KR20100062413A - Method and apparatus for controling speech recognition of telematics apparatus - Google Patents

Abstract

PURPOSE: A method and an apparatus for controlling speech recognition of telematics apparatus are provided to improve the accuracy of a control operation through voice recognition by confirming a detected motion. CONSTITUTION: A voice recognition unit(200) recognizes the voice of a speaker through a voice input unit(1), and a lip reading processor(100) receives the lip shape of the speaker through a photographing unit(2). The lip reading processor recognizes the lip shape of the speaker. According to the matching of lip reading results with the voice recognition results, a multi modal recognition unit(300) outputs a corresponding command for the control of a telematics device(400).

Description

Speech recognition device for telematics device and method thereof {METHOD AND APPARATUS FOR CONTROLING SPEECH RECOGNITION OF TELEMATICS APPARATUS}

The present invention relates to a speech recognition device and a method for a telematics device, and in particular, to recognize the voice of the speaker and the shape of the lips of the speaker at the time of inputting a voice command for the telematics device to increase the reliability of the speech recognition function. A voice recognition device for a telematics device and a method thereof.

The present invention is derived from the research conducted as a part of the development of core technologies for IT new growth engines of the Ministry of Knowledge Economy and the Ministry of Information and Telecommunication Research and Development. [Task name: Development of embedded SW platform technology for future intelligent vehicle electronics].

In general, telematics refers to a combination of telecommunication and information science. In the early days of telematics, a method of connecting a vehicle to a service center using wireless communication and providing a variety of information and services required by the service center during the operation of the vehicle was applied. Such telematics has been developed in the multimedia field such as car radio, tape, car CD, car TV, and the like, and is being developed as an electronic device for providing road information through an information communication field such as navigation technology. In addition, the input / output method of the telematics electronic device has been developed from a simple input method such as a keypad and a button to a method of providing more convenience such as a touch screen.

In recent years, the input method for the telematics device has been increasingly using a voice recognition technology, the multimedia and navigation, etc. installed in the vehicle is using the voice recognition technology as an input method. In the speech recognition technology, a voice of a speaker is received through a voice recognition device, and the received voice is compared with a preset voice pattern to determine the meaning of the corresponding voice to recognize the command. For example, when the speaker's voice is 'instruction', the voice is compared with a preset voice pattern to recognize a command 'instruction'. Thus, the telematics electronic device such as multimedia or navigation by receiving the voice of the speaker is easily operated without the user's separate operation by performing the corresponding command.

However, one of the most problematic parts of speech recognition technology is acoustic noise collected with speech. These acoustic noises range from the drive noise of the device itself to network noise and environmental noise. Thus, the speech recognition method is superior in operability compared to the existing input method, but the accuracy is still not satisfactory. In particular, since telematics electronic devices and the like are used in automobiles, the recognition rate of voice recognition is further reduced if the situation in the vehicle is not quiet. That is, the accuracy of speech recognition is relatively lowered due to ambient noise during speech recognition. Therefore, there is a need for a method for improving the operation accuracy of telematics electronic devices through in-vehicle speech recognition.

In order to solve the above problems, the present invention provides a speech recognition device for a telematics device to increase the accuracy of the speech recognition through the process of detecting and confirming the lip shape to increase the recognition rate of the conventional control device through the speech recognition in the vehicle and its The purpose is to provide a method.

Voice recognition device for a telematics device according to a preferred embodiment of the present invention for solving the above problems is a voice recognition means for receiving a voice of the speaker through the voice input means and recognize the voice of the speaker to generate a voice recognition result; Reading means for receiving the speaker's lip shape through the photographing means and recognizing the speaker's lip shape to generate a reading result; It is characterized in that it comprises a multi-modal recognition means for determining whether the generated speech recognition result and the reading results match, and based on this output as a corresponding voice command for controlling the telematics device.

In addition, the reading processing means receives an image signal photographing the lip shape of the speaker through the photographing means, recognizes the face region of the speaker from the input image signal, and tracks the position of the lips within the face region to detect the speaker's lips. Feature extraction unit for detecting a lip feature for the shape; And a comparison determination unit that compares the lip features detected by the feature extraction unit with a predetermined reading pattern to generate a reading result.

The voice recognition unit may include a voice recognition unit that receives a voice signal generated by converting the voice of the speaker through the voice input unit and recognizes the voice by comparing the input voice signal with a preset voice pattern; And a speech processing unit for processing the speech recognized by the speech recognition unit to generate a speech recognition result.

In addition, the multi-modal recognition means, the multi-modal interface for simultaneously receiving the voice recognition result and the reading results in a multi-modal recognition method; An analysis and determination unit for comparing a voice recognition result and a reading result inputted through a multi-modal interface to determine whether they match; And an operation control unit for outputting a corresponding voice command interlocked when the voice recognition result and the reading result coincide with the determination result in the analysis and determination unit.

On the other hand, the speech recognition method for a telematics device according to a preferred embodiment of the present invention comprises the steps of receiving a voice and lip shape of the speaker; Generating a voice recognition result by recognizing the voice of the received speaker; Recognizing the shape of the received speaker's lips to generate a reading result; And determining whether or not the voice recognition result and the reading result match, and outputting the corresponding voice command for controlling the telematics device based on the matching result.

In addition, the step of generating a reading result by recognizing the lip shape of the speaker, receiving an image signal photographing the lip shape of the speaker; Recognizing a face region of the speaker from the input image signal; Detecting lip features of the speaker's lip shape by tracking the position of the lips within the face area of the speaker; And comparing the lip features with a predetermined reading pattern to generate a reading result.

The generating of the voice recognition result by recognizing the received voice of the speaker may include: receiving a voice signal generated by converting the voice of the speaker; Recognizing a voice by comparing an input voice signal with a preset voice pattern; And processing the recognized voice to generate a voice recognition result.

The outputting of the voice command for controlling the telematics device may include: simultaneously receiving a voice recognition result and a reading result in a multimodal recognition method; Comparing the input speech recognition result with the reading result to determine whether or not a match is made; And outputting a corresponding voice command interlocked with the voice recognition result when the voice recognition result and the reading result coincide.

Here, in the step of simultaneously inputting the voice recognition result and the reading result in a multi-modal recognition method, it is preferable to generate the multi-modal information by receiving the voice recognition result and the reading result at the same time and correcting the time difference for each.

In addition, the voice command is preferably a predetermined control command for controlling the telematics device.

According to the present invention, in addition to the existing in-vehicle telematics terminal, as a secondary device of the user interface so as to facilitate the control by the voice recognition of the control device mounted on the vehicle, the movement of the lip shape is detected and confirmed, thereby recognizing the voice recognition. It has the effect of increasing the accuracy of the control through.

Through this, when the telematics device is controlled by the voice recognition method, it has an excellent effect of improving the reliability of the voice recognition by reducing the probability that the voice recognition fails or misunderstanding due to noise or unclear pronunciation of the speaker.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. Repeated descriptions, well-known functions and configurations that may unnecessarily obscure the subject matter of the present invention, and detailed description of the configuration will be omitted. Embodiments of the present invention are provided to more completely describe the present invention to those skilled in the art. Accordingly, the shape and size of elements in the drawings may be exaggerated for clarity.

1 is a block diagram showing a speech recognition device for a telematics device according to an embodiment of the present invention.

Referring to FIG. 1, a speech recognition apparatus for a telematics apparatus according to an exemplary embodiment of the present invention includes a speech input means 1, a photographing means 2, a reading order means 100, a speech recognition means 200, and a multimodal mode. It comprises a recognition means (300).

The voice input unit 1 serves to receive a voice spoken by the speaker. Preferably, the voice input means 1 may employ a microphone or the like capable of receiving a speaker's sound. The voice input means 1 is continuously receiving voice and collects not only the voice of the speaker but also the noise generated in the surrounding environment. In this way, the voice input means 1 receives the voice of the speaker for voice recognition to generate a voice signal.

The photographing means 2 plays a role of photographing the lip shape of the speaker while the speaker inputs the voice. Preferably, the photographing means 2 may be an image camera, a digital camera, or the like, capable of photographing the face and lips of the speaker. In this way, the photographing means 2 detects the voice of the speaker, preferably the shape of the lip, which produces voice, thereby generating an image signal.

The reading processing means 100 receives an image signal of the speaker's lip shape from the photographing means 2 and generates a corresponding reading result. The reading processing means 100 detects lip features by tracking the shape of the speaker's lips in the input image signal. The reading processing means 100 determines whether the detected lip feature coincides with a predetermined reading pattern and generates a corresponding reading result. That is, when the lip feature detected in the lip shape of the speaker coincides with the preset reading pattern, the reading processing means 100 outputs the reading result for the matching reading pattern. Details of the reading process means 100 may be easily understood by those skilled in the art, and thus detailed description thereof will be omitted.

In other words, the reading processing means 100 recognizes the face area of the speaker from the image signal input from the photographing means 2. The reading processing means 100 detects the lip characteristics of the lip shape of the corresponding speaker by tracking the position of the lips in the recognized speaker's face area. For example, if the speaker's face and background are included in a single frame in the video signal, the reading processing means 100 recognizes the speaker's face area based on the color or shape of the face. In addition, the reading processing means 100 tracks the position of the liquor based on the color and shape of the lips in the face area and extracts the feature of the shape of the lips. The method of extracting such lip features may be applied in any manner.

Thus, the reading process means 100 determines whether the extracted lip feature of the speaker's lip shape matches the reading pattern set in the database and generates a corresponding reading result. The reading processing means 100 searches the database based on the lip characteristics of the speaker to find a reading pattern matching the input lip characteristics. If the lip feature detected in the lip shape of the speaker coincides with the preset reading order pattern, the reading order processing unit 100 generates a reading result for the matching reading pattern. The reading processing means 100 sequentially outputs the reading results.

The voice recognition means 200 recognizes a voice signal generated by receiving voice spoken from the speaker input from the voice input means 1 and outputs a voice recognition result. The voice recognition means 200 recognizes a voice signal for the voice spoken by the speaker, processes the recognized information, and outputs the voice signal as a voice recognition result. For example, the speech recognition means 200 may generate corresponding binary or text data in a manner in which a speech signal is compared with a predetermined speech pattern for a predetermined period, a sound band, or a height.

In other words, the voice recognition means 200 recognizes a corresponding voice by comparing the voice signal received from the voice input means 1 with a preset voice pattern. The voice recognition means 200 processes the recognized voice to generate and output a voice recognition result.

The multi-modal recognition means 300 analyzes and processes the multi-modal information and outputs a single form of information. The multi-modal recognition means 300 reads the reading result obtained by the reading order processing means 100 and the voice output by the speech recognition means 200. Receives the recognition result and compares the reading result and the voice recognition result with each other and plays the role of outputting the final voice command. Here, multimodal refers to a method of receiving two or more different types of inputs at the same time and finally generating an optimal single output. For example, when the output of the voice recognition unit 210 may be "map" or "try", if the lip movement is determined as data information of "map", the output is determined to be the "map" by the user. It means to be done.

In other words, the multi-modal recognition means 300 has a configuration for receiving multi-modal information. The multi-modal recognition means 300 receives two types of input, for example, a reading result of the inventor's lip movement and a voice recognition result of the voice. That is, the two pieces of input information received from the multimodal recognition means 300 are the outputs of the reading device means and the voice recognition means 200, respectively, and have a configuration for receiving them simultaneously with or without time difference.

In addition, the multi-modal recognition means 300 makes a judgment based on the time information and the analyzed information of each of the received input signals (audio and video). The multi-modal recognition stage 300 compares the information of the voice command of the voice recognition result with the information of the voice command of the reading result when the judgment is the same and outputs the corresponding voice command, and if two or more different voice command candidates are present, the database It is determined by inferring the case of the correct voice command through the preset history and the situation in the. Then, a control signal corresponding thereto is generated and output. Thereafter, the control signal is input to the telematics device 400 which can be controlled through the in-vehicle voice recognition to operate the corresponding telematics device 400. For example, when the reading result coinciding with the voice of the speaker is 'start' or 'end', when the voice command 'start' or 'end' is output, the corresponding telematics device takes a corresponding operation. Here, the database is storage means including the above-described voice pattern, voice recognition pattern, voice command information, and the like. Voice command information stored in the database is preferably set according to the applied telematics electronic device. For example, when the present invention is applied to navigation, voice command information stored in a database may include directions information such as place names, roads, stores, and names, and control commands such as start, end, search, and guidance for driving navigation. do. Such a database may be of any configuration provided that the database is configured and arranged in a known storage format. Of course, the database can be configured separately for each configuration, it is natural if it can be integrated.

Here, the telematics device 400 may include all devices such as a navigation system and a multimedia system that can be used in a vehicle as the telematics device.

In this way, as well as the existing in-vehicle telematics terminal, as well as the auxiliary device of the user interface to detect the movement of the lip shape as a user interface to facilitate the control by the voice recognition of the control device mounted on the vehicle, the accuracy of the control is improved. Have

FIG. 2 is a block diagram showing an internal configuration of the reading process means 100 shown in FIG. 1, FIG. 3 is a block diagram showing an internal configuration of the voice recognition means 200 shown in FIG. Is a block diagram showing the internal configuration of the multi-modal recognition means 300 shown in FIG.

Referring to FIG. 2 in more detail with reference to FIG. 2, the reading process means 100 includes a feature extraction unit 110, a comparison determination unit 120, a first output unit 130, and a database ( 140).

The feature extractor 110 processes the image signal input through the photographing means 2 to extract the lip feature of the speaker. The feature extractor 110 recognizes the face area of the speaker from the image signal input from the photographing means 2. The feature extractor 110 detects a lip feature of the speaker's lip shape by tracking the position of the lips in the recognized speaker's face area.

For example, if the speaker's face and background are included in a single frame in the image signal, the feature extractor 110 recognizes the speaker's face area based on the color or shape of the face. The feature extracting unit 110 extracts a feature of a lip shape after tracking a position of a liquor based on the color and shape of the lips in the face region.

The comparison determination unit 120 determines whether the lip feature of the speaker's lip shape extracted from the feature extraction unit 110 matches the reading pattern set in the database 140 and generates a corresponding reading result. When the comparison determination unit 120 receives the lip feature from the feature extractor 110, the comparison determination unit 120 searches the database 140 to find a reading pattern matching the input lip feature. Thus, the comparison determination unit 120 generates a reading result for the matching reading pattern when the lip feature detected in the lip shape of the speaker matches the predetermined reading pattern. Then, the first output unit 130 sequentially outputs the reading result generated by the comparison determination unit 120.

In addition, when the voice recognition means 200 is described in more detail with reference to FIG. 3, the voice recognition means 200 includes a voice recognition unit 210, a voice processing unit 220, and a second output unit 230. It is configured by. The voice recognition unit 210 recognizes a corresponding voice by comparing the voice signal received from the voice input unit 1 with a preset voice pattern. The voice processor 220 processes the voice recognized by the voice recognizer 210 to generate a voice recognition result.

In addition, the multi-modal recognition means 300 described above in more detail with reference to Figure 4, the multi-modal recognition means 300 is a multi-modal interface 310, the analysis and determination unit 320, the database 340, It is configured to include an operation control unit 330.

The multi-modal interface 310 has a configuration for receiving multi-modal information. The multi-modal interface 310 simultaneously receives two types of inputs, for example, a reading result of the inventor's lip movement and a speech recognition result of the voice.

The analysis and determination unit 320 makes a judgment based on the time information and the analyzed information of each input signal (audio and video) received by the multi-modal interface 310. The analysis and determination unit 320 compares the information of the voice command of the voice recognition result and the voice command of the reading result when the judgment is the same and outputs the corresponding voice command. In this case, when two or more different voice command candidates appear, the analysis and determination unit 320 infers the case of the correct voice command based on the history and situation of the speaker in the database 340. Then, the control signal corresponding thereto is generated and output by the operation controller 330. Thereafter, the control signal is input to the telematics device 400 which can be controlled through the in-vehicle voice recognition to operate the corresponding telematics device 400. For example, when the reading result coinciding with the voice of the speaker is 'start' or 'end', the voice command 'start' or 'end' is output.

The database 340 is a storage means including voice command information. Voice command information stored in the database 340 is preferably set according to the applied telematics electronic device. For example, when the present invention is applied to navigation, voice command information stored in a database may include directions information such as place names, roads, stores, and names, and control commands such as start, end, search, and guidance for driving navigation. do.

As described above, when the telematics device is controlled by a voice recognition method, the present invention has an excellent effect of improving the reliability of voice recognition by reducing the probability of failure or misrecognition due to noise or unclear pronunciation of a speaker.

Hereinafter, a speech recognition method for a telematics device according to an embodiment of the present invention will be described.

5 is a flowchart illustrating a voice recognition method for a telematics device according to an embodiment of the present invention. In the description, the same reference numerals shown in FIGS. 1 to 4 are assumed to perform the same functions.

The speech recognition method for the telematics device according to the preferred embodiment of the present invention will be briefly described. It is a way to increase. Through this, as well as the existing in-vehicle telematics terminal, the lip shape movement is detected and confirmed as an auxiliary device of the user interface to facilitate the control by voice recognition of the control device mounted on the vehicle. Height has an effect.

Referring to Figure 5 will be described in more detail voice recognition method for a telematics device according to an embodiment of the present invention. In the description, a voice recognition device for a telematics device according to a preferred embodiment of the present invention described above will be described as an example.

The present invention is disclosed while a talker speaks a voice for driving a telematics device in a vehicle. The voice input means 1 receives the voice spoken by the speaker, and the photographing means 2 captures the shape of the speaker's lips while the speaker inputs the voice (S10). Here, since the voice input means 1 corresponds to a microphone capable of receiving the voice of the speaker, the voice input means 1 continues to receive voice, and not only the voice of the speaker but also noise generated in the surrounding environment. Will be collected. In this way, the voice input means 1 receives the voice of the speaker for voice recognition to generate a voice signal. In addition, when the photographing means 2 is an image camera capable of photographing the face and lip shape of the speaker, the photographing means 2 continuously generates the image signal by continuously detecting the face and lips of the speaker that speaks the voice.

Next, the voice recognition means 200 recognizes a corresponding voice by comparing the voice signal received from the voice input means 1 with a preset voice pattern and the like, and processes the recognized voice to generate a voice recognition result (S20). . In other words, the speech recognition means 200 generates the first speech recognition result of the speech processed by the speaker, and the speech recognition result includes information that is incorrectly recognized by noise in the vehicle or unclear pronunciation of the speaker. There may be.

At this time, the reading process means 100 receives the image signal for the shape of the speaker's lips from the photographing means 2 while generating the corresponding reading result while the voice recognition means 200 recognizes the voice (S30). The reading order processing means 100 processes the image signal input through the photographing means 2, extracts the lip feature of the speaker, and then outputs the reading result by comparing with the preset reading order pattern.

In other words, the reading processing means 100 determines whether or not the extracted lip feature of the speaker's lip shape coincides with the reading pattern set in the database, and generates a corresponding reading result. That is, the reading order processing means 100 searches the database to find a reading pattern matching the input lip feature. Thus, if the lip feature detected in the lip shape of the speaker coincides with the predetermined reading order pattern, the reading order processing unit 100 generates reading results for the matching reading order pattern sequentially.

Here, if the image signal photographed by the photographing means 2 includes a face, a background, etc. together with the shape of the lip of the speaker, first, the reading order means 100 may determine the speaker's speech from the image signal input from the photographing means 2. Recognize facial areas. In addition, the reading processing means 100 detects lip characteristics of the lip shape of the corresponding speaker by tracking the position of the lips in the recognized speaker's face region. For example, if the face and background of the speaker are included in a single frame in the video signal, the face region of the speaker is recognized based on the color or shape of the face. Then, the position of the liquor is tracked based on the color and shape of the lips in the face area, and the features of the shape of the lips are extracted.

Subsequently, the multi-modal recognition means 300 receives the voice recognition result and the reading result simultaneously (S40). That is, the multi-modal recognition means 300 receives two types of inputs, for example, a reading result of the inventor's lip movement and a voice recognition result of the voice. That is, the two pieces of input information received from the multimodal recognition means are output from the reading processing means 100 and the voice recognition means 200, respectively.

At this time, the multi-modal recognition means 300 corrects the time difference between the input voice recognition result and the reading result. This time correction can be time corrected by comparing the number of data of the speech recognition result with the reading result or by inputting the start and end flags in the respective information.

Next, the multi-modal recognition means 300 receives the reading results obtained by the reading order processing means 100 and the speech recognition results issued by the speech recognition means 200, and compares the reading results and the speech recognition results with each other ( S50). Therefore, the multi-modal recognition means 300 compares the information of the voice command of the voice recognition result with the information of the voice command of the reading result when the determination is the same and outputs the corresponding voice command (S70). Then, the control signal corresponding thereto is generated and output by the operation controller 330 (S70). Thereafter, the control signal is input to the telematics device 400 which can be controlled through the in-vehicle voice recognition to operate the corresponding telematics device 400. For example, when the reading result coinciding with the voice of the speaker is 'start' or 'end', the voice command 'start' or 'end' is output. Here, as described above, the voice command information stored in the database is set according to the applied telematics electronic device. For example, when the present invention is applied to navigation, the voice command information stored in the database may be a place name, a road, a store, a human name, or the like. This includes control commands such as start, end, search, and guidance for driving the navigation information and navigation.

On the other hand, in step S50, when the multimodal recognition means 300 comes out with two or more different voice command candidates, the multimodal recognition means 300 infers the case of the correct voice command based on the history and situation of the speaker in the database (S60). Then, the control signal corresponding thereto is generated and output by the operation controller 330 (S70). Thereafter, the control signal is input to the telematics device 400 which can be controlled through the in-vehicle voice recognition to operate the corresponding telematics device 400. For example, when there is no reading result that matches the voice of the speaker, if the reading result is 'start' or 'end', the voice command 'start' or 'end' is output.

As described above, in addition to the existing in-vehicle telematics terminal, the lip shape movement is detected and confirmed as an auxiliary device of the user interface to facilitate the control by the voice recognition of the control device mounted on the vehicle. Has the effect of raising.

Through this, when the telematics device is controlled by the voice recognition method, it has an excellent effect of improving the reliability of the voice recognition by reducing the probability that the voice recognition fails or misunderstanding due to noise or unclear pronunciation of the speaker.

As mentioned above, although one preferred embodiment of the present invention has been illustrated and described, the present invention is not limited to the specific embodiments described above, and the present invention belongs to the present invention without departing from the gist of the present invention as claimed in the claims. Various modifications may be made by those skilled in the art, and these modifications should not be individually understood from the technical spirit or the prospect of the present invention.

1 is a block diagram showing a telematics speech recognition device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an internal configuration of the reading process means shown in FIG.

3 is a block diagram showing the internal configuration of the voice recognition means shown in FIG.

4 is a block diagram showing the internal configuration of the multi-modal recognition means shown in FIG.

5 is a flowchart illustrating a telematics speech recognition method according to an embodiment of the present invention.

<Description of main parts of drawing>

1: Voice input means 2: Shooting means

100: poisonous processing means 110: feature extraction unit

120: comparison determination unit 130: first output unit

200: speech recognition means 210: speech recognition unit

220: voice processing unit 230: second output unit

300: multimodal recognition means 310: multimodal interface

320: analysis and determination unit 330: motion control unit

140, 340: Database 400: Telematics device

Claims (10)

In a speech recognition device for a telematics device, Voice recognition means for receiving a voice of a speaker through voice input means and recognizing the voice of the speaker to generate a voice recognition result; Reading means for receiving a lip shape of the speaker through the photographing means and recognizing the lip shape of the speaker to generate a reading result; And a multi-modal recognition means for determining whether the generated speech recognition result and the reading result are matched, and outputting the corresponding speech command for controlling the telematics device. The method according to claim 1, The reading process means, Receiving an image signal photographing the lip shape of the speaker through the photographing means, recognizes the face region of the speaker from the input image signal, and tracks the position of the lips within the face region to characterize the lips of the speaker. Feature extraction unit for detecting; And Speech recognition device for a telematics device, characterized in that it comprises a comparison determination unit for comparing the lip feature detected by the feature extraction unit and a predetermined reading order pattern to generate a reading result. The method according to claim 1, The voice recognition means, A voice recognition unit receiving a voice signal generated by converting a voice of a speaker through the voice input unit and recognizing the voice by comparing the input voice signal with a preset voice pattern; And And a speech processing unit for processing the speech recognized by the speech recognition unit to generate a speech recognition result. The method according to claim 1, The multimodal recognition means, A multi-modal interface for simultaneously receiving the voice recognition result and the reading result in a multi-modal recognition method; An analysis and determination unit comparing the voice recognition result and the reading order inputted through the multi-modal interface to determine whether they match; And And a motion control unit for outputting a corresponding voice command interlocked when the voice recognition result and the reading result coincide with the determination result of the analysis and determination unit. In the speech recognition method for a telematics device, Receiving voice and lip of a speaker; Generating a voice recognition result by recognizing the voice of the received speaker; Recognizing the shape of the lips of the received speaker to generate a reading result; And And determining whether the voice recognition result and the reading result match, and outputting the corresponding voice command for controlling the telematics device based on the matching result. The method according to claim 5, Recognizing the shape of the lips of the received speaker to generate a reading result, Receiving an image signal photographing the lips of the speaker; Recognizing a face region of a speaker from the input image signal; Detecting lip characteristics of the speaker's lip shape by tracking the position of the lips within the face area of the speaker; And And generating a reading result by comparing the lip features with a predetermined reading pattern. The method according to claim 5, Recognizing the received voice of the speaker and generating a voice recognition result, Receiving a voice signal generated by converting a voice of a speaker; Recognizing a voice by comparing the input voice signal with a preset voice pattern; And Processing the recognized voice to generate a voice recognition result. The method according to claim 5, The step of outputting the voice command for controlling the telematics device, Simultaneously receiving the voice recognition result and the reading result in a multi-modal recognition method; Comparing the input voice recognition result with the reading result to determine whether or not a match exists; And And outputting a corresponding voice command interlocked with the voice recognition result when the voice recognition result and the reading result coincide with each other. The method according to claim 8, Simultaneously receiving the voice recognition result and the reading result in a multi-modal recognition method, Speech recognition method for a telematics device, characterized in that for receiving the voice recognition result and the reading result at the same time and correcting the time difference for each of the multi-modal information. The method according to claim 5 or 8, The voice command is a voice recognition method for a telematics device, characterized in that the preset control command for the control of the telematics device.

Images