KR101151571B1 - Speech recognition environment control apparatus for spoken dialog system and method thereof - Google Patents

Abstract

The present invention relates to an apparatus and method for controlling a speech recognition environment for a speech conversation system for improving a speech conversation recognition level by improving a surrounding environment using a speech conversation processing technology. The apparatus for controlling a speech recognition environment for the speech conversation system of the present invention is an induction apparatus for adjusting from a different speech recognition environment to a speech recognition environment in which a user is well recognized. The apparatus includes a speech and control interface for recognizing a target sentence and a user environment controller. Such a device may be connected to the leading end of all devices related to the voice conversation system to provide a smooth voice recognition environment.

Description

Speech recognition environment control apparatus for speech dialogue system and method thereof

The present invention relates to an apparatus and method for controlling a speech recognition environment for a voice conversation system. More particularly, the present invention relates to a voice recognition environment control apparatus and a method thereof. The present invention relates to a voice recognition environment control apparatus and a method for requesting speech and grasping and adjusting an ambient state based on the degree of speech recognition.

In general, a voice conversation system is aimed at smooth communication between a user and a system, and the system recognizes a command from a user's speech.

Therefore, the voice conversation system should be well-behaved from recognizing what a person's speech is. Recently, the voice conversation system has been developed with the focus on the performance of the voice recognizer itself, and the environmental control for the voice recognition has been developed with respect to the signal processing.

However, in the case of the existing voice conversation system, even though the environment variable affects the voice recognition level of the voice recognizer, it is not taken into consideration. In the voice recognition environment control part, the variables must be adjusted in relation to the recognition level for the voice conversation system. Nevertheless, there are limitations because they are independent of signal processing such as noise reduction and sound effects regardless of the intention of the voice conversation system.

On the other hand, the environment control apparatus for the voice conversation system ultimately evaluates the degree of speech recognition, and the environment is controlled in a direction in which speech recognition is good, and such feedback should be provided to the user.

Korea Patent Registration 10-0940629 (2010. 01. 28) (Noise removing device and method)

Accordingly, the inventors of the present invention have been made to solve the above-mentioned limitations of the prior art, which is an object of the present invention, that is, an object of the present invention is to provide a voice conversation that provides an environmental control feedback in accordance with the intention of the voice conversation system to have an appropriate recognition degree. An apparatus and method for controlling a speech recognition environment for a system are provided.

Another technical problem to be achieved by the present invention, that is, another object of the present invention is to provide an apparatus and method for controlling a speech recognition environment for a voice conversation system capable of grasping the degree of environment variables for robust speech recognition and making feedback on adjustment requests. It is.

Another technical problem to be achieved by the present invention, that is, another object of the present invention is to make a voice utterance request in consideration of a voice recognition domain of a voice conversation system, and have a voice conversation capable of grasping a degree of an environmental variable according to the voice made in the speech. An apparatus and method for controlling a speech recognition environment for a system are provided.

Another object of the present invention, that is, another object of the present invention is to provide an apparatus and method for controlling a voice recognition environment for a voice conversation system in which a user can adjust his / her own environmental parameters according to feedback.

Another technical problem to be achieved by the present invention, that is, another object of the present invention is to provide feedback on which environmental variables and how much are adjusted for feedback of the voice conversation system, and have a voice with the control unit. An apparatus and method for controlling a speech recognition environment for a conversation system are provided.

Another object of the present invention, that is, another object of the present invention is to provide an apparatus and method for controlling a voice recognition environment for a voice conversation system that terminates the control of the voice recognition environment when it is determined that it is robust.

Another technical problem to be achieved by the present invention, that is, another object of the present invention is to provide a robust environment without storing the user's profile for the case where the hardware system configuration dependent on the environment variable is the same. An apparatus and method for controlling a voice recognition environment for a voice conversation system are provided.

In order to achieve the above technical problem, the present invention provides a voice recognition environment control apparatus for a voice conversation system,

A user's current device search and environment initialization unit by an environment control start command;

A noise degree evaluator for estimating noise from a microphone input level of the searched microphone before requesting a speech from a current user;

It provides a voice recognition environment control apparatus including a microphone volume evaluation unit for completing the environmental control by evaluating the volume of the microphone according to the estimated noise level.

Preferably the microphone volume evaluator

A speech and control interface for requesting a user to speak a sentence and outputting a sentence as a result of the sentence that the user has spoken; And

According to the result of the ignition and adjustment of the ignition and adjustment interface, it is characterized in that it comprises a microphone environment control unit for determining whether the current microphone volume adjustment value is appropriate, whether to increase or decrease and send out the degree of adjustment to the user.

Preferably the ignition and conditioning interface is

A speech recognition unit for recognizing a speech input by a user through a microphone according to a speech request and outputting a sentence as a result of which sentence the user has uttered; And

It includes a microphone volume control that allows the user to adjust the input level of the microphone,

Randomly extracts a speech request sentence from a first database storing a voice conversation system domain model for the speech request of the speech recognition unit and requests a speech, and adjusts the microphone volume in the microphone volume control unit and completes the microphone volume control. And if the adjustment value is stored in a second database storing user environment profile data including environment control information.

Preferably the microphone environment control unit

A speech recognition evaluator for evaluating a speech recognition score from sentences obtained from the speech recognizer of the speech and control interface;

A microphone input level evaluator configured to evaluate a microphone input level obtained during speech recognition of the speech recognition unit of the speech and control interface; And

Evaluate the current microphone environment from the voice recognition score obtained from the voice recognition evaluation unit, the microphone input level obtained from the microphone level evaluation unit, and the microphone volume adjustment value obtained from the microphone volume control unit of the speech and control interface. The microphone environment evaluation part which feeds back is included.

Preferably, the speech recognition evaluator compares the sentence received from the speech recognizer with the current speech request sentence and measures the actual speech recognition similarity, and outputs the recognition score.

Preferably, the microphone input level evaluator measures the amount of signal physical input at the moment of the user's speech and outputs the microphone input level of the average level.

In accordance with another aspect of the present invention, there is provided a voice recognition environment control method for a voice conversation system.

A device search and environment initialization step of performing a user's current device search and environment initialization by an environment control start command;

A noise level estimation step of estimating noise from a microphone input level of the searched microphone before requesting a speech from a current user; And

It provides a voice recognition environment control method comprising a microphone volume evaluation step of evaluating the volume of the microphone according to the evaluated noise level to complete the environmental control.

Preferably the microphone volume evaluation step

An utterance and adjustment step of requesting a user to speak a sentence and outputting a sentence as a result of the sentence that the user uttered; And

According to the result of the ignition and adjustment of the ignition and adjustment step, it is characterized in that it comprises a microphone environment control step of determining whether the current microphone volume control value is appropriate, whether to increase or decrease and send out the degree of adjustment to the user and feedback. .

Preferably the ignition and control step

Randomly extracts a speech request sentence from a first database storing a voice conversation system domain model for a speech request and requests a speech, and recognizes a speech input by a user through a microphone according to the speech request, so that the user can recognize any sentence. A speech recognition step of outputting a sentence as a result of uttering speech; And

And adjusting the microphone volume and storing the adjustment value in a second database storing user environment profile data including environmental control information when the microphone volume adjustment is completed.

Preferably the microphone environment control step

A speech recognition evaluation step of evaluating a speech recognition score from sentences obtained from the speech recognition step of the speech and adjustment step;

A microphone input level evaluating step of evaluating a microphone input level obtained during speech recognition of the speech recognition step of the utterance and adjustment step; And

Evaluate the current microphone environmental condition from the voice recognition score obtained from the voice recognition evaluation step, the microphone input level obtained from the microphone level evaluation step, and the microphone volume adjustment value obtained from the microphone volume adjustment step of the utterance and adjustment step. And a microphone environment evaluation step of feeding back.

Preferably, the speech recognition evaluating step is characterized in that the actual speech recognition similarity is measured by comparing the sentence from the speech recognition step with the current speech request, and outputs the recognition score.

Preferably, the microphone input level evaluation step is characterized by measuring the amount of signal physical input at the moment the user speaks and outputs the microphone input level of the average degree.

The apparatus and method for controlling a voice recognition environment for a voice conversation system according to the present invention can solve the problem caused by the environment of the voice recognition of the voice conversation system, thereby improving the recognition performance and enabling a smooth conversation.

Since the present invention can be applied to any environment in which a voice conversation system can be used, it can be expected that the ripple effect on the related apparatus and industry is large.

1 is a block diagram showing the structure of a voice recognition control device for a voice conversation system of the present invention.
FIG. 2 is a detailed block diagram of a specific structure of the microphone volume evaluator of FIG. 1.
3 is a flowchart illustrating a voice recognition control method for a voice conversation system according to the present invention.
4 is a detailed flowchart of FIG. 3.
5 is a detailed flowchart of FIG. 4.

Hereinafter, a voice recognition control apparatus for a voice conversation system according to a preferred embodiment of the present invention will be described in detail with reference to FIGS. 1 and 2.

1 is a block diagram showing the overall structure of a voice recognition environment control apparatus for a voice conversation system according to the present invention.

In FIG. 1, the apparatus 1-1 for recognizing an environment for a voice conversation system according to the present invention includes a device search and environment control initialization unit 1-3, a noise level evaluator 1-4, and a microphone volume evaluator. (1-5), including, in order to proceed with the process.

The device search and initialization unit 1-3 shown in FIG. 1 searches for the devices required for the user's voice chat system by the environment control start command 1-2 and checks whether the requirements of the voice chat system are satisfied. Initialize the environment variables of the device.

The noise level evaluator 1-4 shown in FIG. 1 is a step before evaluating the speech recognition level of the microphone volume evaluator 1-5, and before a current user is requested to speak, the microphone of the signal received from the searched microphone. Noise is estimated by evaluating its noise level from the input level.

The microphone volume evaluator 1-5 shown in FIG. 1 completes the environmental control by evaluating the volume of the microphone according to the noise degree evaluated by the noise degree evaluator 1-4 (1-6). This is illustrated in more detail in FIG.

FIG. 2 is a block diagram illustrating the microphone volume evaluator 1-5 shown in FIG. 1 in more detail. The microphone volume evaluator 1-5 is composed of two parts. The microphone volume evaluator 1-5 includes a control and utterance interface 2-3 and a microphone environment control unit 2-8 in which a user directly utters and controls an external device.

In FIG. 2, the control and utterance interface 2-3 asks the user to utter a sentence, outputs a sentence as a result of which sentence the user uttered, and the microphone environment controller 2-8 According to the result of the ignition and adjustment of the ignition and adjustment interface 2-3, it is determined whether the current microphone volume adjustment value is appropriate, and whether to increase or decrease and send out the degree of adjustment to the user to feed back.

More specifically, the control and utterance interface (2-3) is composed of the same voice recognition unit (2-4) and the microphone volume control unit (2-5) and the microphone environment evaluation unit (2-) 8 is composed of a speech recognition evaluator 2-9, a microphone input level evaluator 2-10, and a microphone environment evaluator 2-11.

The speech recognition unit 2-4 of the speech and control interface 2-3 recognizes the speech input by the user through the microphone according to the speech request, and outputs the sentence as a result of which sentence the user has spoken, and the microphone. The volume control unit 2-5 allows the user to adjust the input level of the microphone.

In this case, the speech and control interface 2-3 receives a sentence to request speech from the first database 2-6, which stores the voice conversation system domain model for the speech request of the speech recognition unit 2-4. A second database for randomly extracting and requesting utterance, and storing user environment profile data including environment control information, the adjustment value after the user adjusts the microphone volume in the microphone volume control unit 2-5. 7) wherein the first and second databases 2-6 and 2-7 are included inside or outside the ignition and conditioning interface 2-3.

In addition, the speech recognition evaluator 2-9 of the microphone environment evaluator 2-8 obtains the speech recognition score from the sentences obtained from the speech recognizer 2-4 of the speech and control interface 2-3. The microphone input level evaluator 2-10 evaluates the microphone input level obtained during voice recognition of the speech recognition unit 2-4 of the speech and control interface 2-3, and the microphone environment evaluator. (2-11) indicates the speech recognition score obtained from the speech recognition evaluator 2-9, the microphone input level obtained from the microphone level evaluator 2-10, and the speech and control interface 2-3. From the microphone volume control value obtained from the microphone volume control unit 2-5, the current microphone environmental condition is evaluated and fed back to the user.

Here, the speech recognition evaluator 2-9 measures the actual speech recognition similarity from the sentences passed from the speech recognizer 2-4 and compares the sentences with the current speech request, and outputs the recognition scores to the microphone input level. The evaluation unit 2-10 measures the amount of signal physical input at the moment of the user's speech and emits the microphone input level of the average level.

Hereinafter, a voice recognition control method for a voice conversation system according to a preferred embodiment of the present invention will be described in detail with reference to FIGS. 3 to 5.

3 is a flowchart illustrating a voice recognition control method for a voice conversation system according to the present invention.

According to another aspect of the present invention, according to another aspect of the present invention, a voice recognition environment control method for a voice conversation system includes a device search and environment initialization step 32 of performing a user device current search and environment initialization by an environment control start command. ); A noise level estimation step (34) of estimating noise from a microphone input level of the retrieved microphone before requesting a speech from a current user; And a microphone volume evaluation step 36 of evaluating the volume of the microphone according to the evaluated noise degree to complete an environmental control.

4 is a detailed flowchart of FIG. 3, which describes the microphone volume evaluation step 36 in more detail.

In Fig. 4, the microphone volume evaluating step 36 includes: a speech and adjustment step 362, which requests a user to speak a sentence and outputs a sentence as a result of which sentence the user has spoken; And a microphone environment control step 364 that determines whether the current microphone volume adjustment value is appropriate, increases or decreases it, and sends back a degree of adjustment to the user according to the result of the ignition and adjustment of the ignition and adjustment step. .

5 is a detailed flowchart of FIG. 4, which details the ignition and adjustment steps 362 and the microphone environment control step 364 of FIG. 4.

In FIG. 5, the speaking and adjusting step 362 may be performed by a user inputting a microphone through a microphone according to a speech request obtained by referring to a first database 2-6 storing a voice conversation system domain model for the speech request. A speech recognition step 3622, which recognizes and outputs a sentence as a result of which sentence the user has uttered; And a microphone volume adjusting step 3642 in which the user adjusts the input level of the microphone and stores the adjustment value in the second database 2-7 storing user environment profile data including environmental control information after the adjustment is completed. And the microphone environment control step 364 includes: a speech recognition evaluation step 3642 for evaluating a speech recognition score from a sentence obtained from the speech recognition step of the utterance and adjustment step; A microphone input level evaluating step (3644) for evaluating a microphone input level obtained during speech recognition of the speech recognition step of the speech and adjustment step; And evaluating the current microphone environment from the voice recognition score obtained from the voice recognition evaluation step, the microphone input level obtained from the microphone level evaluation step, and the microphone volume adjustment value obtained from the microphone volume adjustment step of the utterance and adjustment step. And a microphone environment assessment step 3646 that feeds back.

Here, the speech recognition evaluating step 3642 measures the actual speech recognition similarity from the sentences passed from the speech recognition step 3622 and compares the sentences with the current speech request, and outputs the recognition scores. ) Measures the amount of signal physical input at the moment the user speaks and outputs the average microphone input level.

Hereinafter, with reference to the accompanying drawings will be described the operation of the voice recognition environment control apparatus and method for a voice communication system according to an embodiment of the present invention.

When the user conducts the microphone volume control evaluation, the control and utterance interface (2-3) randomly extracts and requests a sentence to request from the first database (2-6) that stores the voice conversation system domain. The speaker utters the sentence to the voice recognition unit 2-4, receives the feedback, adjusts the microphone volume in the microphone volume adjusting unit 2-5, and stores the user profile data when the adjustment is completed. 2 will be stored in the database (2-7).

In the microphone environment controller 2-8, the score obtained by comparing the sentence transmitted from the speech recognizer 2-4 with the requested sentence in the internal speech recognition evaluator 2-9, and the microphone input level evaluator 2-2-. In consideration of the volume of the microphone input sound measured at the moment of ignition, and the adjustment value of the microphone volume control unit 2-5, the microphone environment evaluator 2-11 has a microphone close or far away, or the volume is high. Low, gives the user feedback such as volume control and re-ignition requests.

In the speech recognition evaluator 2-9, the word error rate score for extracting and comparing the N-Best model of the requested sentence and the recognized sentence, and the voice for determining the similarity between the sentence and the physical signal. The reliability is evaluated in consideration of the confidence score of the recognition unit 2-4.

In the microphone environment evaluator (2-11), the user first requests a re-ignition, performs at least two evaluations, finds a correlation between the microphone input level and the microphone control value, and then evaluates the optimal speech recognition. Calculations to be made and feedback. As more speech is spoken, the data accumulates, and various machine learning techniques can be applied to calculate the best speech recognition score.

As described above, the method and apparatus for controlling a voice recognition environment for a voice conversation system according to the present invention improve the recognition performance by solving a problem caused by the environment of the voice recognition of the voice conversation system, and facilitate a smooth conversation. The voice dialogue system can be applied to any environment where it can be used, so the ripple effect will be great for the relevant devices and industry.

Advantages and features of the present invention, and methods for achieving them will be apparent with reference to the above-described embodiments in detail with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, only the embodiments are to make the disclosure of the present invention complete, and the general knowledge in the technical field to which the present invention belongs. It is provided to fully convey the scope of the invention to those skilled in the art, and the invention is defined only by the scope of the claims. Unless otherwise defined, all terms (including technical and scientific terms) used in the present specification may be used in a sense that can be commonly understood by those skilled in the art. In addition, the terms defined in the commonly used dictionaries are not ideally or excessively interpreted unless they are specifically defined clearly.

1-1 ... Voice Recognition Environment Control Device 1-2 ... Start Environment Control
1-3 ... Device Discovery and Environment Control Initiator
Noise level evaluation section
1-5 ... Mic volume evaluation unit
1-6 ... Environmental control complete
2-1 ... User
2-3 ... ignition and conditioning interface
2-4 ... Voice Recognition Unit
2-5.Microphone volume control
2-6 ... First Database
2-7 ... Second Database
2-8 ... microphone environment control
2-9 ... Voice Recognition Evaluation Unit
Microphone input level evaluation unit
2-11 ... Mike environmental evaluation department

Claims (12)

In the voice recognition environment control apparatus for a voice conversation system,
A device search and environment initialization unit for performing a device search and environment initialization of the user by an environment control start command;
A noise level evaluator for estimating noise from a microphone input level of a microphone searched through the device search before requesting a utterance from a user;
A voice recognition unit for requesting a user to speak a predetermined sentence stored in advance, recognizing a user's speech according to the request, and outputting the sentence spoken by the user as a result;
A microphone volume control unit for adjusting a microphone volume value according to a user's operation; And
A microphone environment controller configured to determine whether a current microphone volume value satisfies a preset criterion based on the user's speech recognized by the voice recognition unit and an adjustment of the microphone volume controller, and feed back the adjustment information of the microphone volume value to the user; Including,
The microphone environment control unit,
A speech recognition evaluator for evaluating a speech recognition score from the sentence recognized by the speech recognizer;
A microphone input level evaluator configured to evaluate a microphone input level obtained during a speech recognition process of the speech recognizer; And
And a microphone environment evaluator configured to evaluate a current microphone environment state from the voice recognition score, the microphone input level, and the microphone volume value and feed back control information of the microphone volume value to the user. delete The method of claim 1,
The speech recognition unit,
Request a speech by randomly extracting the predetermined sentence previously stored in a first database in which a voice conversation system domain model for a speech request is stored and presenting the same to a user;
The microphone volume control unit,
And the microphone volume value adjusted in a second database storing user environment profile data as environment control information. delete The method of claim 1,
The speech recognition evaluation unit,
And a speech recognition similarity is measured by comparing the sentence spoken by the user provided by the speech recognition unit with the sentence requested to speak, and calculating the measurement result as the speech recognition score. The method of claim 1,
The microphone input level evaluation unit,
And a signal physical input amount measured at the instant of the user's speech to calculate an average, and output the calculated average at the microphone input level. In the voice recognition environment control method for a voice conversation system,
A device search and environment initialization step of performing a device search and an environment initialization of a user by an environment control start command;
A noise level estimation step of estimating noise from a microphone input level of a microphone found through the device search before requesting a utterance from a user;
A voice recognition step of requesting a user to speak a predetermined sentence stored in advance, recognizing a user's speech according to the request, and outputting the sentence spoken by the user as a result; And
And a microphone environment control step of determining whether the speech recognized in the speech recognition step and the microphone volume value set by the user satisfy a preset criterion, and feeding back the adjustment information of the microphone volume value to the user.
The microphone environment control step,
A speech recognition evaluation step of evaluating a speech recognition score from a sentence recognized in the speech recognition step;
A microphone input level evaluating step of evaluating a microphone input level obtained in the speech recognition process of the speech recognition evaluating step; And
And a microphone environment evaluation step of evaluating a current microphone environment state from the voice recognition score, the microphone input level, and the microphone volume value and feeding back control information of the microphone volume value to the user. . delete The method of claim 7, wherein
The speech recognition step,
Request a speech by randomly extracting the predetermined sentence previously stored in a first database in which a voice conversation system domain model for a speech request is stored and presenting the same to a user;
The microphone volume adjustment step,
And storing the adjusted microphone volume value as environmental control information in a second database in which user environment profile data is stored. delete The method of claim 7, wherein
The speech recognition evaluation step,
And comparing the sentence spoken by the user provided in the speech recognition step with a sentence requested for speech, and measuring similarity in speech recognition, and calculating the measurement result as the speech recognition score. The method of claim 7, wherein
The microphone input level evaluation step,
And a mean is measured by measuring the amount of signal physical input at the moment of the user's utterance, and outputting the calculated mean at the microphone input level.

Images