KR102394510B1 - Apparatus and method for recognizing voice in vehicle - Google Patents

Abstract

The present invention proposes a vehicle voice recognition apparatus and method for removing a beep sound from sound information input through a microphone and recognizing a user's voice using a user-specified database. An apparatus according to the present invention includes: a beep removal unit configured to obtain a user's voice by removing a beep sound from the sound information when sound information is input by a user in a vehicle; a user information acquisition unit for acquiring information about the user; a vocabulary list detection unit that detects a vocabulary list from a database based on user information; and a voice recognition unit for recognizing the user's voice based on the vocabulary list.

Description

Apparatus and method for recognizing voice in vehicle}

The present invention relates to a voice recognition apparatus and method. More particularly, it relates to a voice recognition apparatus and method for a vehicle.

In general, many devices are installed in a vehicle, and recently, devices for the convenience of drivers or passengers have been continuously developed and installed. A typical example is an audio device or an air conditioning device, and furthermore, the number of vehicles installed with a navigation system that guides a route to a destination by locating the vehicle through a satellite is also increasing.

However, while it is true that such convenience devices provide various conveniences to the driver and passengers, there is a problem in that the concentration of driving decreases and the risk of accidents increases due to distraction while the driver operates the convenience devices while driving. .

In order to solve this problem, recently, a voice recognition device for a vehicle has been developed that recognizes the driver's voice and controls various convenience devices according to the driver's voice command when the driver commands by voice without the need to directly manipulate the convenience devices.

However, since there is only a microphone for recognizing the driver's voice in the conventional voice recognition apparatus for a vehicle, noise generated inside or outside the vehicle is input into the microphone together with the voice, so that voice control is not properly performed.

Korean Patent Application Laid-Open No. 2013-0063091 proposes a voice recognition device for a vehicle. However, since this device is selectively activated as needed to recognize a user's voice command, the above problem cannot be solved.

The present invention has been devised to solve the above problems, and it is an object of the present invention to propose a vehicle voice recognition apparatus and method for removing a beep sound from sound information input through a microphone and recognizing a user's voice using a user-specified database do it with

However, the object of the present invention is not limited to the above, and other objects not mentioned will be clearly understood by those skilled in the art from the following description.

The present invention has been devised to achieve the above object, and when sound information is input by a user in a vehicle, a beep sound removing unit for obtaining the user's voice by removing a beep sound from the sound information; a user information obtaining unit obtaining information about the user; a vocabulary list detector configured to detect a vocabulary list from a database based on the information on the user; and a voice recognition unit for recognizing the user's voice based on the vocabulary list.

Preferably, the beep sound removing unit removes noise along with the beep sound from the sound information.

Preferably, the beep canceling unit removes the beep sound using an echo canceller, and removes the noise using a noise reduction.

Preferably, the beep removal unit uses an echo canceller including an adaptive filter, a residual echo suppressor, and a comfort noise generator as the echo canceller.

Preferably, the beep removal unit removes the beep sound from the sound information first and then removes the noise.

Preferably, the beep removal unit converts a stereo component into a mono component with respect to the audio data included in the sound information, and uses down sampling to make the frequency of each component included in the sound information the same. conversion to remove the beep sound from the sound information.

Preferably, the vocabulary list detector collects and collects at least one of words, sentences, and commands recorded as uttered by the user more than a reference number of times from the database, and detects it as the vocabulary list.

Preferably, the vehicle voice recognition apparatus is driven when the spoken voice is input from the user after the beep sound is reproduced.

In addition, the present invention includes the steps of: when sound information is input by a user in a vehicle, removing a beep sound from the sound information to obtain the user's voice; obtaining information about the user; detecting a vocabulary list from a database based on the information about the user; and recognizing the user's voice based on the vocabulary list.

Preferably, the step of obtaining the user's voice removes noise as well as the beep sound from the sound information.

Preferably, in the obtaining of the user's voice, the beep sound is removed using an echo canceller, and the noise is removed using a noise reduction.

Preferably, in the step of obtaining the user's voice, an echo canceller including an adaptive filter, a residual echo suppressor, and a comfort noise generator as the echo canceller is used. use it

Preferably, in the obtaining of the user's voice, the beep sound is first removed from the sound information and the noise is removed later.

Preferably, the step of obtaining the user's voice converts a stereo component into a mono component with respect to the audio data included in the sound information, and uses down sampling of each component included in the sound information. The beep sound is removed from the sound information by converting the same frequency.

Preferably, in the detecting, at least one of words, sentences, and commands recorded as uttered by the user more than a reference number of times from the database is collected and collected and detected as the vocabulary list.

Preferably, the vehicle voice recognition method is performed when a spoken voice is input from the user after the beep sound is reproduced.

The present invention can obtain the following effects by removing a beep sound from sound information input through a microphone and recognizing the user's voice using a user-specified database.

First, when a command is uttered early, a recognition operation is possible even if it is mixed with a beep sound, so that the user's sensible performance is improved. In addition, by optimizing the recognition command for each user, it is possible to see the improvement of the recognition performance optimized for the user.

Second, the usability of voice recognition is improved so that a novice user of voice recognition or a user who is in a hurry can recognize even if the timing of utterance is wrong. In addition, it is possible to present user-optimized usability by improving recognition performance based on commands mainly used by the user.

Third, while the existing ignition timing is limited, it is possible to give a change in the ignition timing, so that a system with a timing suitable for a user can be designed. In addition, when a user-specific command list is provided, it is possible to infer which functions are frequently used by the user.

Fourth, it not only removes beeps, but also cancels ambient noise to deliver a clean voice rather than a complex voice to the voice recognition engine.

Fifth, the recognition speed can be improved by providing a list corresponding to the user when recognizing commands.

Sixth, when a clean voice is transmitted to the engine, the reliability of the recognition operation is improved.

Seventh, it is economical and not harmful to the environment at all because it is developed on a system that uses the eco-canceller as SW.

1 is a conceptual diagram illustrating a voice pre-processing system according to an embodiment of the present invention.
FIG. 2 is a conceptual diagram illustrating a SW hierarchy of the voice preprocessing system shown in FIG. 1 .
3 is a diagram comparing analysis results of a signal input through a microphone and a signal from which an echo is removed.
4 is an operation block diagram of an echo canceller constituting a voice pre-processing system.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. First of all, it should be noted that in adding reference numerals to the components of each drawing, the same components are given the same reference numerals as much as possible even though they are indicated on different drawings. In addition, in describing the present invention, if it is determined that a detailed description of a related known configuration or function may obscure the gist of the present invention, the detailed description thereof will be omitted. In addition, preferred embodiments of the present invention will be described below, but the technical spirit of the present invention is not limited thereto and may be variously implemented by those skilled in the art without being limited thereto.

Speech recognition is an item that requires an emotional approach rather than a logical one, and users often complain emotionally if they feel that recognition is difficult or difficult to use after using it. In other words, it is judged that there are many problems in terms of the sensibility index that customers feel, and they complain.

Among these IQS user complaints, many complaints about the usability of voice recognition are because not only the user's voice but also various noises and beep sounds are mixed in the voice information input through the microphone.

In addition, in the conventional voice recognition system, commands used by the user among all commands for voice recognition are limited, whereas the recognition rate is inevitably limited because all commands compete for recognition under the same conditions every time.

The present invention has been devised to solve these problems, and proposes a system for removing a beep sound from a user's voice by using an echo canceller during a voice recognition operation and recognizing the user's voice using a user-specified database.

It will be described in more detail below with reference to the drawings.

1 is a conceptual diagram illustrating a voice pre-processing system according to an embodiment of the present invention.

The voice pre-processing system 100 according to the present invention is a pre-processing system that removes beeps from the speaker's voice using an echo canceller 125 . That is, the voice preprocessing system 100 removes the beep sound from the speaker's voice when the novice speaker speaks together with the voice recognition beep by using the echo canceller 125 , and then transmits the beep sound to the recognition engine 150 . The voice pre-processing system 100 uses the voice characteristics of each speaker to classify the speakers, and configures a recognition DB mainly for commands used by each user to increase the usability of voice recognition.

1 , the voice preprocessing system 100 includes a microphone 105 , an audio data buffer 110 , a type conversion unit 115 , a sampling unit 120 , an echo canceller 125 , and noise reduction; 130), a speaker determination unit 135, a recognition command list construction unit 140 for each speaker, a recognition command list calling unit 145 for each speaker, and a recognition engine 150.

The microphone 105 performs a function of receiving a voice input from the driver. At this time, the driver's voice, beeps, ambient noise, etc. are inputted to the microphone 105 and transmitted to the audio data buffer 110 .

The audio data buffer 110 temporarily stores sound information in which the driver's voice, beeps, ambient noise, etc. are mixed in the form of digital data. Thereafter, the audio data buffer 110 transmits sound information to the type conversion unit 115 .

The type conversion unit 115 performs a function of converting a stereo characteristic into a mono characteristic with respect to audio data (eg, a driver's voice) constituting the sound information.

The sampling unit 120 performs a function of equally converting the frequencies of each component constituting the sound information. The sampling unit 120 converts the same frequencies of the driver's voice and beeps to 16 KHz using a down sampling technique.

The echo canceller 125 performs a function of removing a beep sound from the sound information passed through the type conversion unit 115 and the sampling unit 120 .

The noise reduction 130 performs a function of removing noise from the sound information from which the beep sound has been removed. In the present embodiment, the echo canceller 125 first removes the beep sound from the sound information, and then the noise reduction 130 removes the noise, but it is not necessarily limited thereto. It is also possible to perform a function before (125).

The recognition engine 150 performs a function of receiving, from the noise reduction 130 , a clean voice remaining after the beep and noise are removed from the sound information, that is, the driver's pure voice. When the driver's voice is received by the recognition engine 150 , the speaker determination unit 135 , the speaker recognition command list construction unit 140 , the speaker recognition command list call unit 145 , and the like are sequentially driven.

The recognition engine 150 determines the driver's recognition command based on the list of recognized commands for each speaker obtained from sequential driving of the speaker determining unit 135, the speaker-specific recognition command list construction unit 140, and the speaker-specific recognition command list calling unit 145. It performs the function of recognizing voice.

When the driver's pure voice is input to the recognition engine 150 , the speaker determining unit 135 receives personal information such as an ID from the driver to determine the speaker.

When a speaker is identified by the speaker identification unit 135, the speaker-specific recognition command list construction unit 140 detects commands recorded as having been uttered by the speaker in the database and configures the list into a list.

The per-speaker recognition command list calling unit 145 performs a function of transferring the list generated by the per-speaker recognition command list construction unit 140 to the recognition engine 150 .

FIG. 2 is a conceptual diagram illustrating a SW hierarchy of the voice preprocessing system shown in FIG. 1 . In FIG. 2 , a method of optimizing speech recognition for each speaker based on the speaker's (driver's) pure voice from which beeps and noises have been removed and the speaker-specific voice recognition command list will be described. The following description refers to FIGS. 1 and 2 .

The audio data buffer 110 receives sound information such as a voice recognition beep sound and ambient noise in addition to the speaker's voice input through the microphone (MIC) 105 through the Audio Data Buffer Management 235 . The audio data buffer 110 receives sound information in the form of audio data through an audio driver: codec 240 .

Thereafter, the type conversion unit 115 converts the speaker's voice and the beep sound constituting the sound information from the stereo type to the mono type through the first layer 220 and the second layer 230 .

Thereafter, the sampling unit 120 down-samples the sampling frequency of the beep sound from 44.1 KHz to 16 KHz through the third layer 225 .

Thereafter, the echo canceller 125 removes beeps from the speaker's voice according to the existing beep frequency characteristics from the audio data including ambient noise through the echo canceling layer 215 .

Meanwhile, the noise reduction 130 removes ambient noise that exists after the beep is removed through the echo canceling layer 215 .

Thereafter, the recognition engine 150 receives the voice data from which the beep sound and the noise are removed through the OS framework layer 210 .

Thereafter, the speaker determining unit 135 determines whether the speaker is an existing speaker or a new speaker based on the speaker characteristics output from the recognition engine 150 through the application layer 205 .

Thereafter, the recognition command list construction unit 140 for each speaker creates a new speaker voice recognition list if it is a new speaker through the application layer 205 , and calls a voice recognition list corresponding to the speaker if it is an existing speaker.

Thereafter, when the recognition engine 150 receives the voice recognition list from the speaker-specific recognition command list calling unit 145 , the recognition engine 150 recognizes the commands with priority in the voice recognition list through the application layer 205 .

FIG. 3 is a diagram comparing analysis results of a signal input through a microphone and a signal from which an echo is removed.

3A is an analysis result of a signal input through a microphone. In FIG. 3A , reference numeral 310 denotes a TTS guidance prompt section (about 1.8 seconds), and reference numeral 320 denotes a section 320 in which a spoken voice (Dial Number; 321) and a beep sound 322 are mixed. .

As shown in (a) of FIG. 3 , when a signal input through a microphone is analyzed, a beep sound 322 is present in the middle of the speaker's voice 321, so that voice recognition may fail.

3B is an analysis result of a signal from which an echo has been removed. In FIG. 3B , reference numeral 330 denotes a filter convergence section (about 0.5 seconds), and reference numeral 350 denotes a recognizable section transmitted to the recognition engine. Reference numeral 340 denotes an extendable section of the recognizable section.

As shown in (b) of FIG. 3 , when the present invention is applied, it can be confirmed that only the speaker's pure voice is detected because the beep sound is effectively removed according to the echo cancellation.

4 is an operation block diagram of an echo canceller constituting a voice pre-processing system.

The MIC 405 performs a function of receiving a user's utterance (s(n): utterance). At this time, an echo signal (d(n): echo) including a beep from the SPEAKER 410 is also input to the MIC 405 .

The adaptive filter (Adaptive Filter h(n); 425), the sound emitted through the SPEAKER 410, is introduced back into the MIC 405 in the form of echo, and performs a function of predicting and removing this echo signal through adaptation. .

The detection and control unit 420 performs a double talk detection function on a signal input to the MIC 405 . Also, the detection and control unit 420 adjusts the adaptation rate so that the adaptive filter 425 can predict and remove the echo signal through adaptation, and provides it to the adaptive filter 425 .

The residual echo suppressor (Residual Echo Suppressor) 415 performs a function of canceling an echo that is not sufficiently removed by the adaptive filter 425 .

A silence section exists due to the echo cancellation, and therefore the noise characteristic is not constant. In order to prevent this, the communication noise generator 430 generates and adds comfort noise similar to background noise.

Even beginners or intermediate users who are using voice recognition for the first time, if they have a quick temper or are unfamiliar with voice recognition, usually have to hear a beep when uttering a voice recognition command in order to recognize it normally.

However, in the case where only a guide message is heard and uttered without waiting for a beep sound, voice recognition fails. In other words, voice recognition users do not blame their inexperience in use, but complain that the system does not recognize them.

A summary of the above is as follows.

Use Case 1 : In case of firing after beep sound → Recognition success

Use Case 2 : In case of firing before beep sound → However, it is difficult to occur because the interval between microphone open time and beep sound is short as 30ms → Recognition failure

Use Case 3 : In case of firing while beep sound is playing → Recognition failure

In the case of the present invention, by applying the present invention as the time between the guidance message reproduction and the beep sound reproduction, the microphone opening time can be implemented to achieve recognition success in Use Cases 2 and 3.

An embodiment of the present invention has been described above with reference to FIGS. 1 to 4 . Hereinafter, a preferred embodiment of the present invention that can be inferred from such an embodiment will be described.

A voice recognition apparatus for a vehicle according to a preferred embodiment of the present invention includes a beep removal unit, a user information acquisition unit, a vocabulary list detection unit, a voice recognition unit, a power supply unit, and a main control unit.

The power supply unit performs a function of supplying power to each component constituting the vehicle voice recognition device. The main controller performs a function of controlling the overall operation of each component constituting the vehicle voice recognition device. Considering that the vehicle voice recognition device is provided in the vehicle, the power supply unit and the main control unit can be replaced by the vehicle battery and the MCU (or ECU), so the power supply unit and the main control unit do not need to be provided in this embodiment.

When sound information is input by the user in the vehicle, the beep sound removal unit performs a function of acquiring the user's voice by removing a beep sound from the sound information.

The beep removal unit may remove noise along with the beep sound from the sound information.

The beep removal unit may remove the beep sound using an echo canceller, and may remove the noise using noise reduction.

The beep canceller may use an echo canceller including an adaptive filter, a residual echo suppressor, and a comfort noise generator as an echo canceller.

The beep removal unit may first remove the beep sound from the sound information and then remove the noise.

The beep removal unit converts a stereo component into a mono component with respect to the audio data included in the sound information, and converts the frequency of each component included in the sound information to the same frequency by using down sampling, and converts the beep from the sound information. sound can be removed.

The user information obtaining unit performs a function of obtaining information about the user.

The vocabulary list detector detects the vocabulary list from the database based on information about the user.

The vocabulary list detection unit may collect and collect at least one of words, sentences, and commands recorded as utterances by the user more than a reference number of times from the database, and detect it as a vocabulary list. To this end, the vocabulary list detector may increase the number of utterances for the corresponding vocabulary by one whenever the user's uttered voice is recognized.

The voice recognition unit performs a function of recognizing the user's voice based on the vocabulary list.

The above-described vehicle voice recognition apparatus may be driven when a spoken voice is input from the user after the beep sound is reproduced.

Next, a method of operating the voice recognition device for a vehicle will be described.

First, when sound information is input by the user in the vehicle, the beep removal unit removes a beep sound from the sound information to obtain the user's voice.

Thereafter, the user information acquisition unit acquires information about the user.

Thereafter, the vocabulary list detection unit detects the vocabulary list from the database based on the information about the user.

Thereafter, the voice recognition unit recognizes the user's voice based on the vocabulary list.

Even though all the components constituting the embodiment of the present invention described above are described as being combined or operated in combination, the present invention is not necessarily limited to this embodiment. That is, within the scope of the object of the present invention, all the components may operate by selectively combining one or more. In addition, all of the components may be implemented as one independent hardware, but a part or all of each component is selectively combined to perform some or all of the functions of the combined hardware in one or a plurality of hardware program modules It may be implemented as a computer program having In addition, such a computer program is stored in a computer readable media such as a USB memory, a CD disk, a flash memory, etc., read and executed by a computer, thereby implementing an embodiment of the present invention. The computer program recording medium may include a magnetic recording medium, an optical recording medium, a carrier wave medium, and the like.

In addition, all terms including technical or scientific terms have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs, unless otherwise defined in the detailed description. Terms commonly used, such as those defined in the dictionary, should be interpreted as being consistent with the contextual meaning of the related art, and are not interpreted in an ideal or excessively formal meaning unless explicitly defined in the present invention.

The above description is merely illustrative of the technical idea of the present invention, and various modifications, changes, and substitutions are possible within the range that does not depart from the essential characteristics of the present invention by those of ordinary skill in the art to which the present invention pertains. will be. Accordingly, the embodiments disclosed in the present invention and the accompanying drawings are for explaining, not limiting, the technical spirit of the present invention, and the scope of the technical spirit of the present invention is not limited by these embodiments and the accompanying drawings . The protection scope of the present invention should be construed by the following claims, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of the present invention.

Claims (13)

a beep sound removing unit for obtaining the user's voice by removing a beep sound from the sound information when sound information is input by the user in the vehicle;
a user information obtaining unit obtaining information about the user;
a vocabulary list detector configured to detect a vocabulary list from a database based on the information on the user; and
A voice recognition unit for recognizing the user's voice based on the vocabulary list,
The beep sound canceling unit removes the beep sound using an echo canceller,
The echo canceller may include: an adaptive filter for predicting and removing an echo signal from the sound information through adaptation;
a detection and control unit that adjusts an adaptation rate and provides it to the adaptive filter so that the adaptive filter can predict and remove the echo signal;
a residual echo suppressor that removes a residual echo that has not been removed by the adaptive filter; and
a communication noise generator that generates a comfort noise similar to background noise and adds it to the output of the residual echo suppressor;
The beep removal unit converts a stereo component into a mono component with respect to the audio data included in the sound information, and equally converts the frequency of each component included in the sound information by using down sampling, the remove the beep sound from sound information,
Voice recognition for a vehicle, characterized in that by applying the microphone opening time, at which sound information is input by the user, as a time between the reproduction of the guidance message and the reproduction of the beep sound, the user's voice can be recognized even if the timing of the speech is mismatched. Device. The method of claim 1,
The voice recognition apparatus for a vehicle, characterized in that the beep removal unit removes noise along with the beep sound from the sound information. 3. The method of claim 2,
The voice recognition apparatus for a vehicle, wherein the beep removal unit removes the noise by using noise reduction. delete 3. The method of claim 2,
The beep sound removing unit first removes the beep sound from the sound information and then removes the noise later. delete The method of claim 1,
The vocabulary list detector collects and collects at least one of words, sentences, and commands recorded as uttered by the user more than a reference number of times from the database, and detects the vocabulary list as the vocabulary list. The method of claim 1,
The in-vehicle voice recognition device is driven when the spoken voice is input from the user after the beep sound is reproduced. obtaining, by a beep sound controller, a beep sound from the sound information when sound information is input by the user in the vehicle, to obtain the user's voice;
obtaining, by a user information obtaining unit, information about the user;
detecting, by a vocabulary list detection unit, a vocabulary list from a database based on the information about the user; and
A voice recognition unit comprising the step of recognizing the user's voice based on the vocabulary list,
The step of obtaining the user's voice includes removing the beep sound using an echo canceller,
The echo canceller includes an adaptive filter that predicts and removes an echo signal from the sound information through adaptation, and an adaptation rate so that the adaptive filter predicts and removes the echo signal. A detection and control unit provided to the adaptive filter by adjusting and a comfort noise generator that adds to the output of the residual echo suppressor;
In the step of obtaining the user's voice by removing the beep sound,
The beep removal unit converts a stereo component into a mono component with respect to the audio data included in the sound information, and equally converts the frequency of each component included in the sound information by using down sampling, the remove the beep sound from sound information,
Voice recognition for a vehicle, characterized in that by applying a microphone open time point at which sound information is input by the user as a time between a guidance message reproduction and a beep sound reproduction, the user's voice can be recognized even if the speech timing is wrong Way. 10. The method of claim 9,
The obtaining of the user's voice comprises removing noise as well as the beep sound from the sound information. 11. The method of claim 10,
The obtaining of the user's voice comprises removing the noise using noise reduction. delete 11. The method of claim 10,
The acquiring of the user's voice comprises first removing the beep sound from the sound information and then removing the noise later.

Images