup keyword is detected from audio stream data will be described.

5 is a flowchart showing an example of an operation method that can be performed by the voice control apparatus according to another embodiment.

FIG. 6A shows an example in which a single command keyword is uttered when the voice control apparatus according to an embodiment executes the operation method of FIG. 5, and FIG. 6B shows an example in which the voice control apparatus according to an embodiment performs an operation method The general conversation voice is uttered.

The method of operation of Figure 5 includes substantially the same steps as the method of operation of Figure 4. Steps substantially the same as the steps of FIG. 4 among the steps of FIG. 5 are not described in detail. 6A and 6B show the audio signals corresponding to the audio stream data and the user's voice corresponding to the audio signals. 6A shows audio signals corresponding to a voice " Pause " and FIG. 6B shows audio signals corresponding to a voice " Stop Here ".

Referring to FIG. 5 together with FIGS. 6A and 6B, in step S210, the processor 120, for example, the audio processing unit 121 receives audio signals corresponding to ambient sounds.

In step S220, the processor 120, for example, the audio processing unit 121 generates audio stream data based on the audio signal from the microphone 151. [

In step S230, the processor 120, for example, the audio processing unit 121 temporarily stores the audio stream data generated in step S220 in the memory 110. [

In step S240, the processor 120, for example, the keyword detection unit 122 detects a candidate keyword corresponding to a predefined single command keyword from the audio stream data generated in step S220. The single command keyword may be a voice keyword that can directly control the operation of the voice control apparatus 100. [ For example, the single command keyword may be a word such as " stop " as shown in FIG. 6A. In this case, the voice control apparatus 100 may be reproducing music or moving pictures, for example.

In the example of Fig. 6A, the keyword detection unit 122 may detect a candidate keyword " stop " in the audio signals. In the example of Fig. 6B, the keyword detecting section 122 can detect a candidate keyword " stop ", which is a word having pronunciation similar to the keyword " stop " in audio signals.

In step S250, the processor 120, for example, the keyword detection unit 122 identifies the keyword detection period in which the candidate keyword is detected in the audio stream data, and determines the start point and the end point of the keyword detection period. The keyword detection interval may be referred to as the current interval. The data corresponding to the current section in the audio stream data may be referred to as first audio data.

In the example of Fig. 6A, the keyword detection unit 122 can identify the section in which the candidate keyword " stop " is detected as the current section, and determine the start and end points of the current section. The audio data corresponding to the current section may be referred to as first audio data AD1.

In the example of FIG. 6B, the keyword detecting unit 122 can identify the section in which the candidate keyword " STOP " is detected as the current section, and determine the start and end points of the current section. The audio data corresponding to the current section may be referred to as first audio data AD1.

In step S250, the processor 120, for example, the keyword detection unit 122 can determine whether the detected candidate keyword is a candidate keyword corresponding to a wakeup keyword or a single command keyword. 6A and 6B, the keyword detecting unit 122 may determine that the detected candidate keywords, that is, "stop" and "stop" are candidate keywords corresponding to a single command keyword.

In step S260, the processor 120, for example, the speaker feature vector extracting unit 123 reads the second audio data corresponding to the previous section from the memory 110. [ The previous period may be immediately before the current period, and the end point of the previous period may be the same as the current period. The speaker feature vector extracting unit 123 may read the first audio data from the memory 110 together.

In the example of FIG. 6A, the speaker feature vector extractor 123 may read from the memory 110 the second audio data AD2 corresponding to the previous section, which is a section immediately before the current section. In the example of FIG. 6B, the speaker feature vector extractor 123 may read the second audio data AD2 corresponding to the previous section, which is a section immediately before the current section, from the memory 110. FIG. In the example of FIG. 6B, the second audio data AD2 may correspond to a voice called " Voice. &Quot; The length of the previous section may be variably set according to the detected candidate keyword.

In step S270, the processor 120, for example, the speaker feature vector extractor 123 receives the third audio data corresponding to the next section after the current section from the audio processor 121. [ The next interval is the interval immediately after the current interval, and the point of the next interval may be the same as the end point of the current interval.

In the example of FIG. 6A, the speaker feature vector extractor 123 may receive the third audio data AD3 corresponding to the next section immediately after the current section from the audio processor 121. FIG. In the example of FIG. 6B, the speaker feature vector extracting unit 123 may receive the third audio data AD3 corresponding to the next section immediately after the current section from the audio processing unit 121. FIG. In the example of FIG. 6B, the third audio data AD3 may correspond to a voice called " harm. &Quot; The length of the next section may be variably set according to the detected candidate keyword.

In step S280, the processor 120, for example, the speaker feature vector extraction unit 123 extracts the first to third speaker feature vectors from the first to third audio data, respectively. Each of the first to third speaker characteristic vectors are indicators for identifying a speaker of a voice corresponding to the first to third audio data. The processor 120, for example, the wake-up determination unit 124, determines the similarity between the first speaker feature vector and the second speaker feature vector, and the similarity between the first speaker feature vector and the third speaker feature vector, It is possible to judge whether or not a single command keyword is included in the command. The wake-up determination unit 124 may wake up some components of the voice control apparatus 100 when determining that the single command keyword is included in the first audio data.

In the example of Fig. 6A, the first speaker feature vector corresponding to the first audio data AD1 is an index for identifying a speaker who uttered the voice " stop ". Since the second audio data AD2 and the third audio data AD3 are substantially silent, the second and third speaker feature vectors may have vectors corresponding to silence. Thus, the similarity between the first speaker feature vector and the second and third speaker feature vectors may be low.

As another example, when a person other than the speaker who uttered the voice "stop" in the previous section and the next section utters the voice, the second and third speaker characteristic vectors will have the vector corresponding to the other person, The similarity between the first speaker feature vector and the second and third speaker feature vectors may be low.

In the example of FIG. 6B, one person uttered "stop here". Therefore, a first speaker feature vector extracted from the first audio data AD1 corresponding to the " stop ", a second speaker feature vector extracted from the second audio data AD2 corresponding to the " The third speaker feature vector extracted from the third audio data AD3 corresponding to the first speaker feature vector is a vector for identifying the speaker substantially identical to each other, so that the similarity between the first through third speaker feature vectors may be high.

In step S290, the processor 120, for example, the wakeup determination unit 124 compares the similarity between the first speaker feature vector and the second speaker feature vector with a preset reference value, And compares the similarity between the feature vectors with preset reference values. If the similarity between the first speaker feature vector and the second speaker feature vector is equal to or less than a preset reference value and the similarity between the first speaker feature vector and the third speaker feature vector is equal to or less than a predetermined reference value, Since the speaker of the first audio data is different from the speaker of the second audio data of the previous section and the speaker of the third audio data of the next section, it can be determined that the single command keyword is included in the first audio data. In this case, as in step S300, the processor 120, for example, the wakeup determination unit 124 provides the single command keyword to the function unit 126, and the function unit 126 controls the wakeup determination unit 124 In response to the determination that the single command keyword is included in the first audio data by the single command keyword.

In the example of FIG. 6A, the first speaker feature vector is a vector corresponding to a speaker uttered " stop ", and the second and third speaker feature vectors are vectors corresponding to silence, The similarities between the third speaker feature vectors may be lower than a preset reference value. In this case, the wake-up determination unit 124 may determine that the single command keyword " stop " is included in the first audio data AD1. In this case, the function unit 126 may be woken up in response to the determination, and may perform an operation or function corresponding to the sole command keyword " stop ". For example, if the voice control apparatus 100 is playing music, the function unit 126 may stop the music reproduction in response to the sole command keyword " stop ".

However, if the degree of similarity between the first speaker feature vector and the second speaker feature vector is greater than a preset reference value, or if the similarity between the first speaker feature vector and the third speaker feature vector is greater than a preset reference value The speaker of the first audio data of the current section is the same as the speaker of the second audio data of the previous section or the speaker of the third audio data of the next section and therefore judges that the keyword is not included in the first audio data, It may not proceed. In this case, the flow advances to step S210 to receive an audio signal corresponding to the ambient sound.

In the example of Fig. 6B, since a person uttered " stop here ", the similarity between the first to third speaker feature vectors would be high. 6B, the wake-up determination unit 124 determines that the wake-up determination unit 124 determines that the audio data AD1 is a stand-alone command It can be determined that the keyword is not included and the function unit 126 can not perform the function or operation corresponding to " stop " or " stop ".

According to the general technique, the voice control apparatus can detect a voice called " stop " during utterance " stop here " to perform a function or an operation corresponding to " stop ". Such a function or operation is not intended by the user, and the user may feel inconvenience when using the voice control device. However, according to the embodiment, since the voice control apparatus 100 can correctly recognize the single command keyword from the voice of the user, unlike the general technique, the voice control apparatus 100 may not perform malfunction.

7 is a flowchart showing an example of an operation method that the voice control apparatus can perform according to another embodiment.

FIG. 8A shows an example in which a wake-up keyword and a natural language voice command are uttered when the voice control apparatus according to an embodiment executes the operation method of FIG. 7, and FIG. 8B shows an example in which the voice control apparatus according to an embodiment 7 shows an example in which a general conversation voice is uttered.

The method of operation of Figure 7 includes substantially the same steps as the method of operation of Figure 4. Steps substantially the same as the steps of FIG. 4 among the steps of FIG. 7 are not described in detail. 6A and 6B show the audio signals corresponding to the audio stream data and the user's voice corresponding to the audio signals. Fig. 8A shows audio signals corresponding to the wakeup keyword " Clover " and the natural voice command " Tell Tomorrow Weather " and Fig. 6B shows audio signals corresponding to the conversation voice "How to find a four- do.

Referring to FIG. 7 together with FIGS. 8A and 8B, in step S410, the processor 120, for example, the audio processing unit 121 receives audio signals corresponding to ambient sounds. In step S420, the processor 120, for example, the audio processing unit 121 generates audio stream data based on the audio signal from the microphone 151. [ In step S430, the processor 120, for example, the audio processing unit 121 temporarily stores the audio stream data generated in step S120 in the memory 110. [

In step S440, the processor 120, for example, the keyword detection unit 122 detects candidate keywords corresponding to predefined wakeup keywords from the audio stream data generated in step S420. The wake-up keyword is a voice-based keyword that can switch the voice control device in the sleep mode to the wake-up mode. For example, the wakeup keyword may be a spoken keyword such as " clover, " " high computer,

In the example of FIG. 8A, the keyword detecting section 122 can detect a candidate keyword " Clover " in the audio signals. In the example of Fig. 6B, the keyword detecting unit 122 can detect a candidate keyword " clover " which is a word having a similar pronunciation to the keyword " clover " in the audio signals.

In step S450, the processor 120, for example, the keyword detection unit 122 identifies the keyword detection period in which the candidate keyword is detected in the audio stream data, and determines the start point and the end point of the keyword detection period. The keyword detection interval may be referred to as the current interval. The data corresponding to the current section in the audio stream data may be referred to as first audio data.

In the example of FIG. 8A, the keyword detection unit 122 can identify the section in which the candidate keyword " Clover " is detected as the current section, and determine the start and end points of the current section. The audio data corresponding to the current section may be referred to as first audio data AD1. In the example of FIG. 8B, the keyword detecting section 122 can identify the section in which the candidate keyword " clover " is detected as the current section, and determine the start and end points of the current section. The audio data corresponding to the current section may be referred to as first audio data AD1.

In step S450, the processor 120, for example, the keyword detection unit 122 can determine whether the detected candidate keyword is a candidate keyword corresponding to a wakeup keyword or a single command keyword. 8A and 8B, the keyword detection unit 122 can determine that the detected candidate keywords, i.e., "Clover" and "Clover" are candidate keywords corresponding to the wakeup keyword.

In step S460, the processor 120, for example, the speaker characteristic vector extracting unit 123 reads the second audio data corresponding to the previous section from the memory 110. [ The previous period may be immediately before the current period, and the end point of the previous period may be the same as the current period. The speaker feature vector extracting unit 123 may read the first audio data from the memory 110 together.

In the example of FIG. 8A, the speaker feature vector extractor 123 may read from the memory 110 the second audio data AD2 corresponding to the previous section, which is a section immediately before the current section. In the example of FIG. 8B, the speaker feature vector extractor 123 may read from the memory 110 the second audio data AD2 corresponding to the previous section, which is a section immediately before the current section. In the example of Fig. 8B, the second audio data AD2 may correspond to a voice of " four leaves ". The length of the previous section may be variably set according to the detected candidate keyword.

In step S470, the processor 120, for example, the speaker feature vector extraction unit 123 extracts the first and second speaker feature vectors from the first and second audio data, respectively. The processor 120, for example, the wakeup determination unit 124 may determine whether or not the first audio data includes the wakeup keyword based on the degree of similarity between the first speaker feature vector and the second speaker feature vector. The wakeup determination unit 124 may wake up some components of the voice control apparatus 100 when determining that the first audio data includes the wakeup keyword.

In the example of FIG. 8A, the first speaker feature vector corresponding to the first audio data AD1 is an index for identifying a speaker who uttered the voice "Clover". Since the second audio data AD2 is substantially silent, the second speaker feature vector may have a vector corresponding to silence. Thus, the similarity between the first speaker feature vector and the second speaker feature vector may be low.

As another example, when a person other than the speaker who uttered the voice of " Clover " in the previous section utters the voice, the second speaker characteristic vector will have the vector corresponding to the other person, The similarity between the second speaker feature vectors may be low.

In the example of FIG. 8B, one person uttered "How can I find a four-leaf clover?" Therefore, the first speaker characteristic vector extracted from the first audio data AD1 corresponding to " clover " and the second speaker characteristic vector extracted from the second audio data AD2 corresponding to " four leaves & Since the vector is for identifying the speaker, the similarity between the first and second speaker feature vectors may be high.

In step S480, the processor 120, for example, the wake-up determination unit 124 compares the similarity between the first and second speaker feature vectors with preset reference values. If the similarity degree between the first and second speaker feature vectors is greater than a preset reference value, the wake-up determination unit 124 determines that the speaker of the first audio data in the current section and the speaker of the second audio data in the previous section are the same , It is judged that the keyword is not included in the first audio data, so that the wake-up can not proceed. In this case, the process proceeds to step S410, and the processor 120, for example, the audio processing unit 121 receives the audio signal corresponding to the ambient sound.

In the example of Fig. 8B, since a person utters "four-leaf clover ...", the similarity between the first and second speaker feature vectors will be high. 8B, it is determined that the person who uttered "four-leaf clover" does not intend to wake up the voice control apparatus 100, and the wake-up determination unit 124 determines that the wakeup keyword is included in the first audio data AD1 And does not wake up the voice control apparatus 100. [0052]

If the similarity degree between the first and second speaker feature vectors is equal to or less than a preset reference value, the speaker of the first audio data of the current section and the speaker of the second audio data of the previous section are different from each other , It can be determined that the keyword is included in the first audio data. In this case, the wake-up determination unit 124 may wake up some components of the voice control apparatus 100. [ For example, the wake-up determination unit 124 may wake up the voice recognition unit 125. [

8A, since the first speaker feature vector is a vector corresponding to a speaker uttered as " Clover " and the second speaker feature vector is a vector corresponding to silence, the similarity between the first and second speaker feature vectors is And may be lower than a preset reference value. In this case, the wake-up determination unit 124 may determine that the wakeup keyword " Clover " is included in the first audio data AD1. In this case, the voice recognition unit 125 can wake up to recognize a natural voice command.

In step S490, the processor 120, for example, the voice recognition unit 125 receives the third audio data corresponding to the next section after the current section from the audio processing unit 121. [ The next interval is the interval immediately after the current interval, and the point of the next interval may be the same as the end point of the current interval.

The voice recognition unit 125 can determine the end point of the next section when silence of a predetermined length is detected in the third audio data. The voice recognition unit 125 can recognize the third audio data. The voice recognition unit 125 can recognize the third audio data in various ways. According to another example, in order to obtain the voice recognition result of the third audio data, the voice recognition unit 125 transmits the third audio data to an external device, for example, the server 200 having the voice recognition function shown in Fig. 2 . The server 200 receives the third audio data, generates a character string (text) corresponding to the third audio data by voice recognition of the third audio data, and outputs the generated character string (text) (125).

In the example of FIG. 8A, the third audio data of the next section is a natural language voice command such as " Tell Tomorrow Weather ". The speech recognition unit 125 may directly recognize the third audio data to generate a speech recognition result, or may transmit the third audio data to the external (e.g., the server 200) so that the third audio data is recognized as speech.

In step S500, the processor 120, for example, the functional unit 126 may perform a function corresponding to the speech recognition result of the third audio data. In the example of FIG. 8A, the function unit 126 may be a voice information providing unit for searching for tomorrow's weather and providing a result, and the function unit 126 searches for tomorrow's weather using the Internet, can do. The function unit 126 may provide the search result of tomorrow's weather by using the speaker 152 as a voice. The function unit 126 can be woken up in response to the speech recognition result of the third audio data.

The embodiments of the present invention described above can be embodied in the form of a computer program that can be executed on various components on a computer, and the computer program can be recorded on a computer-readable medium. At this time, the medium may be a program that continuously stores a computer executable program, or temporarily stores the program for execution or downloading. In addition, the medium may be a variety of recording means or storage means in the form of a combination of a single hardware or a plurality of hardware, but is not limited to a medium directly connected to a computer system, but may be dispersed on a network. Examples of the medium include a magnetic medium such as a hard disk, a floppy disk and a magnetic tape, an optical recording medium such as CD-ROM and DVD, a magneto-optical medium such as a floptical disk, And program instructions including ROM, RAM, flash memory, and the like. As another example of the medium, a recording medium or a storage medium managed by a site or a server that supplies or distributes an application store or various other software to distribute the application may be mentioned.

In this specification, " part, " " module, " and the like, may be a hardware component, such as a processor or circuit, and / or a software component, executed by a hardware configuration, such as a processor. For example, " part, " " module, " and / or the like, may refer to elements such as software components, object oriented software components, class components and task components, Microcode, circuitry, data, databases, data structures, tables, arrays, and variables, as will be appreciated by those skilled in the art.

It will be understood by those skilled in the art that the foregoing description of the present invention is for illustrative purposes only and that those of ordinary skill in the art can readily understand that various changes and modifications may be made without departing from the spirit or essential characteristics of the present invention. 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 invention 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.

100: Voice control device (electronic device)
110: Memory
120: Processor
121: Audio processor
122: Keyword detection unit
123: speaker characteristic vector extracting unit
124: Wake-up judgment unit
125:
126:

Claims (10)

An audio processing unit for receiving audio signals corresponding to ambient sounds and generating audio stream data;
A keyword detection unit detecting a candidate keyword corresponding to a predetermined keyword from the audio stream data and determining a first section corresponding to the first audio data in which the candidate keyword is detected in the audio stream data;
Extracting a first speaker feature vector for the first audio data and extracting a second speaker feature vector for second audio data corresponding to a second section preceding the first section in the audio stream data A vector extractor; And
And a wake-up determination unit for determining whether the keyword is included in the first audio data based on the degree of similarity between the first speaker feature vector and the second speaker feature vector. The method according to claim 1,
Wherein the wake-up determination unit determines that the keyword is included in the first audio data when the degree of similarity between the first speaker feature vector and the second speaker feature vector is equal to or less than a preset reference value. The method according to claim 1,
When the keyword detection unit detects the candidate keyword corresponding to the single command keyword in the audio stream data,
Wherein the speaker feature vector extraction unit receives third audio data corresponding to a third section following the first section from the audio stream data, extracts a third speaker feature vector of the third audio data,
The wake-up determination unit may further include a wake-up determination unit for determining whether the wake-up determination unit determines that the wake-up determination unit determines that the wake-up determination unit determines that the wake- Is included in the audio signal. The method of claim 3,
The wake-up determination unit may determine that the degree of similarity between the first speaker feature vector and the second speaker feature vector is equal to or less than a preset reference value and the similarity between the first speaker feature vector and the third speaker feature vector is equal to or less than a preset reference value And determines that the single command keyword is included in the first audio data. The method according to claim 1,
When the keyword detection unit detects the candidate keyword corresponding to the wakeup keyword in the audio stream data,
In response to a determination that the wakeup keyword is included in the first audio data by the wakeup determination unit, a third audio corresponding to a third interval following the first interval in the audio stream data, Further comprising a voice recognition unit for receiving the data and for recognizing the third audio data or transmitting the third audio data to the outside so as to recognize the third audio data. Receiving audio signals corresponding to ambient sounds to generate audio stream data;
Detecting a candidate keyword corresponding to a predetermined keyword from the audio stream data and determining a first section corresponding to the first audio data in which the candidate keyword is detected in the audio stream data;
Extracting a first speaker feature vector for the first audio data;
Extracting a second speaker feature vector for second audio data corresponding to a second section preceding the first section in the audio stream data; And
Determining whether the keyword is included in the first audio data based on the similarity between the first speaker feature vector and the second speaker feature vector, and determining whether to wake up the audio data. The method according to claim 6,
Wherein the step of determining whether to wake-
Comparing the similarity between the first speaker feature vector and the second speaker feature vector with a preset reference value;
Determining that the keyword is included in the first audio data and waking up if the similarity is less than or equal to the preset reference value; And
Determining that the keyword is not included in the first audio data and not waking up if the similarity exceeds the predetermined reference value. The method according to claim 6,
When the detected candidate keyword is the candidate keyword corresponding to the single command keyword,
Receiving third audio data corresponding to a third interval following the first interval in the audio stream data;
Extracting a third speaker feature vector of the third audio data;
When the similarity degree between the first speaker feature vector and the second speaker feature vector is equal to or less than a preset reference value and the similarity between the first speaker feature vector and the third speaker feature vector is equal to or less than a predetermined reference value, Further comprising determining that the data includes the single command keyword. The method according to claim 6,
If the detected keyword is the candidate keyword corresponding to the wakeup keyword,
Receiving third audio data corresponding to a third interval following the first interval in the audio stream data in response to the determination that the wakeup keyword is included in the first audio data; And
Further comprising the steps of: recognizing the third audio data or transmitting the third audio data to the outside so as to recognize the third audio data. A computer program comprising instructions for causing a processor of a voice control device to perform the method of any one of claims 6 to 9.

Images