JP2009104047A - Information processing method and information processing apparatus - Google Patents

Abstract

In speech registration type speech recognition, the accuracy of speech recognition at the time of execution is brought close to the accuracy at the time of utterance twice or more at the time of registration without incurring the burden of uttering speech twice or more at the time of speech registration.
SOLUTION: A registration target voice uttered by a user is acquired, voice information corresponding to the acquired registration target voice is registered in a memory, and a recognition target voice uttered by a user is separately registered in the memory. Speech recognition is performed using speech information, and one or a plurality of recognition results are output. A recognition result intended by the user is specified from the output one or more recognition results, and the specified recognition result As the registered voice information corresponding to, voice information corresponding to the voice uttered as a recognition target is registered.
[Selection] Figure 2

Description

  The present invention relates to an information processing method for performing speech registration type speech recognition.

  In speech recognition, in addition to a method of recognizing based on a grammar describing a recognizable vocabulary, one or a plurality of previously spoken voices are registered, and at the time of execution, a registered voice most similar to the input voice is obtained. There is a method. The latter is called voice registration type voice recognition. As an example, consider a telephone that can make a call by speaking the name of the person to call instead of dialing a telephone number. In the destination voice registration mode, “Ichiro” and the voice uttered are registered in association with the telephone number “03-1XXX-XXXX” of “Ichiro Suzuki”. After registration, the user can call “03-1XXX-XXXX” by saying “Ichiro” in the transmission mode. Similarly, if you register the voice by saying “Hi Sha san” to the telephone number “045-2XX-XXXX” of “○ × Dental Clinic”, then you can simply say “Hi Sha San” and “ You can call the clinic. You can use any language you like to register.

A general method for realizing such voice registration type voice recognition is as follows. A speech recognition grammar 1 capable of accepting an arbitrary syllable string is prepared, the utterance at the time of registration is recognized using the speech recognition grammar 1, and a syllable string is output as a recognition result. This syllable string is registered in the speech recognition grammar 2 for execution in association with the target telephone number. In the transmission mode, the input speech is recognized using the speech recognition grammar 2.
JP 2000-276187 A

  In the above example, the case where the voice is uttered only once at the time of voice registration has been described. However, there is a problem in that the accuracy at the time of execution is not sufficient, so that the voice is uttered twice or more and registered. Often designed. Although this design can improve accuracy, there is a problem that the burden on the user increases.

  Patent Document 1 discloses a method of registering, as a registered voice, a second input voice that the user has continuously rephrased when the first input voice at the time of execution is unrecognizable or misrecognized. . That is, when the first input voice and the second input voice are similar, it is determined that this is a rephrase, and when there is a manual operation by the user, the command instructed by the manual operation is The first voice is additionally registered as the registered voice. In this way, by additionally registering the input voice at the time of execution as a registered voice, it is possible to improve the subsequent recognition accuracy, and instead, it is possible to complete with one utterance at the time of registration. However, this method is based on the premise that the user rephrases the same word for a recognition error and further that manual operation is performed. This situation is not certain and there is a problem that it is uncertain whether the effect of additional registration will appear.

  In order to solve the above problems, an information processing apparatus according to the present invention includes an acquisition unit that acquires a voice uttered by a user, and a memory that stores voice information corresponding to the registration target voice acquired by the acquisition unit as registered voice information. A speech recognition means for recognizing the speech to be recognized acquired by the acquisition means using the speech information registered in the memory and outputting one or more recognition results; A specifying unit for specifying a recognition result intended by the user from one or a plurality of recognition results output by the voice recognition unit, and the registration unit corresponds to the recognition result specified by the specifying unit. Voice information corresponding to the voice to be recognized is registered as the registered voice information.

  According to the present invention, at the time of voice registration, the accuracy of voice recognition at the time of execution can be made close to the accuracy when uttered twice or more at the time of registration without burdening the voice to be uttered twice or more. .

  Embodiments according to the present invention will be described below in detail with reference to the accompanying drawings. However, the constituent elements described in this embodiment are merely examples, and are not intended to limit the scope of the present invention only to them.

  Hereinafter, Embodiment 1 of the present invention will be described in detail with reference to the drawings. In this embodiment, as an example of the information processing apparatus according to the present invention, a case of operating a mobile phone like 101 in FIG. 1 will be described. The mobile phone 101 has a liquid crystal screen 102 and can be operated by operating a key 103 and inputting voice from a microphone 104.

  FIG. 2 is a block diagram illustrating the configuration of the mobile phone 101 according to the present embodiment. In the figure, reference numeral 201 denotes a voice input unit. Reference numeral 202 denotes a voice recognition unit. A voice recognition grammar holding unit 203 is used by the voice recognition unit 202. A determination unit 204 determines whether or not to present a confirmation dialog to the user. Reference numeral 205 denotes a confirmation unit that presents a confirmation dialog to the user via a display screen (not shown) or a voice output unit (not shown) when the determination unit 204 determines to present a confirmation dialog, and receives input from the user. . 206 is a syllable recognition unit that recognizes a voice acquired via the voice input unit 201 when a user's instruction input in the confirmation unit 205 is finished, and outputs a syllable sequence as voice information corresponding to the voice. . A syllable recognition grammar holding unit 207 is used by the syllable recognition unit 206. A registration unit 208 registers the syllable sequence output by the syllable recognition unit 206 in the speech recognition grammar holding unit 203 as registered speech information.

  FIG. 3 is a configuration diagram illustrating a hardware configuration of the mobile phone 101 according to the present embodiment. In the figure, reference numeral 301 denotes a CPU which operates according to a program for realizing an operation procedure of the mobile phone 101 described later. A RAM 302 provides a storage area necessary for the operation of the program. Reference numeral 303 denotes a ROM which holds a program for realizing the operation procedure of the program. Reference numeral 304 denotes various buttons as indicated by 103 in FIG. Reference numeral 305 denotes a microphone as shown by 104 in FIG. Reference numeral 306 denotes an LCD as indicated by 102 in FIG. Reference numeral 307 denotes a bus.

  In this embodiment, a case will be described in which the voice corresponding to the registered telephone number is registered and the registered telephone number is called with the registered voice in the memory registration function of the telephone number.

  First, a process for registering a telephone number in memory will be described with reference to the flowchart of FIG. Reference numeral 601 in FIG. 6 represents a scene in which the memory registration screen is called by a predetermined operation of the mobile phone. First, as shown at 601 in FIG. 6, the user is prompted to input a telephone number to be registered, and the telephone number is input using the key 103 (S 401). Subsequently, on a screen such as 602, the user is prompted to input a registered voice for performing an operation of calling the telephone number. The user utters a registration voice to be registered toward the microphone 104 (S402). The registered voice may be any utterance. Here, it is assumed that the registered telephone number is that of a friend, Mr. Taro Yamada, and that "Tarokun" is spoken. This input speech is recognized by the syllable recognition unit 206 (S403). The syllable recognition unit 206 performs recognition processing based on the grammar of the syllable recognition grammar holding unit 207. This grammar is written to accept any syllable string pattern. The speech recognition result is output in the form of a syllable string. The syllable string of the recognition result when the input speech “Tarokun” can be accurately recognized is “ta-roo-ku-n”, but the recognition of the syllable string is not always 100% accurate, and “toa A syllable string different from an accurate syllable string, such as -reo-ku-n ", may be output. This output syllable string is registered in the speech recognition grammar of the speech recognition grammar holding unit 203 in association with the registered telephone number 0901111XXX (S404), and a message such as 603 in FIG. 6 is displayed. Similarly, FIG. 9 shows an example of the speech recognition grammar 203 as a result of registering speech corresponding to a plurality of telephone numbers. In this way, the phone number corresponding to the syllable string of the syllable recognition result of the registered voice is registered as a pair. On the right side, I showed what I made as a reference.

  Next, the operation for making a call by calling a telephone number registered in the memory by voice input will be described with reference to the flowchart of FIG. Reference numeral 701 in FIG. 7 represents a scene in which a screen for calling and calling a telephone number registered in a memory by voice input is called by a predetermined operation of the mobile phone. As indicated by reference numeral 701 in FIG. 7, voice input is prompted and voice input is performed (S501). The speech recognition unit 202 recognizes the input speech that is the recognition target (S502). This speech recognition is performed using the grammar of the speech recognition grammar holding unit 203. In this example, FIG. 9 is a speech recognition grammar. The speech recognition result is output to the display screen as N-best with certainty by the display control means. For example, when a voice is input as “Tarokun” in order to call a friend, Mr. Taro Yamada, the recognition result is output as shown in FIG. It is determined whether or not the first candidate of the recognition result has a significant difference with respect to the second or lower candidate (S503). Here, if the certainty score value is 0.7 or more, it is determined that the difference is significant. In FIG. 10, since the first certainty score of the recognition result is 0.75, it is determined that there is a significant difference. In this case, it is determined that the first-ranked telephone number matches the one intended by the user, and a call operation to the telephone number is performed (S507). Reference numeral 702 in FIG. 7 denotes a screen display during transmission. On the other hand, when the recognition result is as shown in FIG. 11, since the first certainty score is not 0.75 or more, a confirmation dialog with an N-best list such as 802 in FIG. 8 is displayed (S504). The user selects a desired telephone number from these by a predetermined key operation (S505). Here, the second telephone number is selected. Then, a call is made to the selected telephone number (S507), but before that, the process of S506 is performed. When the desired telephone number is selected by the user, it can be seen that the voice uttered in S501 corresponds to the telephone number 0901111XXXX. Therefore, the voice information corresponding to this voice is additionally registered as registered voice information corresponding to the operation information for transmitting to the telephone number 0901111XXXX. The registration method is the same as the procedure of S403 and S404 in FIG. As a result, the speech recognition grammar stored in the speech recognition grammar holding unit 203 is as shown in FIG. 12, and two types of syllable strings are registered for the phone number 0901111XXX (Taro Yamada's phone number).

  As described above, the registration operation of the registered voice can be performed only once. However, it is possible to create a state in which two or more registered voices are registered by using the input voice at the time of calling the telephone number. The rate can be improved.

  In Example 1, in S503 of FIG. 5, it is determined whether or not there is a significant difference depending on whether or not the certainty score of the first candidate is greater than a predetermined value. On the other hand, when the certainty score of the first candidate and the certainty score of the second candidate are equal to or greater than a predetermined difference, it may be determined that the difference is significant.

  In the above embodiment, whether or not to display the dialog 802 in FIG. 8 is determined based on the certainty score of the recognition result. On the other hand, when the operation for the first result is executed before the dialog is displayed and the operation is canceled by the user, the N-best dialog may be displayed. For example, as shown in FIG. 13, when canceling is performed by a predetermined operation as in 1303 immediately after calling the first phone number of the recognition result for the input voice in 1301, a dialog is displayed as in 1304. You may make it do.

In the above-described embodiment, an example has been shown in which a telephone number registered in a memory in a mobile phone is called by voice. However, it goes without saying that the present invention can be applied to voice operations of various devices and various systems. A user interface using voice registration type voice recognition can be applied to voice operation of a copying machine, voice operation of a digital camera, voice operation of a digital television, and the like. In addition, an audio-only dialogue without display may be used. For example, in a call center voice conversation,
Sys: Press “1” for product inquiries and “2” for fault repair inquiries.
User: [Press “2”]
Sys: “1” for a compact camera.
User: [Press “1”]
Sys: I will connect to the person in charge. When registering the operations up to this point, say the sound you want to register. If you do not register, please wait.
User: Camera show Sys: Voice has been registered. I will connect you to the person in charge.
As described above, it is assumed that the registered voice corresponding to the operation information for inquiring about the repair of the compact camera is registered. Thereafter, when the user calls the call center again, the following dialogue is performed.
(1) 'Sys: Press "1" for product inquiries and "2" for fault repair inquiries.
(2) 'User: Camera show (3)' Sys: Are you sure you want to have a camera show?
(4) 'User: Yes (5)' Sys: We will connect to the person in charge.
(3) Based on the confirmation such as' and the response to (4) ', it is determined that the input voice of (2)' is intended to be connected to the person in charge of inquiring the repair of the compact camera. Therefore, the process of connecting to the person in charge at (5) ′ is executed, and the input voice of (2) ′ is registered as a registered voice. The system voice of (3) ′ is generated by combining the voice data “Are you sure?” With the registered voice of (6) used for the “camera show” part.

  In the above embodiment, the syllable string obtained as a result of recognition of the input speech at the time of registration by the syllable recognition unit 206 is registered in the speech recognition grammar holding unit 203 by the registration unit 208, and the input speech at the time of execution is registered as the speech recognition grammar holding unit 203. The voice recognition unit 202 recognizes the voice recognition grammar. On the other hand, the configuration shown in FIG. 14 can also be implemented. In the figure, reference numeral 1401 denotes a voice input unit. Reference numeral 1402 denotes a syllable recognition unit. Reference numeral 1403 denotes a syllable recognition grammar holding unit used by the syllable recognition unit 1402. Reference numeral 1404 denotes an evaluation unit. A determination unit 1405 determines whether to present a confirmation dialog to the user. Reference numeral 1406 denotes a confirmation unit that presents a confirmation dialog to the user and receives an input from the user when the determination unit 1405 determines to present the confirmation dialog. Reference numeral 1407 denotes a syllable string holding unit. A registration unit 1408 registers the syllable sequence output from the syllable recognition unit 1402 in the syllable string holding unit 1407 when the user's instruction input in the confirmation unit 1406 is completed.

  The input speech is recognized by the syllable recognition unit 1402 at the time of registration and execution. The syllable recognition unit 1402 outputs a syllable string as a recognition result. The registered speech is registered in the syllable string holding unit 1407 in the form of a syllable string. The registered contents are the same as in FIG. The evaluation unit 1404 calculates the degree of coincidence between the syllable string output by the syllable recognition unit 1402 and each syllable string registered in the syllable string holding unit 1407, and outputs N-best together with the degree of coincidence score in descending order of degree of coincidence. The determination unit 1405 operates based on the content of the N-best output as in the first embodiment.

  In the above-described embodiment, the user's intention is specified by the confirmation dialog or the confirmation message for the voice recognition result, but the present invention is not limited to such a form. When the first candidate of the speech recognition result is executed and the user does not cancel the execution result, it is determined that the first candidate of the recognition result matches the user's intention, and the first candidate is registered. Information of input voice may be registered as voice.

  In the above-described embodiment, an example in which a syllable string is registered as voice information has been described. However, the present invention is not limited to this, and voice data may be registered. In that case, first, the voice data corresponding to the registration target voice uttered by the user is stored in the memory, and the recognition target voice uttered separately by the user is compared with the voice data stored in the memory. One or more recognition results are output. Then, the recognition result intended by the user is identified from the output one or more recognition results, and the speech data corresponding to the speech to be recognized is stored in the memory as registered speech information corresponding to the identified recognition result. It is configured to register.

  The object of the present invention can also be achieved as follows. That is, a storage medium in which a program code of software that realizes the functions of the above-described embodiments is recorded is supplied to the system or apparatus. Then, the computer (or CPU or MPU) of the system or apparatus reads and executes the program code stored in the storage medium. It goes without saying that the purpose is achieved even in this way.

  In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention.

  As a storage medium for supplying the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

  Further, the embodiments according to the present invention are not limited to the case where the functions of the above-described embodiments are realized by executing the program code read by the computer. For example, an OS (operating system) running on a computer performs part or all of actual processing based on an instruction of the program code, and the functions of the above-described embodiments may be realized by the processing. Needless to say, it is included.

  Furthermore, the functions of the embodiment according to the present invention are also realized as follows. That is, the program code read from the storage medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer. Then, based on the instruction of the program code, the CPU provided in the function expansion board or function expansion unit performs part or all of the actual processing. It goes without saying that the functions of the above-described embodiments are realized by this processing.

1 is a mobile phone according to a first embodiment of the present invention. It is a block diagram of the mobile telephone 101 which concerns on Example 1 of this invention. It is a hardware block diagram of the mobile telephone 101 which concerns on Example 1 of this invention. It is a flowchart explaining the process which registers the audio | voice with respect to a registration telephone number on the mobile telephone 101 which concerns on Example 1 of this invention. It is a flowchart explaining the operation | movement in the case of making a telephone call by calling the telephone number registered in the memory by voice input on the mobile phone 101 according to the first embodiment of the present invention. It is a figure which shows the example of a transition of the screen display at the time of registering the audio | voice with respect to a registration telephone number on the mobile telephone 101 which concerns on Example 1 of this invention. It is a figure which shows the example of a transition of a screen display in the case of making a telephone call by calling the telephone number registered in memory by voice input on the mobile phone 101 according to the first embodiment of the present invention. It is a transition example of a screen display including a display of a confirmation dialog when calling a telephone by calling a telephone number registered in a memory by voice input on the mobile phone 101 according to the first embodiment of the present invention. It is an example of the speech recognition grammar based on Example 1 of this invention. It is an example of N-best with certainty which the voice recognition part concerning Example 1 of the present invention outputs. It is another example of N-best with reliability which the speech recognition part which concerns on Example 1 of this invention outputs. It is an example figure at the time of the speech recognition grammar which concerns on Example 1 of this invention when the syllable string of a registration speech is newly added. An example of screen display transition when a confirmation dialog is displayed when a cancel process is performed immediately after calling a phone number registered in a memory by voice input on the mobile phone 101 according to the third embodiment of the present invention is shown. FIG. It is a block diagram of the mobile telephone 101 which concerns on Example 5 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 201 Speech input part 202 Speech recognition part 203 Speech recognition grammar holding part 204 Judgment part 205 Confirmation part 206 Syllable recognition part 207 Syllable recognition grammar holding part 208 Registration part

Claims (12)

Obtaining means for obtaining the voice uttered by the user;
Registration means for registering voice information corresponding to the registration target voice acquired by the acquisition means in a memory as registered voice information;
Speech recognition means for recognizing speech to be recognized acquired by the acquisition means using registered voice information registered in the memory, and outputting one or a plurality of recognition results;
A specifying unit for specifying a recognition result intended by the user from one or a plurality of recognition results output by the voice recognition unit;
The information processing apparatus according to claim 1, wherein the registration unit registers voice information corresponding to the recognition target voice as registered voice information corresponding to the recognition result specified by the specifying unit. The registration unit registers voice information corresponding to the registration target voice in association with predetermined operation information as registered voice information,
The voice recognition means outputs operation information corresponding to the voice information as the recognition result,
The specifying unit specifies operation information intended by the user from one or more pieces of operation information output as a recognition result,
The information processing apparatus according to claim 1, wherein the registration unit registers voice information corresponding to the recognition target voice as registered voice information corresponding to the operation information specified by the specifying unit. Further comprising display control means for displaying one or more recognition results output by the voice recognition means on a display screen;
The identification unit identifies a recognition result selected by the user among one or more recognition results displayed by the display control unit as a recognition result intended by the user. The information processing apparatus described. A determination means for determining whether or not to output a confirmation dialog for presenting the user with one or more recognition results output by the voice recognition means and selecting a recognition result intended by the user;
2. The identification unit according to claim 1, wherein when the determination unit determines to output a confirmation dialog, the identification unit identifies a recognition result selected by the user according to the confirmation dialog as a recognition result intended by the user. The information processing apparatus described. The voice recognition means outputs a recognition result with certainty,
5. The determination unit according to claim 4, wherein the determination unit determines to output the confirmation dialog when the certainty factor of the first candidate of the recognition result is not significantly different from other candidates. Information processing device.   The significant difference is that the value of the certainty factor of the first candidate is greater than a predetermined value, or the difference between the certainty factor of the first candidate and the certainty factor of the second candidate is a predetermined value. The information processing apparatus according to claim 5, wherein the information processing apparatus is larger. Execution means for executing processing corresponding to one of the recognition results output by the voice recognition means;
The information processing apparatus according to claim 4, wherein the determination unit determines to output the confirmation dialog when a user performs a cancel operation on the process executed by the execution unit. Execution means for executing processing corresponding to one of the recognition results output by the voice recognition means;
When the user does not cancel the process executed by the execution unit, the specifying unit sets a recognition result corresponding to the process executed by the execution unit as a recognition result intended by the user. The information processing apparatus according to claim 1, wherein the information processing apparatus is specified.   The information processing apparatus according to claim 1, wherein the voice information corresponding to the voice is voice data corresponding to the voice or a syllable string obtained by voice recognition of the voice. An acquisition step of acquiring voice uttered by the user;
A first registration step of registering voice information corresponding to the voice to be registered acquired in the acquisition step in a memory as registered voice information;
A speech recognition step of recognizing the recognition target speech acquired in the acquisition step using registered speech information registered in the memory and outputting one or more recognition results;
A specifying step of specifying a recognition result intended by the user from one or a plurality of recognition results output in the voice recognition step;
An information processing method comprising: a second registration step of registering voice information corresponding to the recognition target voice in a memory as registered voice information corresponding to the recognition result specified by the specifying step.   A program for causing a computer to execute the information processing method according to claim 10.   The computer-readable storage medium which recorded the program of Claim 11.

Images