KR100913130B1 - Method and Apparatus for speech recognition service using user profile - Google Patents

Abstract

The present invention provides a method of receiving a voice signal from a user terminal, analyzing user identification information corresponding to the voice signal, and recognizing the voice signal using a finite state network (FSN) and a dictionary corresponding to the analyzed user identification information. It provides a speech recognition method of a large-capacity speech recognition device comprising the step.

Speech recognition, profile

Description

Method and device for speech recognition service using user profile {Method and Apparatus for speech recognition service using user profile}

1 is a general flow diagram for providing a service using a speech recognition system.

2 is a flowchart for providing a speech recognition service using a conventional large capacity speech recognition apparatus.

3 is a flowchart illustrating another method for providing a voice recognition service using an existing large capacity speech recognition apparatus.

4 is a conceptual diagram of a system according to an embodiment to which the present invention is applied.

5 is a flowchart of processing a service in a large-capacity speech recognition apparatus according to an exemplary embodiment of the present invention.

6 is a flowchart for storing user profile information in a mass speech recognition device according to an embodiment of the present invention.

7 is a diagram illustrating an embodiment of user profile information according to an embodiment of the present invention.

8 is a flow diagram of a voice recognition service using a user profile in an embodiment of the present invention.

<Explanation of symbols for the main parts of the drawings>

401: user terminal

403: speech recognition device

405: Service device

407 network

The present invention relates to a voice recognition service method and apparatus using a user profile.

The large speech recognition system takes a long time to output recognition results because there are more than hundreds of thousands of registered vocabularies, and it is difficult to respond in real time due to the large amount of calculations. Worse

 Since a user cannot know what speech to provide a speech recognition service from a system point of view, a speech that can be spoken must be defined as a registered vocabulary in order to provide a service through speech recognition. The registered vocabulary is designed to be flexible by designing the order in which the user needs to speak in a multi-step manner or by a one-step method according to the grammar and service scenario defined in the system.

In the case of designing in a multi-stage manner according to the conventional service scenario, the service usage time is long because a multi-stage procedure is used every time a service is used from the user's point of view. As a concept of changing a task according to a service scenario from a system point of view, since the number of words to be recognized is limited according to the task, the real-time response is fast and the recognition performance is generally good. In this way, the user is asked to speak out in isolation form. Thus, for complex services, the steps are complicated.

Another method of the conventional method is to design a large-capacity speech recognition system in one step, that is, when a user uses a service at a time. This method is used to complement the multi-level method. Since the service is used in one step from the user's point of view, the service usage time is fast, but the real-time response is difficult due to the large number of words to be recognized and the recognition performance is not good.

An object of the present invention is to provide a voice recognition service method and apparatus using a user profile.

Another object of the present invention is to increase the speed and accuracy of the voice recognition service by using the profile information prepared in advance.

In order to achieve the above objects, according to an aspect of the present invention, receiving a voice signal from a user terminal, analyzing the user identification information corresponding to the voice signal and the FSN corresponding to the analyzed user identification information A voice recognition method of a large-capacity speech recognition apparatus may be provided, which includes recognizing the voice signal using a finite state network and a dictionary.

In a preferred embodiment, the method may further include sending a service request signal corresponding to the recognized voice signal to a service device. The method may further include receiving a service access signal from the user terminal, sending a user profile information request signal to the user terminal in response to the service access signal, receiving user profile information from the user terminal, The method may further include storing user profile information and generating a finite state network (FSN) and a dictionary according to the stored user profile information.

The user profile information may include at least one of a finite state network (FSN) and at least one word that can be generated in a dictionary.

According to another aspect of the present invention, means for receiving a voice signal from a user terminal, means for analyzing user identification information corresponding to the voice signal, and a finite state network (FSN) and dictionary corresponding to the analyzed user identification information It is possible to provide a large-capacity speech recognition apparatus comprising a means for recognizing the speech signal using.

In a preferred embodiment, the method may further include means for sending a service request signal corresponding to the recognized voice signal to a service device. Further, means for receiving a service access signal from the user terminal, means for sending a user profile information request signal to the user terminal in correspondence with the service access signal, means for receiving user profile information from the user terminal, the received The apparatus may further include means for storing user profile information and means for generating a finite state network (FSN) and a dictionary corresponding to the stored user profile information.

The user profile information may include at least one of a finite state network (FSN) and at least one word that can be generated in a dictionary.

Next, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a general flowchart for providing a service using a voice recognition system.

Referring to FIG. 1, first, the user terminal 110 attempts an access request to a service device to receive a service using a voice recognition system (step 101). The service device connection request signal 101 is then transmitted to the voice recognition system 130 via the communication network 120 (step 103). The voice recognition system 130 then sends a guidement for using the service device 140 to the user terminal (step 105), and the guidement is transmitted to the user terminal 110 through the communication network 120 (step 107).

The user terminal 110 then sends a voice for receiving the service according to the announcement (step 109), and this voice information is transmitted to the voice recognition system 130 through the communication network 120 (step 111). Upon receiving the voice information, the voice recognition system 130 processes the voice based on the received voice information to obtain necessary information (step 113), and transmits the service requested by the user terminal to the service device 140 (step 115). ). The service device 140 then transmits the service requested by the user terminal to the user terminal (step 117).

Such a speech recognition service has a disadvantage in that its accuracy depends on the performance of the speech recognition system, and in particular, the performance of speech recognition is poor when the connected words are processed.

2 is a flowchart for providing a speech recognition service using a conventional large capacity speech recognition apparatus.

Referring to FIG. 2, first, a service request is made from the user terminal 210 to the mass speech recognition apparatus 220 (step 201). The mass speech recognition device 220 then transmits the first step announcement to the user terminal (step 203). Then, the user terminal transmits the first-stage isolated language voice to the mass speech recognition apparatus 220 according to the announcement (step 205), and the mass-speech speech recognition apparatus 220 processes the transmitted first-stage isolated language speech to provide the first-stage service information. Acquire (step 207). Thereafter, the mass speech recognition apparatus 220 transmits the two-stage announcement to the user terminal (step 209), and the user terminal transmits the two-stage isolated speech to the mass speech recognition apparatus 220 according to the announcement (step) 211, the mass speech recognition apparatus 220 processes the transmitted two-stage isolated word speech to obtain two-stage service information (step 213).

In this way, if it is necessary to service the connection word consisting of the isolated words of n words, the large-capacity speech recognition apparatus 220 transmits the announcement of step n to the user terminal 210 (step 215), and the step n of the user terminal 210 is performed. Receives the isolated speech information (step 217), processes the isolated speech of step n (step 219), processes the entire voice information, and then requests the service device 230 to request the desired service from the service device 230 (step 221). The service device transmits the requested service to the user terminal (step 223).

Here, each step is a step of requesting voice transmission by separating each isolated word when the service requested by the user terminal is composed of a complex continuous word. For example, when the service requested by the user terminal is "SBS Drama Sky", the large-capacity voice recognition device 220 "tell the desired over-the-air broadcasting channel." Transmits the first step announcement, and receives the isolated information "SBS" voice information from the user terminal 210, processes the information, and then transmits the second step announcement, "speak the desired broadcast type." . Then, when the voice information "drama" is received, the three-step announcement is sent again, "tell the title of the desired drama" to receive the information of the final isolated word "heaven".

Through this step of speech recognition to obtain information about the "SBS drama sky poetry" to transmit the final speech recognition information to the service device.

This method of processing large-capacity speech recognition has the advantage of real-time response and good recognition performance due to the limited number of words for recognition of each speech information. There is a disadvantage that it may be inconvenient from the user's point of view.

3 is a flowchart illustrating another method for providing a voice recognition service using an existing large capacity speech recognition apparatus.

Referring to FIG. 3, first, a service request is made from the user terminal 310 to the mass speech recognition apparatus 320 (step 301). Thereafter, the mass voice recognition apparatus 320 transmits the announcement to the user terminal (step 303). Then, the user terminal transmits the connection word voice including a plurality of isolated words to the mass speech recognition apparatus 320 according to the announcement (step 305), and the mass speech recognition apparatus 320 processes the transmitted connection speech voice to provide service information. Acquire (step 307). Thereafter, after processing the connection word voice information, the user device 310 requests the desired service from the service device 330 (step 309), and the service device transmits the requested service to the user terminal 310 (step 311). .

This method of processing large-capacity speech recognition has the advantage that the service can be used quickly and conveniently because the user can use the service with only one step from the user's point of view. As the amount increases, real-time response is difficult and recognition performance is poor.

4 is a conceptual diagram of a system according to an embodiment to which the present invention is applied.

Referring to FIG. 4, the voice recognition service system according to the present invention includes a user terminal 401 coupled with a network 407, a mass voice recognition device 403, and a service device 405. This division is for understanding the present invention, and each component may be a functional entity rather than a physical entity. Therefore, as can be seen in this figure, the user may directly connect to the large-capacity voice recognition device 403 without going through the user terminal 401 in some cases.

The user terminal 401 may connect to the network 407 and may be any type of device capable of transmitting voice. In general, the voice recognition service provides a voice-based service, so that the voice recognition service will be a telephone or mobile communication terminal capable of transmitting voice. However, it also includes general personal computers, handheld PCs, notebook PCs, PDAs, home appliances including computing functions, etc., which can be connected to the high-capacity voice recognition device 403 via the network 407 to set up services according to the present invention. A computing device including a computing system. In addition, the terminal 401 may mean only some of the functions of the above-described embodiments as a functional entity, not as a physical entity.

Network 407 includes any type of communication network over which voice services can be delivered. Such a communication network may typically include a public switched telephone network (PSTN), a mobile communication network, and the Internet. In addition, such a network is only for the understanding of the present invention, and the network 407 is a concept including not only the above-described examples but also a micro network, for example, a bus interface inside a computer.

The mass speech recognition apparatus 403 is responsible for analyzing a voice signal received from the user terminal 401 and requesting a desired service from the user terminal 401 to the service device 405. When the large-capacity speech recognition apparatus 403 analyzes the speech signal received from the user terminal 401, the speech recognition may be performed quickly and accurately using the stored user profile information.

Such profile information is received and stored in advance from the user terminal 401, and provides a faster analysis of the voice signal when generating a finite state network (FSN) and a dictionary using the stored user profile information to provide a real speech recognition service. do. Therefore, a profile generally includes words frequently used by a user, frequently used service categories, and user authentication information.

The large-capacity speech recognition apparatus 403 is also a functional entity, not a physical entity, but a device having a physical entity generally referred to by a person in general as well as a functional entity having the functions described above may be represented as a device.

The service device 405 is responsible for transmitting the requested service to the user terminal 401 according to the service request signal requested by the mass voice recognition device 405. In this case, if the user does not need to provide the service from the user terminal 401, the user does not need to provide the user terminal 401 if the user can receive the appropriate service.

The service device 405 may store various service data according to a service. For example, in the case of a video on demand (VOD) service, the service device 405 may store broadcast data that can be transmitted to a user terminal, and in the case of another service, other data corresponding to the service may be stored. will be.

The service device 405 may also be represented as the service device 405 as well as the service device 405 having the functions described above as well as the service device 405 having a physical entity generally as a functional entity as a functional entity.

However, the description of such a system is only an embodiment of the present invention.

In the present invention, the user terminal 401, the large-capacity voice recognition device 403, and the service device 405 may be physical entities exemplified above, but have a meaning as a functional division.

For example, in the case of the notebook PC represented as an embodiment of the terminal 401 as described above, the terminal 401 may recognize a voice and modulate the voice into a signal recognizable by a mass recognition device. In addition, as a large-capacity voice recognition device 403, the voice signal received from the terminal 401 may be analyzed to include a role of requesting a desired service from the user terminal 401 to the service device 405. As the service device 405, all of the roles of providing a service to a user according to the service request signal requested by the large-capacity voice recognition device 405 may be performed together.

In this case, the network 407 is a small computer system interface (SCSI), an integrated drive electronics (IDE), a peripheral component interconnect (PCI), and an industry standard architecture (ISA), which are data transmission / reception networks in a notebook PC, not the Internet or a mobile communication network. Can be.

For the convenience of description, the embodiments described below assume that all of the terminals, the large-capacity speech recognition apparatus, and the service apparatus are physically separated entities, but as described above, each component is not only a physical entity but also a functional entity. It can also be understood as.

5 is a flowchart of processing a service in a large-capacity speech recognition apparatus according to an exemplary embodiment of the present invention.

Referring to FIG. 5, first, a large-capacity speech recognition apparatus receives a speech signal for speech recognition (step 501). The mass speech recognition apparatus then analyzes the user of the speech signal (step 503). This analysis of the user is a necessary step to utilize the user profile information. The user's analysis can be performed by voice signal, which analyzes the user's voice information, analyzes the characteristics of each user, and checks the user. A method of receiving back voice information is also available.

After analyzing the user's information after these steps, it is determined whether the user is a service user (step 505). If the service user is not a service user, the service termination message is sent and the service is terminated (step 507). Step 509).

In this case, if the user profile does not exist, voice recognition may be performed using the existing voice recognition procedures described with reference to FIGS. 2 and 3 (step 511).

Speech may be recognized by a finite state network (FSN) generated by the user profile information and a dictionary (step 513). The speech recognition method based on the preset profile information has a higher accuracy and a faster speech recognition process than the conventional method, thereby enabling highly efficient speech recognition processing.

When speech recognition is completed in this manner (step 515), the recognized information is sent to the service device (step 517). In this manner, a faster and more efficient speech recognition service can be provided than the conventional speech recognition scheme.

6 is a flowchart for storing user profile information in a mass voice recognition apparatus according to an exemplary embodiment of the present invention.

Referring to FIG. 6, first, the user terminal 610 accesses the large-capacity speech recognition apparatus 620 (step 601). Then, the mass speech recognition apparatus 620 requests the user profile information from the user terminal 610 (step 603), and the user terminal 610 received the request of the user profile information is requested by the mass speech recognition apparatus 620. Profile information of the form is created (step 605).

Thereafter, the created user profile information is transmitted to the mass speech recognition apparatus 620 (step 607). The mass speech recognition apparatus 620 then stores the user profile information transmitted from the user terminal 610 (step 609) and analyzes the profile information (step 611). The user profile information includes user authentication information and a dictionary. And keywords for narrowing the scope of the finite state network (FSN). Analysis of these keywords allows the creation of dictionaries and finite state networks (FSNs) that narrow the search appropriately.

 Thereafter, a finite state network (FSN) and a dictionary are generated and stored using the words included in the user profile information (step 613).

The words included in the profile information serve to define a database of words to be searched by the large-capacity speech recognition device. That is, if the existing speech recognition method is a method of recognizing speech by searching all the speech information contained in the database regardless of the user, the present invention applies different criteria to each user using profile information to apply finite state network (FSN). And dictionaries can be created, each having a different grammar structure and dictionary according to the user. This approach eliminates the need to search through large amounts of voice information for fast and accurate voice recognition.

The stored finite state network (FSN) and the dictionary will be applied when providing a voice recognition service to the user terminal 610.

7 is a diagram illustrating an embodiment of user profile information according to an exemplary embodiment of the present invention.

Referring to FIG. 7, such user profile information is prepared in advance before receiving a voice recognition service from a user terminal. Therefore, the large-capacity speech recognition apparatus storing the user profile information in advance may generate a finite state network (FSN) and a dictionary based on words included in the user profile information, and provide a fast and accurate speech recognition service based on the word.

Since user profile information is often different personal information for each user, different information may need to be stored for each user. Therefore, the first thing to be recognized in the user profile information will be a user ID and a user password (701). Of course, the user may be determined by the characteristics of the voice information, not the ID or password, and in this case, what is included in the user profile information may be the voice fingerprint of the user, not the user ID and password.

Since the user profile information of FIG. 7 represents an embodiment of a voice recognition TV guide, the user profile shown in the figure is represented by a list of service types, service channels, and favorite words. Such a user profile may be changed according to the type of voice recognition service.

The information of the service type 703 and the service channel 705 serves to reduce the range of services to be searched for when the large-capacity speech recognition device recognizes the voice.

Favorite words 707 are words most frequently used when a user uses a voice recognition service. This user profile information enables the mass-capacity speech recognition device to create a finite state network (FSN) and dictionary within a database defined by the user profile without having to search all the speech databases included in the device.

The generated finite state network (FSN) and the dictionary are stored so that when a user terminal requests a voice recognition service, real time response is possible and accurate voice recognition is possible.

8 is a flowchart illustrating a voice recognition service using a user profile according to an embodiment of the present invention.

Referring to FIG. 8, first, a user terminal 810 transmits a service access signal to a mass voice recognition apparatus 820 (step 801). Then, the mass speech recognition apparatus 820 requests user profile information to the user terminal 810 (step 803), and the user terminal 810 receiving the request of the user profile information is requested by the mass speech recognition apparatus 820. Profile information of the form is created (step 805).

Thereafter, the created user profile information is transmitted to the mass speech recognition apparatus 820 (step 807). Then, the mass speech recognition apparatus 820 stores the user profile information transmitted from the user terminal 810 (step 809), analyzes the profile information, and uses the words included in the user profile information to find a finite state network (FSN). And generate and store the dictionary (step 811).

Thereafter, in order to use the voice recognition service, the user terminal 810 transmits a service request signal to the large-capacity voice recognition device 820 (step 813). The mass voice recognition device 820 then sends a service announcement to the user terminal 810 (step 815). The user terminal 810 receiving the service announcement sends an appropriate speech signal to the mass speech recognition apparatus 820 according to the announcement (step 817), and the mass speech recognition apparatus 820 receiving the speech signal identifies the user. The information is analyzed (step 819). The user identification information may be based on the voice itself, or may receive the encryption key according to the content of the voice and analyze the identification information. The analysis of the user identification information is to determine whether the user who wants to use speech recognition is an authenticated user.

Thereafter, the mass speech recognition apparatus 820 checks whether an authenticated user profile exists (step 821), and if the user profile exists, processes the voice signal using the user profile generated in step 811 (step 825). (Step 823).

Thereafter, the information recognized by the speech processing is transmitted to the service device 830 (step 827). In response to the service request signal, the service device 830 transmits a service corresponding to the request signal to the user terminal 810 (step 829).

The present invention is not limited to the above embodiments, and many variations are possible by those skilled in the art within the spirit of the present invention.

According to the present invention, it is possible to provide a voice recognition service method and apparatus using a user profile.

In addition, according to the present invention, it is possible to increase the speed and accuracy of the voice recognition service by using the profile information prepared in advance.

Claims (8)

Receiving a service access signal from a user terminal; Sending a user profile information request signal to the user terminal in response to the service access signal; Receiving user profile information from the user terminal; Storing the received user profile information; Generating a finite state network (FSN) and a dictionary according to the stored user profile information; Receiving a voice signal from a user terminal; Analyzing user identification information corresponding to the voice signal; And Recognizing the voice signal using the generated finite state network (FSN) and dictionary corresponding to the analyzed user identification information Speech recognition method of a large-capacity speech recognition device comprising a. The method of claim 1, Sending a service request signal corresponding to the recognized voice signal to a service device; Speech recognition method of a large-capacity speech recognition device further comprising. delete The method of claim 1, The user profile information includes at least one finite state network (FSN) and at least one word that can be generated in a dictionary. Speech recognition method of a large-capacity speech recognition device characterized in that. Means for receiving a service access signal from the user terminal; Means for sending a user profile information request signal to the user terminal in response to the service access signal;  Means for receiving user profile information from the user terminal; Means for storing the received user profile information; Means for generating a finite state network (FSN) and a dictionary according to the stored user profile information; Means for receiving a voice signal from a user terminal; Means for analyzing user identification information corresponding to the voice signal; And Means for recognizing the voice signal using the generated finite state network (FSN) and a dictionary corresponding to the analyzed user identification information Large-capacity speech recognition device comprising a. The method of claim 5, Means for sending a service request signal corresponding to the recognized voice signal to a service device Large-capacity speech recognition device further comprising. delete The method of claim 5, The user profile information includes at least one finite state network (FSN) and at least one word that can be generated in a dictionary. A large capacity speech recognition device characterized in that.

Images