JP2015206998A - Portable terminal, editing guiding program, and editing guiding method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a portable terminal where candidates of false recognition strings are displayed so that the candidate can be determined at a glance, and sentences can be efficiently created by using voice recognition, and to provide an editing guiding program and an editing device.SOLUTION: A portable terminal 10 comprises a second microphone 16b for taking in voice of a user, and applies voice recognition to the voice input to the second microphone 16b to generate a character string. In addition, when creating the character string by the voice recognition, calculated likelihood is used as reliability of the voice recognition, and the generated character string and the reliability corresponding to the character string are recorded in a reliability table. The character string whose reliability of a threshold or less is specified on the basis of the reliability table, and background color of the specified character string of low reliability is colored in blue color to be displayed on an LCD monitor 26.

Description

  The present invention relates to a mobile terminal, and more particularly to a mobile terminal that inputs a character string by voice recognition, for example.

  2. Description of the Related Art Conventionally, a portable terminal that inputs a character string by voice recognition, for example, is known, and an example of this type of device is disclosed in Patent Document 1. This background art utterance recognition device performs speech recognition for each single syllable, displays the character image of a single syllable as it is when the recognition result is highly reliable, and displays the vowel when the recognition result is low in reliability. Next to the character image and an image of “?” Indicating that the consonant cannot be recognized. When the recognition result is further low, “*” indicating that the next syllable input is prompted is displayed.

The pronunciation practice support system disclosed in Patent Document 2 can supply pronunciation practice content such as English conversation using a mobile phone. When the learner inputs an audio signal to the mobile phone, the audio signal is transmitted from the mobile phone to the line call control device, converted into an audio data signal by the line call control device, and then transmitted to the pronunciation rating server. The utterance rating server evaluates the degree of similarity of the basic frequency pattern of the utterance by matching the utterance word or sentence of the learner with the data pattern. When the rating result is transmitted to the database server, the content is edited according to the rating result and displayed on the screen of the mobile phone. Thereby, the learner is notified of the correctness of pronunciation in the input voice signal by inputting the voice signal to the mobile phone.
JP-A-2005-128130 [G10L 15/22, G10L 15/28] JP-A-2005-31207 [G09B 19/06, G09B 5/06, G09B 19/04, G10L 15/00]

  However, in the speech recognition apparatus in Patent Document 1, since the user speaks in units of single syllables, it is not suitable for inputting long sentences. Even if a long sentence is input, if the reliability of each single syllable is low, the character string is mixed with “*” and “?”, And it becomes difficult for the user to read the sentence as a sentence.

  Moreover, in the pronunciation practice support system in Patent Document 2, speech can be input in units of words, but a function for creating a sentence by speech recognition is not disclosed.

  Therefore, a main object of the present invention is to provide a novel portable terminal and an editing guide program applied to a processor of such a portable terminal.

  Another object of the present invention is to provide a portable terminal and an editing guide program applied to a processor of such a portable terminal capable of increasing the efficiency of sentence creation by voice recognition.

  The present invention employs the following configuration in order to solve the above problems. The reference numerals in parentheses, supplementary explanations, and the like indicate the corresponding relationship with the embodiments described in order to help understanding of the present invention, and do not limit the present invention.

  1st invention is a portable terminal which has a voice recognition means which produces | generates a character string from the audio | voice signal taken in by the taking-in means and the audio | voice signal taken in by the taking-in means, Comprising: A recording means for recording data indicating the reliability, a specifying means for specifying a character string having a reliability equal to or lower than a predetermined value with reference to the data, and a character string specified by the specifying means are referred to as other character strings. It is a portable terminal provided with the display means to display with a different form.

  In the first invention, the portable terminal (10) has capturing means (16b) such as a microphone for speech recognition, and the speech recognition means (20a, 20b, 30) is obtained from the speech signal captured by the capturing means. A character string is generated. This voice recognition means is composed of, for example, a CPU (20a), a DSP (20b) and a ROM (30) for storing voice dictionary data.

  The recording means (20a, S147) uses the likelihood calculated when the character string is generated by the speech recognition means as the recognition reliability, and the generated character string and the reliability corresponding to the character string. Are recorded as data (336). In the data, a character string having a reliability level equal to or lower than a predetermined value (threshold value) is specified by the specifying means (20a, S161) as a candidate of a misrecognized character string (misrecognized character string). The display means (20a, 26, S163, S165) guides editing by the user by displaying only the specified character string, for example, the background color is displayed in blue.

  According to the first aspect, since the candidate for the misrecognized character string is displayed so that it can be determined at a glance, the user can easily determine whether the candidate character string needs to be edited. Therefore, the user can efficiently create a sentence using voice recognition.

  A second invention is dependent on the first invention, and further includes cursor display means for displaying a cursor for selecting only the specified character string.

  In the second invention, the cursor (CUb) is displayed so as to select only the character string specified by the cursor display means (20a, S201, S289).

  According to the second invention, since the cursor can select only the low reliability character string, the operability of the editing operation performed by the user can be improved.

  A third invention is according to the second invention and further comprises operation means for receiving an operation for moving the cursor, and the cursor selects a character string according to the operation result by the operation means.

  In the third invention, the cursor moves according to the operation result of the operation means (22d) such as a direction key capable of inputting, for example, the vertical direction and the horizontal direction.

  According to the third aspect, since the cursor can be operated with the direction keys, the user can perform a highly reliable cursor operation.

  The fourth invention is dependent on the second invention and further comprises search means for searching for a character string that matches a new character string that is newly voice-recognized after the character string is generated, and the cursor is provided by the search means. Select the searched string.

  In the fourth invention, the search means (20a, S331) searches the character string newly identified by the re-input voice from the specified character string after the character string is generated. The cursor selects a character string based on the search result by the search means.

  According to the fourth invention, the user can operate the cursor only by uttering an incorrect character string. Therefore, the user is provided with a cursor operation method that is convenient for creating a document by voice recognition.

  A fifth invention is dependent on any one of the second to fourth inventions, and edits a character string selected by a cursor based on a character string newly generated by a voice recognition means. (20a, S149, S319).

  In the fifth invention, the voice editing means (20a, S149, S319) replaces, for example, a character string selected by the cursor with a character string newly recognized by the re-input voice.

  According to the fifth aspect, the user can easily edit the misrecognized character string using voice recognition. Therefore, the user is provided with an editing operation that is convenient for editing text by voice recognition.

  A sixth invention is according to the fifth invention, further comprising list display means for displaying a list of character string candidates generated by the voice recognition means, wherein the voice editing means is displayed by the candidate list display means. When the selected candidate is selected, the selected candidate is edited as a newly generated character string.

  In the sixth invention, the list display means (20a, S315) displays, for example, all character strings whose likelihood calculated by the speech recognition means is a predetermined value or more. Of the character strings displayed in a list, the character string selected by the user is replaced with the character string selected by the cursor.

  According to the sixth aspect, since the voice recognition candidates are displayed as a list, the user can edit even if the recognition accuracy of the re-input voice is not high.

  A seventh invention is according to any one of the second to fourth inventions, the character input means for inputting a character string, and the character string selected by the cursor, the character string input by the character input means. Is further provided with character editing means for editing based on the above.

  In the seventh invention, the character input means (22e) is, for example, a character input key for inputting hiragana, and the character editing means (20a, S211, S215) converts the input hiragana into kanji or katakana. Replace with the string selected by the cursor.

  According to the seventh invention, the user can edit the text using the character input key in an environment that is inappropriate for voice recognition, such as in a train or around a noisy place. The user can also perform highly reliable editing operations using the character input keys.

  An eighth invention is dependent on the first invention, and a similar search means for searching for a character string similar to a character string newly generated by the voice recognition means and a character string searched by the similar search means for voice recognition Substitution means for substituting a character string newly generated by the means is further provided.

  In the eighth invention, the similarity search means (20a, S283, S285) is newly generated using the voice when recognizing a newly generated character string from among the specified character strings. Search for a string that is similar to the string. Then, the replacement means (20a, S293) replaces the character string similar to the newly generated character string with the newly generated character string.

  According to the eighth aspect, the user can edit the character string only by speaking the character string for editing. Therefore, the user is provided with an editing operation that is convenient for editing text by voice recognition.

  A ninth invention is according to the second invention, further comprising a similarity search means for searching for a character string similar to the character string newly generated by the speech recognition means, wherein the cursor is searched by the similarity search means Select a string.

  In the ninth invention, the cursor selects a character string similar to the newly generated character string.

  According to the ninth aspect, the user can select an intended character string regardless of the result of speech recognition of the re-input speech. Therefore, similarly to the fourth invention, the user is provided with a cursor operation method that is convenient for creating a document by voice recognition.

  A tenth invention is according to the ninth invention, and further comprises voice editing means for editing the character string selected by the cursor based on the character string newly generated by the voice recognition means.

  According to the tenth invention, similar to the fifth invention, the user is provided with an editing operation that is convenient for editing text by voice recognition.

  An eleventh invention is according to the tenth invention, further comprising list display means for displaying a list of character string candidates generated by the speech recognition means, wherein the speech editing means is displayed by the candidate list display means. When the selected candidate is selected, the selected candidate is edited as a newly generated character string.

  According to the eleventh aspect, similar to the sixth aspect, since the voice recognition candidates are displayed in a list, the user can edit correctly even if the recognition accuracy of the re-input voice is not high. Can do.

  A twelfth invention is according to the ninth invention, and is a character input means for inputting a character string, and a character editing means for editing a character string selected by a cursor based on the character string input by the character input means. Is further provided.

  According to the twelfth invention, similar to the seventh invention, text editing is performed using the character input key in an environment that is inappropriate for voice recognition, such as in a train or around a noisy place. be able to.

  A thirteenth invention is dependent on any of the eighth to twelfth inventions, and a voice dictionary recording means for recording a voice captured by the fetching means and a character string generated from the voice as a voice dictionary. The similarity search unit searches for a similar character string by calculating a correlation value between each of the voices recorded by the voice dictionary recording unit and the newly input voice.

  In the thirteenth invention, the speech dictionary recorded by the speech dictionary recording means (20a, S251) includes the identified character string and the feature pattern of speech or speech data corresponding to the identified character string. It is. The similarity search means searches for a similar character string by calculating a correlation value between the feature pattern of the speech data constituting the speech dictionary and the feature pattern of the newly input speech.

  According to the thirteenth aspect, a correlation function can be used to search for a similar character string.

  A fourteenth invention is dependent on the first to thirteenth inventions, a display means for displaying at least a part of a plurality of character strings, and a scroll means for scrolling the display position of the character strings displayed by the display means. Is further provided.

  In the fourteenth invention, when a part of a plurality of character strings generated by the voice recognition means is displayed on the display means (26) such as an LCD monitor, the scroll means (20a, S201, S237, S261) The display position is scrolled to display a character string that is not displayed.

  According to the fourteenth aspect, the user can search for the specified character string by scrolling the contents of the displayed character string.

  A fifteenth aspect of the present invention is a portable terminal (10) having capturing means (16a, 16b) for capturing a speech signal and speech recognition means (20a, 20b, 30) for generating a character string from the speech signal captured by the capturing means. ) Processor (20a), recording means (S147) for recording the character strings generated by the speech recognition means and data indicating their reliability, and specifying the character string having a reliability below a predetermined value with reference to the data And an editing guidance program that causes the character string specified by the specifying means to function as display means (26, S163, S165) for displaying in a form different from other character strings.

  In the fifteenth invention, similarly to the first invention, the user can efficiently create a sentence using voice recognition.

  A sixteenth aspect of the invention is an editing apparatus having capturing means (16a, 16b) for capturing a speech signal and speech recognition means (20a, 20b, 30) for generating a character string from the speech signal captured by the capturing means. Recording means (20a, S147) for recording character strings generated by the speech recognition means and data indicating their reliability (reliability table), and referring to the data, a character string having a reliability equal to or lower than a predetermined value. An editing device comprising: specifying means (20a, S161) for specifying, and display means (20a, 26, S163, S165) for displaying the character string specified by the specifying means in a form different from other character strings. .

  In the sixteenth invention, as in the first invention, the user can efficiently create a sentence using voice recognition.

  According to this invention, since the candidate for the misrecognized character string is displayed so that it can be determined at a glance, the user can efficiently create a sentence using voice recognition.

  The above object, other objects, features, and advantages of the present invention will become more apparent from the following detailed description of embodiments with reference to the drawings.

FIG. 1 is a block diagram showing a portable terminal of the present invention. FIG. 2 is an illustrative view showing an appearance of the portable terminal shown in FIG. FIG. 3 is an illustrative view showing a display example of a GUI displayed on the LCD monitor shown in FIG. FIG. 4 is an illustrative view showing another display example of the GUI displayed on the LCD monitor shown in FIG. FIG. 5 is an illustrative view showing one example of a reliability table stored in the RAM shown in FIG. FIG. 6 is an illustrative view showing another display example of the GUI displayed on the LCD monitor shown in FIG. FIG. 7 is an illustrative view showing still another display example of the GUI displayed on the LCD monitor shown in FIG. FIG. 8 is an illustrative view showing one example of a memory map of the RAM shown in FIG. FIG. 9 is an illustrative view showing one example of a data storage area in the memory map shown in FIG. FIG. 10 is a flowchart showing the outgoing mail creation process of the CPU shown in FIG. FIG. 11 is a flowchart showing submenu processing of the CPU shown in FIG. FIG. 12 is a flowchart showing the reliability threshold value setting process of the CPU shown in FIG. FIG. 13 is a flowchart showing document editing processing of the CPU shown in FIG. FIG. 14 is a flowchart showing arbitrary cursor editing processing of the CPU shown in FIG. FIG. 15 is a flowchart showing voice recognition input processing of the CPU shown in FIG. FIG. 16 is a flowchart showing the low-reliability part editing process of the CPU shown in FIG. FIG. 17 is a flowchart showing the cursor specification processing of the CPU shown in FIG. FIG. 18 is a flowchart showing voice designation processing of the CPU shown in FIG. FIG. 19 is a flowchart showing the voice search processing of the CPU shown in FIG. FIG. 20 is a flowchart showing conversion site search processing of the CPU shown in FIG. FIG. 21 is a flowchart of another embodiment showing the voice recognition input process of the CPU shown in FIG. FIG. 22 is a flowchart of another embodiment showing the conversion site search process of the CPU shown in FIG.

  Referring to FIG. 1, mobile terminal 10 includes a control unit 20 and a key input device 22, and control unit 20 includes a CPU (also referred to as a processor or a computer) 20a and a DSP (Digital Signal Processor) 20b. Including. Here, when a call operation is performed by the key input device 22, the CPU 20a included in the control unit 20 controls the wireless communication circuit 14 corresponding to the CDMA system and outputs a call signal. The output call signal is transmitted from the antenna 12 and transmitted to the mobile communication network including the base station. When the other party performs a response operation, a call ready state is established.

  When a call end operation is performed by the key input device 22 after shifting to a call ready state, the CPU 20a controls the wireless communication circuit 14 to transmit a call end signal to the other party. Then, after transmitting the call end signal, the CPU 20a ends the call process. Also, when the call end signal is received from the call partner first, the CPU 20a ends the call process. Furthermore, the CPU 20a also ends the call process when a call end signal is received from the mobile communication network regardless of the call partner.

  When a call signal from a call partner is captured by the antenna 12 while the mobile terminal 10 is activated, the wireless communication circuit 14 notifies the CPU 20a of an incoming call. Further, the CPU 20a controls the LCD monitor 26 which is a display means by the LCD driver 24, and causes the LCD monitor 26 to display the caller information described in the incoming call notification. The CPU 20a then outputs a ring tone from an incoming call notification speaker (not shown).

  In the call ready state, the following processing is executed. The modulated audio signal (high frequency signal) sent from the other party is received by the antenna 12. The received modulated audio signal is subjected to demodulation processing and decoding processing by the wireless communication circuit 14. The obtained received voice signal is output from the speaker 18. On the other hand, the transmitted voice signal captured by the first microphone 16 a serving as a capturing unit is subjected to encoding processing and modulation processing by the wireless communication circuit 14. Then, the generated modulated audio signal is transmitted to the other party using the antenna 12 as described above.

  Further, the mobile terminal 10 has a document editing function for inputting and deleting a character string, and can perform character input by voice recognition. That is, when the user reads out the sentence and inputs the voice to the second microphone 16b, the sentence whose voice is recognized is displayed on the LCD monitor 26. Specifically, the audio signal captured by the second microphone 16b is converted into audio data by the DPS 20b, and the CPU 20a and the DSP 20b extract a feature pattern (feature amount) from the audio data. Further, the CPU 20a and the DSP 20b read the reference voice data constituting the voice dictionary for voice recognition from the ROM 32, and extract a feature pattern of the reference voice data (hereinafter referred to as a reference pattern) or directly read the reference pattern from the ROM 32. read out. Then, the CPU 20a and the DSP 20b collate the feature pattern with each reference pattern by a speech recognition method based on feature pattern matching or a speech recognition method based on statistical decision theory, thereby matching the reference speech data with the speech data. Is identified. Since this voice dictionary associates reference voice data with a character string represented by the reference voice data, the CPU 20a and the DSP 20b read the character string corresponding to the identified reference voice data, thereby converting the voice data into a character string. Convert to

  In the speech recognition method based on feature pattern matching, the CPU 20a and the DSP 20b calculate the likelihood between each reference pattern and the feature pattern by a method such as a multi-template method, an NN (Nearest Neighbor) identification method, or a k-NN identification method. Then, the reference pattern having the highest likelihood is specified.

  Further, in the speech recognition method based on the statistical decision theory, the likelihood between each reference speech data and the feature pattern using an HMM (Hidden Markov Model) widely used as a speech recognition method. And the CPU 20a and the DSP 20b specify the reference speech data with the highest likelihood in the same manner as the speech recognition method based on feature pattern matching.

  In the present embodiment, the likelihood described above is used as the reliability of recognition in speech recognition. Further, the portable terminal 10 causes the CPU 20a, DSP 20b, and ROM 32 to function as voice recognition means.

  The portable terminal 10 has a mail function, and can perform data communication with a mail server (not shown) to send and receive mail. During data communication, the antenna 12 and the wireless communication circuit 14 function as communication means, and a mail server or the like is connected to the network by wire or wirelessly.

  FIG. 2 is an illustrative view showing an appearance of the mobile terminal 10. With reference to FIG. 2, the portable terminal 10 has a case C formed in a plate shape. The antenna 12 is an extendable antenna configured to be extendable and provided to protrude from the upper side surface of the case C. Note that the antenna 12 may be a built-in antenna or may be built in the case C.

The first microphone 16a and the speaker 18 (not shown in FIG. 2) are built in the case C. The opening op1 leading to the built-in first microphone 16a is provided on one main surface in the length direction of the case C, and the opening op2 leading to the built-in speaker 18 is provided on the other main surface in the length direction of the case C. It is done. A second microphone 16b (not shown in FIG. 2) is also built in the case C. The opening op3 leading to the built-in second microphone 16b is provided on one main surface in the length direction of the case C so as to be aligned with the opening op1.

  That is, the caller inputs the transmission voice to the first microphone 16a through the opening op1 and listens to the reception voice from the speaker 18 through the opening op2. The user inputs voice for voice recognition to the second microphone 16b through the opening op3. The first microphone 16a can also be used for voice recognition. A technique for canceling ambient noise, which is a distant sound source, can be realized by the difference between the sound signals of the first microphone 16a and the second microphone 16b, which contributes to an improvement in the speech recognition rate.

  The key input device 22 includes a first menu key 22a, a second menu key 22b, an enter key 22c, a direction key 22d (also referred to as operation means), a plurality of character input keys 22e (also referred to as character input means), and the like. Is provided on the main surface. The LCD monitor 26 is attached such that the monitor screen is exposed on the main surface of the case C. The key input device 22 includes a call key and an end key.

  The first menu key 22a, the second menu key 22b, and the enter key 22c correspond to soft keys displayed on the LCD monitor 26, respectively, and are used when operating the soft keys displayed in the LCD monitor 26. . The direction key 22d is used to perform an input operation in the vertical direction or the horizontal direction on a GUI (Graphical User Interface) displayed on the LCD monitor 26. The determination key 22c is also used to determine the GUI operation result by the direction key 22d.

  The user can input a character string by using the plurality of character input keys 22e while the document editing function is being executed. For example, the plurality of character input keys 22e are composed of character keys from "A" line to "WA" line, and a character (Hiragana) to be input can be designated by the number of times each character key is operated. Specifically, the “A” character key is a key for inputting the characters “A, I, U, E, O”. If the user presses the “A” line character key, "Can be entered, and if pressed again, the character" I "can be entered. Further, the user can convert hiragana to kanji or katakana by using the direction key 22d and the enter key 22c.

  Note that the user can switch the characters assigned to the character input keys to alphanumeric characters or symbols by operating the GUI of the document editing function, and can input alphanumeric characters or symbols. The plurality of character input keys 22e can also be used as numeric keys from 0 to 9. For example, the “a” line character key is the number “1”, and the “ka” line character key is the number “ 2 ”respectively. That is, when a user inputs a telephone number to make a call, the user can input the telephone number using a plurality of character input keys 22e and perform a call operation using the call key.

  Here, a document editing function capable of inputting a character string by voice recognition will be described in detail using each image displayed on the LCD monitor 26, a GUI, and the like.

With reference to FIG. 3A, a status display area 50, a function display area 52, and a key display area 54 are set in the LCD monitor 26. The status display area 50 is set on the upper side of the LCD monitor 26, and displays the radio wave reception status by the antenna 12, the remaining battery capacity of the rechargeable battery (battery), the current date and time, and the like. In addition, the display content of the function display area 52 changes according to the function to be executed. Here, a text editing screen of the outgoing mail is displayed, and a cursor CUa indicating the current character input position is displayed.

  A plurality of soft keys are displayed in the key display area 54. The display state changes according to the function to be executed, and the key display area 54 is not displayed for functions that do not require display of the soft key. For example, when the document editing function is executed in order to edit the text of the outgoing mail, the normal input key 56a, the completion key 56b, and the submenu key 56c are displayed. The normal input key 56a corresponds to the first menu key 22a, the completion key 56b corresponds to the enter key 22c, the submenu key 56c corresponds to the second menu key 22b, and the user selects the first menu key 22a. By pressing the key 22a, the enter key 22c, and the second menu key 22b, the corresponding soft keys can be operated.

  In other display examples (drawings), the soft key displayed on the left side corresponds to the first menu key 22a, and the soft key displayed on the center corresponds to the enter key 22c and is displayed on the right side. The soft key corresponds to the second menu key 22b.

  First, when the normal input key 56a is operated, the character string input mode can be switched. The display of the character string input mode is displayed on the mode display 58, and in the state of FIG. When the normal input key 56a is operated in this state, the mode is switched to the “normal input mode” in which a character string is input using the plurality of character input keys 22e, and the normal input key 56a is switched as shown in FIG. A voice input key 56d is displayed instead of. The normal input mode is a mode in which characters are input by key input to the plurality of character input keys 22e, and the voice input mode is a mode in which a character string is input by voice recognition. Then, by operating the normal input key 56a and the voice input key 56d, the user can arbitrarily switch the input mode. In addition, when a non-input state for a predetermined time (2 seconds) is detected in the voice input mode, the voice input mode is switched to the normal input mode.

  Next, when the completion key 56b is operated, the text editing of the outgoing mail by the document editing function is finished, and a GUI for inputting the destination and title of the outgoing mail is displayed. Further, when the sub menu key 56c is operated, a sub menu for changing the initial setting of the outgoing mail creation or the document editing function is displayed.

  When a voice meaning a sentence is input to the second microphone 16b in the state where the voice input mode is set, as shown in FIG. 3B, the result of voice recognition is displayed in the function display area 52. A sentence is displayed. A character string with low reliability (hereinafter referred to as a low reliability character string or a low reliability portion) is colored blue in background. In other words, since the low-reliability character string is likely to be a misrecognized character string (hereinafter referred to as a misrecognized character string), in order to guide editing by the user, Are displayed in different forms.

  Furthermore, the portable terminal 10 displays a window Wa on the function display area 52 as shown in FIG. 3C in order to confirm means for editing. This window Wa includes an edit menu for selecting four modes: “1. cursor designation mode”, “2. voice designation mode”, “3. voice search mode”, and “4. normal input mode”. In addition, the selection of each editing menu may be performed by operating the corresponding numeric keys, and a plurality of character input keys 22e are used.

For example, when the “A” line character key corresponding to “1” is pressed, an arbitrary low-reliability character string is displayed in a state designated by the editing cursor CUb, as shown in FIG. The mode display 58 displays a cursor designation mode. In FIG. 3D, an edit key 56e is displayed as a central soft key. Then, the editing cursor CUb is in a state where “economy” which is a low reliability character string is selected. Since this edit cursor CUb can select only a low-reliability character string, the operability of the editing operation can be improved.

  Since the edit key 56e will be described later, a detailed description thereof will be omitted for the sake of simplicity.

  Here, an outline of each of the four modes will be described. First, in the cursor designation mode, each low-reliability character string is designated (selected) by the edit cursor CUb (see FIG. 3D) that can be operated by the direction key 22d, and the designated character string is edited. Mode. For example, in the state of FIG. 3D, when the right direction input is made with the direction key 22d, “medicine” is selected instead of “economy”. Further, when the direction key 22d is further input to the right direction, "Dai" is selected instead of "Medical technique". Further, in the state where “don't care” is selected, if the left direction input is made with the direction key 22d, “medical technique” is selected instead of “not good”. If an operation in the vertical direction is performed instead of the horizontal direction, the editing cursor CUb may move so that the downward direction corresponds to the right direction and the upward direction corresponds to the left direction.

  Thus, since the edit cursor CUb can be operated by the direction key 22d, the user can perform a highly reliable cursor operation.

  Next, in the voice designation mode, when the low-reliability character string is a misrecognized character string and the voice representing the character string is re-input, a character string similar to the character string represented by the re-input voice is obtained. Selected. Moreover, in order to designate this similar character string, it is realizable by dividing and memorize | storing the audio | voice data input initially for every morpheme. Specifically, in each divided speech data, a low reliability speech dictionary is created from each speech data corresponding to a low reliability character string and a corresponding low reliability character string. Then, the CPU 20a selects the low-reliability character string by specifying the re-input speech data and the reference speech data having the highest likelihood using a speech recognition method based on feature pattern matching. The selected character string is replaced with a character string in which the re-input voice is recognized. That is, the selected misrecognized character string is replaced with a newly recognized character string. Thus, the user can edit a misrecognized character string only by speaking the character string for editing. That is, the user can perform an editing operation that is convenient for editing text by voice recognition.

  Also, in the voice search mode, as in the voice designation mode, when the low-reliability character string is a misrecognized character string, the voice representing the misrecognized character string is re-input to re-start from each misrecognized character string. A misrecognized character string corresponding to the input voice is searched. The search result is indicated by the edit cursor CUb. When searching for a misrecognized character string, as in the voice designation mode, it may be designated using a voice recognition technique based on feature pattern matching, or it matches the recognition result of the re-input voice. A character string may be specified. In this way, the user can operate the editing cursor CUb simply by speaking the wrong character string. That is, the user can execute a cursor operation that is convenient for creating a document by voice recognition.

  The normal input mode shown in FIG. 3C is the above-described normal input mode, and the user moves the cursor CUa to the character string (sentence) input by voice recognition using the direction key 22d. A position for editing a sentence (character string) is arbitrarily determined, and characters are input by a plurality of character input keys 22e.

Next, an operation for editing the low reliability character string will be described. Referring to FIG. 4A, the mode display 58 displays a cursor designation mode, and “some” that is a low reliability character string is designated (selected) by the editing cursor CUb. In the key display area 54, a normal input key 56a is displayed on the left side, an edit key 56e in the center, and an end key 56f on the right side. Then, when the edit key 56e is operated, the voice recognition character input to the edit cursor CUb portion or the character input by the character input key 22e is accepted.

  For example, when a character string “location” is input by a plurality of character input keys 22e, the designated low-reliability character string “some” becomes a character string “location” as shown in FIG. Replaced. That is, the user can edit the text by using the plurality of character input keys 22e in an environment that is inappropriate for voice recognition, such as in a train or around a noisy place. The user can also perform a highly reliable editing operation by using the plurality of character input keys 22e.

  If the character input key 22e is not pressed, it is possible to input a character string by voice recognition using the second microphone 16b. When a voice representing “location” is input, the character string is input in the same manner. The designated low reliability character string “some” is replaced with the character string “location”. That is, the user can easily edit using voice recognition.

  When the enter key 56g is operated after editing, the re-input character string is displayed with the background color colored in the same color as the other character strings and further underlined. Note that the character string with the underline added in this way is referred to as a confirmed character string. The confirmed character string becomes editable when the edit key 56e is operated again after being selected by the edit cursor CUb.

  Further, even if it is a low-reliability character string, if it is not an erroneous input, it can be made a confirmed character string by operating the edit key 56e and the confirm key 56g continuously after selecting with the edit cursor CUb. Furthermore, when the end key 56f is selected in a state where the low reliability character string is displayed, the current mode (here, the cursor designation mode) is ended, and another mode can be selected. . For example, when the end key 56f is operated on any screen of FIG. 3D, the window Wa shown in FIG. 3C is displayed.

  When all the low reliability character strings are replaced with the confirmed character strings, a window Wb is displayed as shown in FIG. In the window Wb, a character string “Do you normally input?” And “1. YES” and “2. NO” are displayed.

  For example, when the number key “1” is operated while the window Wb is displayed, the underline added to the confirmed character string is deleted, and the same display as that of the other character strings is obtained. And based on the display position of cursor CUa, it will be in the state which can perform the character input by the several character input key 22e. On the other hand, when the number key “2” is operated, the text editing of the outgoing mail is finished, and the screen is changed to a screen on which a GUI capable of inputting the destination and title of the outgoing mail is displayed.

  Note that all of the low-reliability character strings may be changed to the confirmed character string by long pressing the enter key 22c corresponding to the edit key 56e or the confirm key 56g. Although the cursor designation mode has been described, even in the voice search mode, only the operation for moving the editing cursor CUb is different, and the editing operation for the low-reliability character string is the same.

Next, a reliability table for displaying a low reliability character string in a form different from other character strings will be described. Referring to FIG. 5, the reliability table is composed of a character string column for recording each character string recognized by speech recognition and a reliability column for recording the reliability corresponding to each character string. Yes. For example, in a character string column, a sentence uttered by a user is divided into morpheme units and stored, and character strings such as “economy”, “no”, and “medicine” are stored. On the other hand, in the reliability column, a numerical value representing the reliability as a percentage is recorded, and is recorded based on the result of speech recognition. In other words, if the reliability in speech recognition of “economy” is 50%, “50%” is recorded corresponding to the “economy” column. If the reliability of “NO” is 80%, “80%” is recorded corresponding to the “NO” column, and if the reliability of “Medical” is 40%, the “Medical” column is recorded. "40%" is recorded corresponding to

  Then, a character string of 60% or less in the reliability column of the reliability table is displayed as a low reliability character string. That is, since the reliability corresponding to “economy” and “medical technique” is 60% or less, the background color of “economic” and “medical technique” is displayed in blue as shown in FIG. Is done.

  Next, another embodiment when editing a low-reliability character string by voice input will be described. Referring to FIG. 6, when a low-reliability character string “Daitai” is designated by the edit cursor CUb and a new voice is input by the user, a character string candidate generated by voice recognition is displayed by a pull-down PD. Displayed in a list. Since the list displayed on the pull-down PD is displayed from the top in the descending order of likelihood (reliability), the character string having the highest likelihood is displayed at the top. That is, here, the character strings “not satisfied” and “dirty” are displayed in the pull-down PD as recognition candidates. When a number corresponding to the displayed character string is selected, the designated low reliability character string is replaced with the selected character string. For example, when the number key “1” is operated, “not” is replaced with “not satisfied”. The selection may be made by moving the cursor with the direction key 22d and operating the confirmation key 56g.

  In this way, by displaying a list of voice recognition candidates in a list, the user can edit correctly even if the recognition accuracy of the re-input voice is not high.

  Note that editing the low-reliability character string using the pull-down PD can be executed not only in the cursor designation mode but also in the voice search mode and the voice designation mode. In addition, character strings below a certain value (for example, 40%) are not displayed on the pull-down PD.

  Next, a case where the sub menu key 56c shown in FIGS. 3A to 3C is operated will be described. Referring to FIG. 7A, window Wc is displayed in function display area 52, and in this window Wc, “1. New save”, “2. Edit content confirmation” and “3. Reliability threshold value” are displayed. 3 sub-menus are displayed. Each menu can be selected by operating a numeric key or the like as in the operation of selecting another menu.

  For example, when the “New Save” menu is selected, a process of saving (storing) the data of the outgoing mail in the RAM 30 is executed. When the “confirmation of editing content” menu is selected, a process for displaying a screen for simultaneously confirming the destination, title, and text of the created outgoing mail is executed.

  When the “reliability threshold value” menu is selected, a GUI for changing a threshold value determined to be a low reliability character string is displayed. That is, when the number key “3” is operated, a window Wd shown in FIG. 7B is displayed, and “1. High”, “2. Normal”, and “3. Low” are further displayed in the window Wd. The threshold menu is displayed. The user can change the threshold by selecting an arbitrary threshold menu. For example, when “high” is selected, the threshold is set to 70%, when “normal” is selected, the threshold is set to 60%, and when “low” is selected, the threshold is set to 50%. . Then, when the return key 56h in FIGS. 7A and 7B is operated, the submenu processing is terminated and the screen returns to the screen shown in FIG. 3A and the like.

  The reliability threshold is not limited to three levels, and may be two levels or four or more levels. Further, the reliability threshold value may be designated by an arbitrary numerical value. Further, the processing of the submenu can be executed not only when the body of the outgoing mail is created but also when the destination and the title are input.

  FIG. 8 is an illustrative view showing a memory map of the RAM 30. Referring to FIG. 8, a memory map 300 of RAM 30 includes a program storage area 302 and a data storage area 304. A part of the program and data is read from the flash memory 28 all at once or partially and sequentially as necessary, stored in the RAM 30, and then processed by the CPU 20 or the like.

  The program storage area 302 stores a program for operating the mobile terminal 10. A program for operating the mobile terminal 10 includes a mail function program 310 and a document editing program 312. The mail function program 310 is a program for creating a sent mail and a reply, and displaying a received mail, and further includes a sent mail creating program 310a and a submenu program 310b. The outgoing mail creation program 310a is a program for creating (inputting) the destination, title, and body of the outgoing mail, and the submenu program 310b is a program for saving the outgoing mail.

  The document editing program 312 is a program that is executed when editing the text of the outgoing mail, and further includes a reliability threshold setting program 312a, an arbitrary cursor editing program 312b, a speech recognition input program 312c, a low reliability part. The editing program 312d, the cursor designation program 312e, the voice designation program 312f, the voice search program 312g, and the conversion site search program 312h are configured.

  The reliability threshold setting program 312a is a program for arbitrarily setting a reliability threshold by the user. The arbitrary cursor editing program 312b is a program for editing a sentence by key input or voice input based on a position determined by the cursor CUa, that is, inputting a character string. The voice recognition input program 312c is a process for inputting a character string by voice recognition, and is a program executed in a voice input mode or the like.

  The low reliability part editing program 312d is a program for editing a low reliability character string by key input or voice input. The cursor designation program 312e is a program for selecting and editing a low-reliability character string with the edit cursor CUb. The voice designation program 312f is a program for editing a character string in a part having a high correlation with a newly input voice. The voice search program 312g is a program for editing the low-reliability character string by operating the editing cursor CUb with newly input voice. The conversion site search program 312h is a subroutine of the voice designation program 312f and the voice search program 312g, and is a program for searching for a low-reliability character string based on a newly input voice.

  In addition, although illustration is abbreviate | omitted, the program for operating the portable terminal 10 also includes the program for performing a telephone call, the program which acquires mail data through a network, etc.

Next, referring to FIG. 9, the data storage area 304 is provided with a speech recognition buffer 330 and an input character buffer 332, and further, setting threshold value data 334, reliability table data 336, and low reliability speech dictionary data 338. Mail data 340 is stored, and a cursor designation flag 342, a voice designation flag 344, and a voice search flag 346 are provided.

  The voice recognition buffer 330 is a buffer used when executing voice recognition processing, and temporarily stores voice data for voice recognition converted by the DSP 20b, for example. The input character buffer 332 is a buffer that temporarily stores a character string edited (created) by executing the document editing program 312. The character string is displayed on the LCD monitor 26 using the data stored in the input character buffer 332. The setting threshold value data 334 is threshold value data determined by the processing of the reliability threshold value setting program 312a, and is composed of a numeric string representing, for example, “60%” or “70%”. The reliability table data 336 is data of the reliability table shown in FIG.

  The low-reliability speech dictionary data 338 is composed of speech data corresponding to a low-reliability character string among speech data input by the user, and the low-reliability character string character string. It is used to search for a character string similar to the character string represented by the voice input in the voice search mode. The mail data 340 is data composed of data such as a main body (character string) of a transmitted mail, and data such as a received mail, a transmitted mail, and an unsent mail.

  The cursor designation flag 342 is a flag for determining whether or not the cursor designation mode is set. For example, the cursor designation flag 342 is composed of a 1-bit register. When the cursor designation flag 342 is turned on (established), a data value “1” is set in the register. On the other hand, when the cursor designation flag 342 is turned off (not established), a data value “0” is set in the register. The voice designation flag 344 is a flag for determining whether or not the voice designation mode is set. The voice search flag 346 is a flag for determining whether or not the voice search mode is set. Note that the configurations of the voice designation flag 344 and the voice search flag 346 are the same as those of the cursor designation flag 342, and thus a detailed description of the configuration is omitted.

  Although not shown, the data storage area 304 stores data for displaying images and character strings to be displayed in the status display area 50 and other counters necessary for the operation of the mobile terminal 10. And flags are also provided.

  Under the control of an RTOS (real-time operating system) such as “Linux” and “REX”, the CPU 20a performs outgoing mail creation processing shown in FIG. 10, submenu processing shown in FIG. 11, and reliability threshold setting shown in FIG. 13, the character editing process shown in FIG. 13, the arbitrary cursor editing process shown in FIG. 14, the speech recognition input process shown in FIG. 15, the low-reliability part editing process shown in FIG. 16, the cursor designation process shown in FIG. A plurality of tasks including a voice designation process shown, a voice search process shown in FIG. 19, a conversion part search process shown in FIG. 20, and the like are executed in parallel.

  FIG. 10 is a flowchart showing the outgoing mail creation process. For example, when the user performs an operation of creating a transmission mail, the CPU 20a determines whether or not the operation is an end operation in step S1. That is, it is determined whether or not the operation is to end the creation of the outgoing mail. If “YES” in the step S1, the outgoing mail creation process is ended, and the process returns to the mail function process which is the upper process. On the other hand, if “NO” in the step S1, a step S3 determines whether or not a transmission operation is performed. That is, it is determined whether or not the operation is an operation for transmitting outgoing mail to the network. If “YES” in the step S3, the transmission process is executed in a step S5, and the transmission mail creating process is ended. That is, in step S5, the outgoing mail data is sent to the network.

  If “NO” in the step S5, that is, if the transmission operation is not performed, it is determined whether or not the sub menu is set in a step S7. That is, it is determined whether or not the sub menu key 56c displayed in the key display area 54 has been operated. If “YES” in the step S7, that is, if the submenu key 56c is operated, a submenu process is executed in a step S9, and the process returns to the step S1. Further, since the submenu process executed in step S9 will be described later, a detailed description thereof is omitted here. On the other hand, if “NO” in the step S7, that is, if the submenu key 56c is not operated, it is determined whether or not the title is edited in a step S11. That is, it is determined whether or not the operation is for editing the title of the outgoing mail.

  If “YES” in the step S11, that is, if the operation is to edit the title, the document editing process is executed in a step S13, and after the process of setting the title in a step S15, the process returns to the step S1. On the other hand, if “NO” in the step S11, that is, if the operation is not an operation for editing the title, it is determined whether or not the text is edited in a step S17. That is, in step S17, it is determined whether or not the operation is to edit the text of the outgoing mail. If “YES” in the step S17, the document editing process is executed in a step S19 similarly to the step S13, and the text is set in a step S21, and then the process returns to the step S1. On the other hand, if “NO” in the step S17, that is, if it is not an operation for editing the text, it is determined whether or not the destination is set in a step S23. The document editing process executed in step S13 or step S19 will be described in detail with reference to the flowchart showing the document editing process shown in FIG.

  If “YES” in the step S23, that is, if the process is to set a destination, the destination setting process is executed in a step S25, and the process returns to the step S1. On the other hand, if “NO” in the step S23, that is, if the operation is not an operation for setting a destination, it is determined whether or not data is attached in a step S27. That is, it is determined whether or not an operation for attaching data to the outgoing mail has been performed. If “YES” in the step S27, that is, if the operation is an operation for attaching data, a data attaching process is executed in a step S29, and the process returns to the step S1. On the other hand, if “NO” in the step S27, that is, if it is not an operation for attaching data, the process returns to the step S1.

  FIG. 11 is a flowchart showing the submenu process executed in step S9 (see FIG. 10). When the process of step S9 is executed, the CPU 20a displays a window Wc on the LCD monitor 26 as shown in FIG. 7A, and determines whether the operation is a return operation in step S41. That is, it is determined whether or not the return key 56h is operated. If “YES” in the step S41, that is, if the return key 56h is operated, the submenu process is ended and the process returns to the transmission mail creating process. On the other hand, if “NO” in the step S41, that is, if the return key 56h is not operated, it is determined whether or not a new saving operation is performed in a step S43. For example, it is determined whether or not the numeric key “1” has been operated. If “YES” in the step S43, the process of saving the mail being created is executed in a step S45, and then the process returns to the step S41. That is, in step S45, the transmitted mail is stored (stored) in the RAM 30 as an untransmitted mail.

  If “NO” in the step S43, that is, if the operation is not a new saving operation, it is determined whether or not an editing content confirmation operation is performed in a step S47. That is, it is determined whether or not the numeric key “2” has been operated.

If “YES” in the step S47, a sent mail confirmation display process is executed in a step S49, and the process returns to the step S41. That is, in step S49, a confirmation display process is executed in which the destination, title, and body of the outgoing mail can be confirmed on the same screen. If “NO” in the step S47, it is determined whether or not a reliability threshold value setting operation is performed in a step S51. That is, it is determined whether or not the numeric key “3” has been operated. If “YES” in the step S51, a reliability threshold setting process is executed in a step S53, and the process returns to the step S41. Since the process of step S53 will be described later, detailed description thereof is omitted here. If “NO” in the step S51, the process returns to the step S41 as it is.

  The submenu process may be executed in parallel with the process of creating the text of the outgoing mail, and may be executed as long as the submenu key 56c is displayed.

  FIG. 12 is a flowchart showing the reliability threshold setting process executed in step S53 (see FIG. 11). In step S71, the CPU 20a displays a reliability setting screen. For example, the window Wd is displayed as shown in FIG. Subsequently, in step S73, it is determined whether the operation is a return operation. That is, it is determined whether or not the return key 56h is operated. If “YES” in the step S73, that is, if the return key 56h is operated, the reliability threshold value setting process is ended and the process returns to the sub menu process. On the other hand, if “NO” in the step S73, that is, if the return key 56h is not operated, it is determined whether or not a reliability changing operation is performed in a step S75. For example, it is determined whether any one of the numeric keys “1” to “3” has been operated. If “NO” in the step S75, that is, if the reliability changing operation is not performed, the process returns to the step S73. On the other hand, if the reliability change operation is performed in step S75, the reliability is set according to the change operation in step S77, and the reliability threshold value setting process is terminated. For example, if the “1” numeric key is operated, the reliability threshold is set to 70% (high), and if the “2” numeric key is operated, the reliability threshold is set to 60% (normal). If the number key “3” is operated, the reliability threshold is set to 50% (low). Further, data indicating the set reliability threshold value is stored in the RAM 30 as the set threshold value data 334.

  FIG. 13 is a flowchart showing the document editing process executed in step S13 or step S19 (see FIG. 10). When the process of step S13 or step S19 is executed, the CPU 20a determines whether or not the completion operation is performed in step S91. That is, it is determined whether or not the completion key 56b shown in FIG. If “YES” in the step S91, that is, if the completion key 56b is operated, the document editing process is ended and the process returns to the transmission mail creating process. On the other hand, if “NO” in the step S91, that is, if the completion key 56b is not operated, it is determined whether or not there is a low reliability portion. That is, with reference to the reliability table data 336, it is determined whether or not reliability equal to or lower than the threshold indicated by the setting threshold data 334 is recorded.

  If “NO” in the step S93, that is, if there is no low reliability portion, an arbitrary cursor editing process is executed in a step S95, and the process returns to the step S91. Further, since the process of step S95 will be described later, detailed description thereof is omitted here. If “YES” in the step S93, that is, if there is a low reliability part, a low reliability part editing process is executed in a step S97. Further, since the process of step S97 will be described later, detailed description thereof is omitted here.

Subsequently, in step 99, it is determined whether or not a mode reselection operation is performed. That is, it is determined whether or not the end key 56f (see FIG. 4A) is operated in a state where the low reliability character string is displayed. If “YES” in the step S99, that is, if it is a mode reselection operation, the process returns to the step 97. On the other hand, if “NO” in the step S99, that is, if it is not a mode reselection operation, it is determined whether or not an arbitrary cursor editing is performed in a step S101. That is, the determination is made based on the operation result of selecting “YES” or “NO” shown in the window Wb shown in FIG. If “YES” in the step S101, that is, if “YES” is selected, the process proceeds to a step S95. On the other hand, if “NO” in the step S101, that is, if “NO” is selected, the document editing process is ended, and the process returns to the transmission mail creating process.

  FIG. 14 is a flowchart showing the arbitrary cursor editing process executed in step S95 (see FIG. 13). When the process of step S95 is executed, the CPU 20a determines whether or not it is a confirmed state in step S111. That is, it is determined whether an unconverted hiragana is confirmed. If “YES” in the step S111, that is, if the hiragana is converted and confirmed, the arbitrary cursor editing process is ended, and the process returns to the document editing process. On the other hand, if “NO” in the step S111, that is, if an undetermined character string is not confirmed, the character string is displayed in a step S113. That is, the character string stored in the input character buffer 332 is read and displayed in the function display area 52. If no character string is stored in the input character buffer 332, only the cursor CUa is displayed.

  Subsequently, in step S115, it is determined whether or not a voice recognition operation is performed. That is, it is determined whether or not the voice input key 56d is operated in the normal input mode. If “YES” in the step S115, that is, if the voice input key 56d is operated, a voice recognition input process is executed in a step S117, and the process returns to the step S111. Further, since the process of step S117 will be described later, detailed description thereof is omitted here.

  If “NO” in the step S115, that is, if the voice input key 56d is not operated, it is determined whether or not the direction key is operated in a step S119. That is, it is determined whether or not the direction key 22d has been operated to move the cursor CUa. In the state where only the cursor CUa is displayed in the function display area 52, the display position of the cursor CUa does not change. If “YES” in the step S119, that is, if the direction key 22d is operated, a cursor moving process is executed in a step S121, and the process returns to the step S111. On the other hand, if “NO” in the step S119, that is, if the direction key 22d is not operated, it is determined whether or not a character input operation is performed in a step S123. That is, it is determined whether any one of the plurality of character input keys 22e has been operated.

  If “YES” in the step S123, that is, if a character input operation is performed, a character input process is executed in a step S125, and the process returns to the step S111. That is, in step S 125, hiragana is displayed according to the pressed character key, and the displayed hiragana data is stored in the input character buffer 332. On the other hand, if “NO” in the step S123, that is, if it is not a character input operation, it is determined whether or not it is a conversion operation in a step S127. That is, it is determined whether or not an operation for converting an undetermined hiragana is performed. If “YES” in the step S127, that is, if a conversion operation is performed, a character converting process is executed in a step S129. On the other hand, if “NO” in the step S127, that is, if it is not a conversion operation, the process returns to the step S111.

  FIG. 15 is a flowchart showing the speech recognition input process executed in step S117 (see FIG. 14), later-described step S207 (see FIG. 17) or later-described step S269 (see FIG. 19). When any of step S117, step S207, or step S263 is executed, the CPU 20a determines whether or not a voice is input in step S141. That is, it is determined whether or not sound is input to the second microphone 16b. If “NO” in the step S141, that is, if no sound is input to the second microphone 16b, the process of the step S141 is repeatedly executed. On the other hand, if “YES” in the step S141, that is, if a sound is input to the second microphone 16b, the sound input in the step S143 is converted into sound data. That is, the voice input to the second microphone 16b is converted into voice data by the DSP 20b, and the voice data is stored in the voice recognition buffer 330.

  Subsequently, in step S145, a character string corresponding to the voice data is extracted from the voice dictionary. That is, the voice data is converted into a character string by specifying the reference voice data corresponding to the voice data stored in the voice recognition buffer 330 from the voice dictionary stored in the ROM 32. Note that as a method for identifying reference speech data corresponding to speech data, the speech recognition method based on the above-described feature pattern matching or the speech recognition method based on statistical decision theory is used.

  In step S147, a reliability table corresponding to each extracted character string is created. That is, each character string converted in morpheme units and the likelihood corresponding to each character string are stored in the RAM 30 as reliability table data 336. The CPU 20a that executes the process of step S147 functions as a recording unit.

  Subsequently, in step S149, the extracted character string is displayed based on the position of the cursor CUa (or edit cursor CUb). That is, in step S149, the converted character strings are stored in the input character buffer 332 based on the character string indicated by the cursor CUa. Subsequently, in step S151, the low-reliability speech dictionary data 338 is created, and after the speech recognition input process is completed, the process returns to the main routine. That is, when the CPU 20a that executes the process of step S151 converts to a character string, only the voice data corresponding to the character string whose reliability is equal to or lower than the threshold value from each voice data corresponding to the character string divided in morpheme units. The selected voice data and the character string corresponding to the selected voice data are stored in the RAM 30 as the low-reliability voice dictionary data 338. The CPU 20a that executes the process of step S151 functions as a voice dictionary recording unit.

  FIG. 16 is a flowchart showing the low reliability part editing process executed in step S97 (see FIG. 13). When the process of step S97 is executed, the CPU 20a specifies a low-reliability character string in step S161. That is, the character string below the threshold in the reliability table data 336 is specified. The CPU 20a that executes the process of step S161 functions as a specifying unit.

  Subsequently, in step S163, the background color of the specified character string is changed and displayed. That is, a character string having a reliability level equal to or lower than the threshold value is specified in the reliability table from each character string stored in the input character buffer 332, and the image data displayed on the LCD monitor 26 is changed. For example, if the character string whose reliability is less than or equal to the threshold is “economy”, the character string “economy” stored in the input character buffer 332 is specified, and the image data for displaying the “economy” is changed. To do. In step S165, a mode selection GUI is displayed. For example, as shown in FIG. 3C, a window Wa for displaying a cursor designation mode, a voice designation mode, a voice search mode, and a normal input mode with keys for numbers is displayed. Note that the CPU 20a that executes the processes of step S163 and step S165 functions as an editing guide.

  Subsequently, in step S167, it is determined whether or not the cursor designation mode is set. That is, it is determined whether or not a numeric key corresponding to the cursor designation mode has been operated. If “YES” in the step S167, that is, if the operation is for selecting the cursor designation mode, the cursor designation flag 342 is turned on, and the cursor designation process is executed in a step S169. Further, when the process of step S169 is completed, the cursor designation flag 342 is turned off, the low reliability part editing process is terminated, and the process returns to the document editing process. Further, since the process of step S169 will be described later, detailed description thereof is omitted here.

If “NO” in the step S167, that is, if the operation is not an operation for selecting the cursor designation mode, it is determined whether or not the voice designation mode is made in a voice in a step S171. That means
It is determined whether or not a numeric key corresponding to the voice designation mode has been operated. If “YES” in the step S171, that is, if an operation for selecting the voice designation mode is performed, the voice designation flag 344 is turned on, and the voice designation process is executed in a step S173. Further, when the process of step S173 is completed, the voice designation flag 344 is turned off, and the low reliability part editing process is terminated. Further, since the process of step S173 will be described later, detailed description thereof is omitted here.

  If “NO” in the step S171, that is, if an operation for selecting the voice designation mode is not performed, it is determined whether or not the voice search mode is set in a step S175. That is, it is determined whether or not a numeric key corresponding to the voice search mode has been operated. If “YES” in the step S175, that is, if an operation for selecting the voice search mode is performed, the voice search flag 346 is turned on, and the voice search process is executed in a step S177. Furthermore, when the process of step S177 is finished, the voice search flag 346 is turned off, and the low-reliability part editing process is finished. In addition, since the process of step S177 will be described later, detailed description thereof is omitted here.

  If “NO” in the step S175, that is, if the operation for selecting the voice search mode is not performed, it is determined whether or not the normal input mode is selected in a step S179. That is, it is determined whether or not the numeric key has been operated to select the normal input mode. If “YES” in the step S175, that is, if the normal input mode is selected, the reliability table is updated in a step S181, and the low reliability part editing process is ended. That is, the reliability below the threshold value recorded in the low reliability table is changed to 100%. Thereby, if the user determines that a plurality of low-reliability character strings are not erroneously recognized, the user can resume normal character input by selecting the normal input mode. If “NO” in the step S179, that is, if the operation is not an operation for selecting the normal input mode, the process returns to the step S167.

  FIG. 17 is a flowchart showing the cursor designation process executed in step S169 (see FIG. 16). The process of step S207 is the same as that of step S117, and the process of steps S211 to S215 is the same as that of steps S125 to S129, and thus detailed description thereof is omitted. When the process of step S169 is executed, the CPU 20a determines whether or not the confirmation operation is performed in step S191. For example, it is determined whether or not the edit key 56e shown in FIG. 4A has been pressed for a long time or the enter key 56g has been operated. If “NO” in the step S191, that is, if the confirming operation is not performed, the process proceeds to a step S197. On the other hand, if “YES” in the step S191, that is, if a confirming operation is performed, the reliability table is updated in a step S193.

  For example, the reliability of the character string selected by the editing cursor CUb is changed to 100%. When the edit key 56e or the enter key 56g is pressed for a long time, the reliability column values in the reliability table are all changed to 100% regardless of the character string selected by the edit cursor CUb. Subsequently, in step S195, it is determined whether or not there is a low reliability portion. That is, it is determined whether or not the reliability below the threshold is recorded in the trust table data 338. If “NO” in the step S195, that is, if the reliability below the threshold is not recorded, the cursor designating process is terminated, and the process returns to the low reliability part editing process.

  If “YES” in the step S195, that is, if a reliability equal to or lower than the threshold is recorded, it is determined whether or not the end operation is performed in a step S197. That is, it is determined whether or not the end key 56f shown in FIGS. 4A and 4B has been operated. If “YES” in the step S197, that is, if the end key 56f is operated, the cursor designation processing is ended. On the other hand, if “NO” in the step S197, that is, if the end key 56f is not operated, it is determined whether or not the direction key is operated in a step S199. That is, it is determined whether or not the direction key 22d has been operated.

  If “YES” in the step S199, that is, if the direction key 22d is operated, the display position of the editing cursor CUb is updated in a step S201, and the process returns to the step S191. That is, in step S201, the reliability table data 336 is referred to, and another low reliability character string is selected according to the input direction. For example, referring to FIG. 3D, when “Economy” is the currently selected low-reliability character string and an operation in the right direction (or downward direction) is performed, it is recorded after “Economy”. The low-reliability character string, that is, “medical technique” is selected by the editing cursor CUb. In addition, “medicine” is the currently selected low-reliability character string, and if a leftward (or upward) operation is performed, the low-reliability character string recorded before “medicine”, that is, “Economy” is selected by the editing cursor CUb.

  In the reliability table, when the low reliability character string recorded at the top is selected and an upward operation is performed, the display position of the edit cursor CUb need not be updated. Alternatively, the reliability character string recorded at the bottom of the reliability table may be selected. If the low-reliability character string selected by the edit cursor CUb is not displayed in the function display area 52, the displayed character string is updated, and the edit cursor CUb and the selected low-reliability character string are updated. Is displayed.

  If “NO” in the step S199, that is, if the direction key 22d is not operated, it is determined whether or not the editing operation is performed in a step S203. That is, it is determined whether or not the edit key 56e has been operated. If “NO” in the step S203, the process returns to the step S191. On the other hand, if “YES” in the step S203, it is determined whether or not a voice recognition operation is performed in a step S205. For example, after the edit key 56e is operated, the presence / absence of voice input is determined. If “YES” in the step S205, that is, if it is a voice recognition operation, thereafter, the voice recognition mode is stored, the voice recognition input process is executed in a step S207, and then the process returns to the step S191. For example, the CPU 20a turns on a voice recognition mode flag (not shown in FIG. 9) in order to store the voice recognition mode.

  If “NO” in the step S205, that is, if it is not a voice recognition operation, it is determined whether or not a character input operation is performed in a step S209. For example, it is determined whether or not the character input key 22e has been operated after the edit key 56e has been operated. If “YES” in the step S209, the character input mode is stored thereafter, and after the character input process is executed in a step S211, the process returns to the step S191. For example, the CPU 20a turns on a character input mode flag (not shown in FIG. 9) in order to store that it is in the character input mode.

  On the other hand, if “NO” in the step S209, that is, if it is not a character input operation, it is determined whether or not it is a conversion operation in a step S213. If “YES” in the step S213, a character conversion process is executed in a step S215, and the process returns to the step S191. On the other hand, if “NO” in the step S213, the process returns to the step S191 as it is.

  In addition, CPU20a which performs step S211 and step S215 functions as a character editing means.

FIG. 18 is a flowchart showing the voice designation process executed in step S173 (see FIG. 16). In addition, since the process of step S231 and step S233 is the same as the process of step S197 and step S195, detailed description is abbreviate | omitted. When the process of step S173 is executed, the CPU 20a determines whether or not the end operation is performed in step S231. If “YES” in the step S231, the voice designating process is ended, and the process returns to the low reliability part editing process. On the other hand, if “NO” in the step S231, it is determined whether or not there is a low reliability part in a step S233. If “NO” in the step S233, the voice specifying process is ended.

  On the other hand, if “YES” in the step S233, it is determined whether or not a direction key operation is performed in a step S235. That is, it is determined whether or not the direction key 22d has been operated. If “YES” in the step S235, that is, if the direction key 22d is operated, a display scrolling process is executed in a step S237, and the process returns to the step S231. That is, in step S237, the character string displayed in the function display area 52 is scrolled (changed) in accordance with the input direction. For example, if an input is made in the downward direction, the screen is scrolled downward so that the text of the outgoing mail that has not yet been displayed is displayed. Further, if an upward input is made, the scroll is performed upward so as to display the text of the already displayed outgoing mail.

  As described above, the user can easily search for other low-reliability character strings by scrolling the contents of the character strings displayed on the LCD monitor 26.

  If there is no character string that is not displayed, the display in the function display area 52 does not change even if an input is made to the direction key 22d. Similarly, when the left / right direction key is input, the display in the function display area 52 does not change.

  If “NO” in the step S235, that is, if the direction key 22d is not operated, it is determined whether or not a sound is input in a step S239. That is, it is determined whether or not sound is input to the second microphone 16b. If “YES” in the step S239, that is, if a voice is input, the conversion part search process is executed in a step S241, and then the process returns to the step S231. Since the process of step S241 will be described later, a detailed description thereof is omitted here. On the other hand, if “NO” in the step S239, that is, if no voice is input, the process returns to the step S231.

  FIG. 19 is a flowchart showing the voice search process executed in step S177 (see FIG. 16). Note that the processing of steps S251 to S257 is the same as the processing of steps S191 to S197, the processing of steps S259 and S261 is the same as steps S235 and S237, the processing of step S265 is the same as step S241, and step S269. Since this process is the same as step S117 or step S207, detailed description thereof is omitted.

  When the process of step S177 is executed, the CPU 20a determines in step S251 whether or not it is a confirmation operation. If “NO” in the step S251, the process proceeds to a step S257, and if “YES”, the reliability table is updated in a step S253. Subsequently, in step S255, it is determined whether or not there is a low-reliability part. If “NO”, the voice search process is terminated and the process returns to the low-reliability editing process. On the other hand, if “YES” in the step S255, it is determined whether or not an end operation is performed in a step S257. If “YES” in the step S257, the voice search process is ended, and if “NO”, it is determined whether or not the direction key operation is performed in a step S259.

If “YES” in the step S259, the display scroll process is executed in a step S261, and the process returns to the step S251. On the other hand, if “NO” in the step S259, it is determined whether or not a voice is input in a step S263. That is, it is determined whether or not sound is input by the second microphone 16b. If “YES” in the step S263, that is, if a voice is inputted, a conversion part search process is executed in a step S265, and the process returns to the step S251. For example, if a voice representing an arbitrary low-reliability character string is input, an arbitrary low-reliability character string is selected by the editing cursor CUb when the process of step S265 ends.

  If “NO” in the step S263, that is, if no sound is input, it is determined whether or not an input operation is performed in a step S267. For example, it is determined whether or not the edit key 56e has been operated. If “YES” in the step S267, that is, if an input operation is performed, a voice recognition input process is executed in a step S269, and the process returns to the step S251. If “NO” in the step S267, the process returns to the step S251. For example, when the process of step S269 ends, the voice input after the editing key 56e is operated is converted into a character string, and is replaced with the selected low-reliability character string.

  Note that after the edit key 56e is operated, the low-reliability character string may be edited by normal input instead of voice input. In other words, after “YES” is determined in step S267, the processing in steps S205 to S215 may be executed instead of S269. And after the process of step S207, S211, and S215 is complete | finished, it is made to return to step S251.

  FIG. 20 is a flowchart showing conversion site search processing executed in step S241 (see FIG. 18) or step S265 (see FIG. 19). When step S241 or step S265 is executed, the CPU 20a converts the input voice into voice data in step S281. That is, the input voice is converted into voice data by the DSP 20a. Subsequently, in step S283, the low reliability speech dictionary data 338 is read. That is, each piece of voice data constituting the low reliability voice dictionary data 338 is read as reference voice data.

  Subsequently, in step S285, a highly correlated part is searched. Specifically, a plurality of feature patterns that change at fixed time intervals are acquired from the reference speech data and the input speech data, and further, correlation is performed from each feature pattern of the reference speech data and the input speech data. Calculate the value. Then, by extracting the character string represented by the reference speech data corresponding to the largest correlation value, the low reliability character string that matches the extracted character string is searched from the reliability table data 336. Thus, in this embodiment, the correlation function can be used to search for a similar character string. Note that the CPU 20a that executes the processes of steps S283 and S285 functions as a similarity search unit.

  Subsequently, in step S287, it is determined whether or not the voice search mode is set. That is, it is determined whether or not the voice search flag 346 is on. If “YES” in the step S287, that is, in the voice search mode, the display position of the editing cursor CUb is updated in accordance with the search result in a step S289, and after the conversion part search process is completed, the voice search process is performed. Return. For example, if the similar character string is “medical technique”, “medical technique” is selected by the editing cursor CUb. In this manner, the low-reliability character string intended by the user can be selected regardless of the result of speech recognition of the re-input speech. As described above, the user can execute a cursor operation convenient for creating a document by voice recognition.

  If “NO” in the step S287, that is, if the voice search flag 346 is turned off and the voice designation flag 344 is turned on, a character string corresponding to the voice data is extracted from the voice dictionary in a step S291. That is, a character string is extracted from the speech dictionary stored in the ROM 32 as in step S145.

Subsequently, in step S293, the extracted character string is displayed based on the highly correlated part. That is, the character string indicated by the search result obtained in step S285 is identified from the input character buffer 332 and replaced with the character string extracted from the speech dictionary. For example, if the low-reliability character string of the search result is “economy” and the character string extracted from the speech dictionary is “present”, the character string “economic” is “current” in the function display area 52. Replaced by a column. Subsequently, in step S295, the reliability table is updated, and after the conversion site search process is completed, the process returns to the voice designation process. For example, in step S295, “economic” recorded in the reliability table is replaced with “present”, and the likelihood calculated when voice recognition of “present” is recorded as reliability.

  The CPU 20a that executes the process of step S201 or step S289 functions as a cursor display unit. Further, the CPU 20a that executes the process of step S293 functions as a replacement unit. Further, the CPU 20a that executes the processes of step S201, step S239, and step S261 functions as a scroll unit.

  Here, the editing of the character string using the pull-down PD shown in FIG. 6 will be described in detail with reference to the flowchart of the speech recognition process shown in FIG. In addition, since the process of step S141-S151 has already been demonstrated in detail, detailed description is abbreviate | omitted here and it demonstrates from the process of step S311.

  In step S311, it is determined whether or not the low-reliability character string is to be edited. That is, it is determined whether any of the cursor designation flag 342, the voice designation flag 344, or the voice search flag 346 is on. If “NO” in the step S311, that is, if the low-reliability character string is not edited, the processing in the step S145 and the subsequent steps is executed. On the other hand, if “YES” in the step S311, a plurality of character strings corresponding to the voice data are extracted from the voice dictionary. That is, not only the character string having the highest likelihood but also all character strings having a likelihood equal to or greater than a certain value are extracted.

  In step S315, a pull-down menu is displayed. That is, the pull-down PD shown in FIG. 6 is displayed, and a plurality of extracted character strings are displayed in the pull-down PD. The CPU 20a that executes the process of step S315 functions as a list display unit. Subsequently, in step S317, the reliability corresponding to the selected character string is recorded. That is, the likelihood of the selected character string, that is, the reliability is recorded in the reliability table. When recording the reliability, the character string selected by the editing cursor CUb and the corresponding reliability are overwritten.

  Subsequently, in step S319, the selected character string is displayed, and the speech recognition input process is terminated. That is, the low reliability character string stored in the input character buffer 332 is replaced with the selected character string. For example, referring to FIG. 6, if “not satisfied” in the pull-down PD is selected, “not required” is replaced with “not satisfied”.

  In addition, a process for searching for a low-reliability character string that matches a newly recognized character string instead of searching for a similar character string in the voice search mode will be described in detail with reference to FIG. Note that the CPU 20a that executes the processes of steps S149 and S319 functions as a voice editing unit.

  Referring to FIG. 22, in another embodiment, the processing contents in steps S281-S287 and S291-S295 in the conversion site search processing are the same, and thus detailed description thereof is omitted.

  The CPU 20a converts the voice input in step S281 into voice data, and then extracts a character string corresponding to the voice data from the voice dictionary in step S291. When the process of step S291 ends, it is determined whether or not the voice search mode is set in step S287.

  If “NO” in the step S287, that is, if the voice designation mode is not the voice search mode, the process is executed in the order of steps S283, S285, S293, and S295, and the conversion site search process is ended. On the other hand, if “YES” in the step S287, that is, if the voice search mode is set, a low-reliability character string that matches the character string extracted in the step S331 is searched. That is, the character string extracted in step S291 is searched from the character string column in the reliability table. The CPU 20a that executes the process of step S331 functions as a search unit.

  Subsequently, in step S289, the display position of the edit cursor CUb is updated according to the search result. That is, the display position of the edit cursor CUb is updated according to the search result in the process of step S331. For example, if the recognition result of the newly input voice is “economy”, “economy”, which is a low recognition character string, becomes the search result. Then, “economy” is selected by the editing cursor CUb.

  As can be understood from the above description, the mobile terminal 10 includes the second microphone 16b that captures the user's voice, and recognizes the voice input to the second microphone 16b to generate a character string. When a character string is generated by speech recognition, the calculated likelihood is used as the reliability of speech recognition, and the generated character string and the reliability corresponding to the character string are recorded in the reliability table. Then, based on the reliability table, a character string having a reliability equal to or lower than the threshold is specified, and the background color of the specified low reliability character string is colored blue and displayed on the LCD monitor 26.

  As a result, the misrecognized character string candidates are displayed so that they can be determined at a glance, so that the user can easily determine whether or not editing is necessary, and can efficiently create a sentence using voice recognition.

  It should be noted that if a function (expression) for calculating likelihood is set for each of reference speech data constituting dictionary data in speech recognition (including speech dictionary and low-reliability speech dictionary data stored in ROM 32). Each function for calculating likelihood may be used to search for a similar character string. That is, a function ID for identifying each function is set, and the function ID used for calculating the reliability of the low reliability character string is recorded. Then, it is possible to search for a similar low-reliability character string by searching the function ID used when recognizing the newly input voice from the recorded function IDs.

  Further, the microphone used for voice recognition is not limited to the second microphone 16b, but may be the first microphone 16a. Furthermore, even if the portable terminal 10 does not include a microphone, it is possible to obtain the effect of the present embodiment by installing a program capable of performing each process shown in FIGS. It may be.

  In addition, the document editing function of the present embodiment is applicable to any function for inputting a character string such as a memo pad function, which is not limited to editing the text of the outgoing mail.

  Further, each menu that can be selected in the windows Wa-Wd may be selected by a dedicated cursor for selecting a menu, instead of being selected by a numeric key.

  Further, the communication method of the mobile terminal 10 is not limited to the CDMA method, but may be a W-CDMA method, a TDMA method, a PHS method, a GSM method, or the like. In addition, each process executed by the CPU 20a of the mobile terminal 10 in the present embodiment is not limited to the mobile terminal 10, but may be a mobile information terminal such as a PDA (Personal Digital Assistant) or a personal computer (PC). Good.

DESCRIPTION OF SYMBOLS 10 ... Portable terminal 16a ... 1st microphone 16b ... 2nd microphone 20a ... CPU
20b DSP
22 ... Key input device 26 ... LCD monitor 30 ... RAM
32 ... ROM

The present invention relates to a mobile terminal , an edit guide program, and an edit guide method , and more particularly to a mobile terminal , an edit guide program, and an edit guide method for inputting a character string by voice recognition, for example.

Therefore, a main object of the present invention is to provide a novel portable terminal , editing guide program, and editing guide method .

Another object of the present invention is to provide a portable terminal , an editing guide program, and an editing guide method that can increase the efficiency of sentence creation by voice recognition.

A first aspect of the present invention is a character input key, have a voice recognition unit for generating a character string from the speech signal captured by the microphone and the microphone, the voice input to the first mode or a microphone capable key input to the character input key A portable terminal capable of inputting characters in a possible second mode, by a first receiving means for receiving a switching operation to the first mode, a second receiving means for receiving a switching operation to the second mode, and a voice recognition means storage means for storing the generated character string, in a state of being switched to the contact and a second mode, the character string stored by the storage means includes a table Shimesuru display means, in a state of switched on second mode, The portable terminal is switched to the first mode when a non-input state for a predetermined time is detected .

In the first invention, the mobile terminal (10) is a character input key (22e), microphone (16a, 16b) for speech recognition and speech recognition means for generating a character string from the audio signal captured by a microphone (20a , 20b, 30) . This voice recognition means includes, for example, a CPU (20a), a DSP (20b), and a ROM ( 32 ) for storing voice dictionary data. The portable terminal can input arbitrary characters in the first mode in which key input with respect to the character input key is possible or in the second mode in which voice input with respect to the microphone is possible. The first accepting unit accepts a switching operation to the first mode. The second accepting unit accepts a switching operation to the second mode. The storage means stores the character string generated by the voice recognition means. A display means displays the character string memorize | stored by the memory | storage means in the state switched to the 2nd mode. Then, when an uninput state for a predetermined time is detected in the state switched to the second mode, the mode is switched to the first mode.

According to the first invention, using a person will be able to efficiently create a sentence that uses speech recognition.

The second invention is dependent on the first invention, and further comprises editing means for editing the character string displayed by the display means.

A third invention is a character input key, have a microphone (16a, 16b) and speech recognition means for generating a character string from the audio signal captured by a microphone (20a, 20b, 30), the key input to the character input keys A first receiving means for receiving a processor (20a) of the portable terminal (10) capable of inputting characters in the first mode capable of inputting voice or in the second mode capable of inputting voice to the microphone ; second receiving means for receiving a switching operation to the second mode, storage means for storing a string generated by the speech recognition means, in a state of being switched contact and the second mode, the character string stored by the storage means to function as a table Shimesuru display means, in a state of switched on second mode, when the non-input state of the predetermined time is detected, the first mode Ri frogs, is editing induction program.

In the third invention, similarly to the first invention, the user can efficiently create a sentence using voice recognition.

4th invention has the voice recognition means (20a, 20b, 30) which produces | generates a character string from the audio | voice signal taken in by the character input key, the microphone, and the microphone, and the 1st key input with respect to a character input key is possible An edit guiding method in a portable terminal capable of inputting characters in a second mode in which voice input to the mode or the microphone can be performed, wherein the processor of the portable terminal accepts a switching operation to the first mode; A second receiving step for accepting an operation for switching to the two mode; a storing step for storing the character string generated by the voice recognition means; and a character string stored by the storing step in a state switched to the second mode. When a non-input state for a predetermined time is detected while the display step is executed and the mode is switched to the second mode Switches to the first mode, a edit induction method.

In the fourth invention as well, as in the first invention, the user can efficiently create a sentence using voice recognition.

Further, the mobile terminal 10 has a document editing function for inputting and deleting a character string, and can perform character input by voice recognition. That is, when the user reads out the sentence and inputs the voice to the second microphone 16b, the sentence whose voice is recognized is displayed on the LCD monitor 26. Specifically, the audio signal captured by the second microphone 16b is converted into audio data by the DSP 20b, and the CPU 20a and the DSP 20b extract a feature pattern (feature amount) from the audio data. Further, the CPU 20a and the DSP 20b read the reference voice data constituting the voice dictionary for voice recognition from the ROM 32, and extract a feature pattern of the reference voice data (hereinafter referred to as a reference pattern) or directly read the reference pattern from the ROM 32. read out. Then, the CPU 20a and the DSP 20b collate the feature pattern with each reference pattern by a speech recognition method based on feature pattern matching or a speech recognition method based on statistical decision theory, thereby matching the reference speech data with the speech data. Is identified. Since this voice dictionary associates reference voice data with a character string represented by the reference voice data, the CPU 20a and the DSP 20b read the character string corresponding to the identified reference voice data, thereby converting the voice data into a character string. Convert to

That is, the caller inputs the transmission voice to the first microphone 16a through the opening op1 and listens to the reception voice from the speaker 18 through the opening op2. The user inputs voice for voice recognition to the second microphone 16b through the opening op3. The first microphone 16a can also be used for voice recognition. And the technique of canceling the ambient noise which is a distant sound source is realizable by the difference of the audio | voice signal of the 1st microphone 16a and the 2nd microphone 16b, and it contributes to the improvement of a speech recognition rate.

Next, in the audio designation mode, when low confidence string is erroneously recognized character strings, the speech representing the character string is re-entered, a character string similar to a string that represents the audio that the reenter Selected. Moreover, in order to designate this similar character string, it is realizable by dividing and memorize | storing the audio | voice data input initially for every morpheme. Specifically, in each divided speech data, a low reliability speech dictionary is created from each speech data corresponding to a low reliability character string and a corresponding low reliability character string. Then, the CPU 20a selects the low-reliability character string by specifying the re-input speech data and the reference speech data having the highest likelihood using a speech recognition method based on feature pattern matching. The selected character string is replaced with a character string in which the re-input voice is recognized. That is, the selected misrecognized character string is replaced with a newly recognized character string. Thus, the user can edit a misrecognized character string only by speaking the character string for editing. That is, the user can perform an editing operation that is convenient for editing text by voice recognition.

If the character input key 22e is not pressed, it is possible to input a character string by speech recognition using the second microphone 16b. The designated low reliability character string “some” is replaced with the character string “location”. That is, the user can easily edit using voice recognition.

FIG. 8 is an illustrative view showing a memory map of the RAM 30. Referring to FIG. 8, a memory map 300 of RAM 30 includes a program storage area 302 and a data storage area 304. Some programs and data are partially and sequentially read out in accordance with all or required at a time from the flash memory 28, is treated with a CPU 20 a after being stored in the RAM 30.

The document editing program 312 is a program that is executed when editing or the like of the transmission mail text, further confidence threshold value setting program 312a, any cursor editing program 312b, a voice recognition input program 312c, a low confidence region The editing program 312d, the cursor designation program 312e, the voice designation program 312f, the voice search program 312g, and the conversion site search program 312h are configured.

Subsequently, referring to FIG. 9, the data storage area 304 is provided with a speech recognition buffer 330 and an input character buffer 332, and further, setting reliability data 334, reliability table data 336, low reliability speech dictionary data. 338 and mail data 340 are stored, and a cursor designation flag 342, a voice designation flag 344, and a voice search flag 346 are provided.

The voice recognition buffer 330 is a buffer used when executing voice recognition processing, and temporarily stores voice data for voice recognition converted by the DSP 20b, for example. The input character buffer 332 is a buffer that temporarily stores a character string edited (created) by executing the document editing program 312. The character string is displayed on the LCD monitor 26 using the data stored in the input character buffer 332. Setting reliability data 334 is data in the threshold determined by the processing of the reliability threshold value setting program 312a, and a string of digits representing the like for example "60%" and "70%". The reliability table data 336 is data of the reliability table shown in FIG.

The CPU 20a, under the control of an RTOS (real-time operating system) such as “Linux (registered trademark) ” and “REX”, creates outgoing mail as shown in FIG. 10, submenu processing as shown in FIG. 11, and as shown in FIG. Reliability threshold setting processing, character editing processing shown in FIG. 13, arbitrary cursor editing processing shown in FIG. 14, speech recognition input processing shown in FIG. 15, low reliability part editing processing shown in FIG. 16, cursor designation processing shown in FIG. A plurality of tasks including a voice designation process shown in FIG. 18, a voice search process shown in FIG. 19, a conversion part search process shown in FIG. 20, and the like are executed in parallel.

If "NO" in the step S 3, that is, if the transmission operation, it is determined whether the setting or not the sub-menu in step S7. That is, it is determined whether or not the sub menu key 56c displayed in the key display area 54 has been operated. If “YES” in the step S7, that is, if the submenu key 56c is operated, a submenu process is executed in a step S9, and the process returns to the step S1. Further, since the submenu process executed in step S9 will be described later, a detailed description thereof is omitted here. On the other hand, if “NO” in the step S7, that is, if the submenu key 56c is not operated, it is determined whether or not the title is edited in a step S11. That is, it is determined whether or not the operation is for editing the title of the outgoing mail.

Figure 12 is a flow diagram illustrating a reliability threshold value setting process executed in step S53 (see FIG. 11). In step S71, the CPU 20a displays a reliability setting screen. For example, the window Wd is displayed as shown in FIG. Subsequently, in step S73, it is determined whether the operation is a return operation. That is, it is determined whether or not the return key 56h is operated. If in step S73 "YES", if it is operated ie return key 56h is, exit the reliability threshold value setting process and returns to the submenu processing. On the other hand, if “NO” in the step S73, that is, if the return key 56h is not operated, it is determined whether or not a reliability changing operation is performed in a step S75. For example, it is determined whether any one of the numeric keys “1” to “3” has been operated. If “NO” in the step S75, that is, if the reliability changing operation is not performed, the process returns to the step S73. On the other hand, if the reliability change operation is performed in step S75, the reliability is set according to the change operation in step S77, and the reliability threshold value setting process is terminated. For example, if the “1” numeric key is operated, the reliability threshold is set to 70% (high), and if the “2” numeric key is operated, the reliability threshold is set to 60% (normal). If the number key “3” is operated, the reliability threshold is set to 50% (low). Data indicating the set reliability threshold value is stored in the RAM 30 as set reliability data 334.

FIG. 13 is a flowchart showing the document editing process executed in step S13 or step S19 (see FIG. 10). When the process of step S13 or step S19 is executed, the CPU 20a determines whether or not the completion operation is performed in step S91. That is, it is determined whether or not the completion key 56b shown in FIG. If “YES” in the step S91, that is, if the completion key 56b is operated, the document editing process is ended and the process returns to the transmission mail creating process. On the other hand, if “NO” in the step S91, that is, if the completion key 56b is not operated, it is determined whether or not there is a low reliability portion. That is, with reference to the reliability table data 336, it is determined whether or not reliability equal to or lower than the threshold indicated by the set reliability data 334 is recorded.

For example, the reliability of the character string selected by the editing cursor CUb is changed to 100%. When the edit key 56e or the enter key 56g is pressed for a long time, the reliability column values in the reliability table are all changed to 100% regardless of the character string selected by the edit cursor CUb. Subsequently, in step S195, it is determined whether or not there is a low reliability portion. In other words, the reliability degree table data 33 6, the following confidence threshold to determine whether it is recorded. If “NO” in the step S195, that is, if the reliability below the threshold is not recorded, the cursor designating process is terminated, and the process returns to the low reliability part editing process.

If there is no character string is not displayed, be the input to the direction key 22 d, the display of the function display area 52 does not change. Similarly, if the left-right direction key is input, the display of the function display area 52 does not change.

It should be noted that if a function (expression) for calculating likelihood is set for each of reference speech data constituting dictionary data in speech recognition (including speech dictionary and low-reliability speech dictionary data stored in ROM 32). Each function for calculating likelihood may be used to search for a similar character string. In other words, a function ID for identifying each function is set, and the function ID used for calculating the reliability of the low-reliability character string is recorded. Then, it is possible to search for a similar low-reliability character string by searching the function ID used when recognizing the newly input voice from the recorded function IDs.

Further, the communication method of the mobile terminal 10 is not limited to the CDMA method, but may be a W-CDMA method, a TDMA method, a PHS method, a GSM (registered trademark) method, or the like. In addition, each process executed by the CPU 20a of the mobile terminal 10 in the present embodiment is not limited to the mobile terminal 10, but may be a mobile information terminal such as a PDA (Personal Digital Assistant) or a personal computer (PC). Good.

Claims (16)

A portable terminal having a capturing means for capturing a speech signal and a speech recognition means for generating a character string from the speech signal captured by the capturing means,
Recording means for recording the character strings generated by the speech recognition means and data indicating their reliability;
A portable unit comprising: a specifying unit that specifies a character string having a reliability equal to or lower than a predetermined value with reference to the data; and a display unit that displays the character string specified by the specifying unit in a form different from other character strings. Terminal.   The mobile terminal according to claim 1, further comprising a cursor display unit that displays a cursor for selecting only the specified character string. An operation means for receiving an operation for moving the cursor;
The portable terminal according to claim 2, wherein the cursor selects a character string according to an operation result by the operation means. Search means for searching for the character string that matches a new character string that is newly voice-recognized after the character string is generated,
The mobile terminal according to claim 2, wherein the cursor selects the character string searched by the search means.   The mobile terminal according to claim 2, further comprising voice editing means for editing a character string selected by the cursor based on a character string newly generated by the voice recognition means. A list display means for displaying a list of character string candidates generated by the voice recognition means;
The portable terminal according to claim 5, wherein when the candidate displayed by the candidate list display unit is selected, the voice editing unit edits the selected candidate as a newly generated character string. The character input means for inputting a character string, and the character edit means for editing the character string selected by the cursor based on the character string input by the character input means. The portable terminal as described in. Similar search means for searching for a character string similar to the character string newly generated by the voice recognition means, and the character string searched by the similarity search means as a character string newly generated by the voice recognition means The mobile terminal according to claim 1, further comprising replacement means for replacing. Similarity search means for searching for a character string similar to the character string newly generated by the voice recognition means,
The portable terminal according to claim 2, wherein the cursor selects a character string searched by the similarity search unit.   The mobile terminal according to claim 9, further comprising voice editing means for editing a character string selected by the cursor based on a character string newly generated by the voice recognition means. A list display means for displaying a list of character string candidates generated by the voice recognition means;
The portable terminal according to claim 10, wherein when the candidate displayed by the candidate list display unit is selected, the voice editing unit edits the selected candidate as a newly generated character string. The portable terminal according to claim 9, further comprising: a character input unit that inputs a character string; and a character editing unit that edits the character string selected by the cursor based on the character string input by the character input unit. Voice dictionary recording means for recording the voice captured by the capture means and the character string generated from the voice as a voice dictionary;
The similarity search means searches for a similar character string by calculating a correlation value between each of the voices recorded by the voice dictionary recording means and the newly input voice. The mobile terminal according to Crab. The portable terminal according to claim 1, further comprising: a display unit that displays at least a part of the plurality of character strings; and a scroll unit that scrolls a display position of the character string displayed by the display unit. . A portable terminal processor having a capturing means for capturing a speech signal and a speech recognition means for generating a character string from the speech signal captured by the capturing means,
Recording means for recording the character strings generated by the speech recognition means and data indicating their reliability;
A specifying means for specifying a character string having a reliability equal to or lower than a predetermined value with reference to the data, and a function for displaying the character string specified by the specifying means in a form different from other character strings; Editing guidance program. An editing apparatus having a capturing means for capturing a speech signal and a speech recognition means for generating a character string from the speech signal captured by the capturing means,
Recording means for recording the character strings generated by the speech recognition means and data indicating their reliability;
Editing means comprising: specifying means for specifying a character string having a reliability equal to or lower than a predetermined value with reference to the data; and display means for displaying the character string specified by the specifying means in a form different from other character strings. apparatus.

Images