JP3982289B2 - Voice recognition device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a speech recognition apparatus that recognizes input speech and presents selection candidates for the input speech to enable selection.
[0002]
[Prior art]
A conventional speech recognition apparatus is disclosed in Japanese Patent Application Laid-Open No. 11-352991. This speech recognition device recognizes speech uttered by dividing into single syllables and displays recognition candidates. The next recognition candidate can be sequentially displayed until the desired recognition candidate is confirmed by the voice input person.
[0003]
[Problems to be solved by the invention]
However, in the conventional speech recognition apparatus, for example, when a high level of noise is mixed during speech input and erroneous recognition occurs, the desired recognition candidate may not be displayed even if the recognition candidates are sequentially displayed. Therefore, the selection operation must be performed while confirming the recognition candidates one by one without knowing whether there is a correct recognition candidate.
[0004]
Further, it is conceivable to change the order of recognition candidates and arrange recognition candidates that are easily misrecognized with candidates having higher matching degree with speech. However, when the recognition candidate with the first matching score is erroneous recognition and the recognition candidate with the second matching score is correct, and the first recognition candidate and other words are likely to be erroneously recognized. Problems arise. In this case, when the above measures are taken, the order of appearance is the first recognition candidate, the word that is easily misrecognized, and the second recognition candidate in this order, and the appearance order of the recognition candidate with the second highest match. Will go down. It is desirable to arrange the recognition candidates in order of the degree of coincidence.
[0005]
The object of the present invention is to select a recognition candidate using an operation device, and if there is a recognition target word that is easily misrecognized among the recognition candidates, a simple operation can be performed regardless of the degree of coincidence with speech. An object of the present invention is to provide a speech recognition apparatus that can select a recognition target word.
[0006]
[Means for solving the problems]
(1) A speech recognition device according to the invention of claim 1 includes a speech input device for inputting speech, a plurality of recognition target words, and individual recognition target words and other recognition target words that are easily misrecognized. A storage device that stores the correspondence; a degree of coincidence between the input speech and each recognition target word; a candidate extraction device that sets a plurality of recognition target words as a plurality of recognition candidates based on the degree of coincidence; and a plurality of recognitions A display device that displays the candidates side by side in the first direction, and displays other recognition target words that are likely to be erroneously recognized with respect to the recognition candidates in the second direction; a first operation corresponding to the first direction; an operation unit for performing a second operation corresponding to the second direction, and selects from the recognition candidates by a first operation, which are easily recognized by other erroneous against recognition candidate selected by the second operation and a selection device for selecting from the recognition target word, If the existing recognition candidates of the recognized easily recognized target word is selected by the first operation, characterized in that it comprises a force generator for applying a force in a second direction relative to the operating device.
(2) The invention of claim 2 stores a correspondence between a speech input device for inputting speech and a plurality of recognition target words and each recognition target word and another recognition target word that is easily misrecognized. A storage device, a candidate extraction device that calculates the degree of coincidence between the input speech and each recognition target word and sets a plurality of recognition target words as a plurality of recognition candidates based on the degree of coincidence, and a plurality of recognition candidates as a first A display device that displays other words to be recognized that are likely to be erroneously recognized with respect to the recognition candidates, and displays the first recognition operation and the second direction corresponding to the first direction. An operation device that performs a second operation corresponding to, and another recognition target word that is selected from the recognition candidates by the first operation and that is easily misrecognized with respect to the recognition candidate selected by the second operation A selection device for selecting from the above and a selection by the first operation. A force generation device that applies a force in the first direction to the operating device when a recognition target word that is easily misrecognized with respect to the recognized candidate is selected to the end by the second operation. To do.
(3) The invention of claim 3 is characterized in that, in the speech recognition apparatus of claim 1 or 2, the selected recognition target words are the same before and after the force is applied by the force generator.
(4) The speech recognition device according to any one of claims 1 to 3, wherein the first direction and the second direction are substantially orthogonal to each other.
(5) The invention of claim 5 is the speech recognition apparatus of claim 1, wherein the display device is a recognition target that is likely to be erroneously recognized with respect to a plurality of recognition candidates and a recognition candidate selected from the plurality of recognition candidates. A recognition target word that is easily misrecognized with respect to an unselected recognition candidate among a plurality of recognition candidates is not displayed .
[0007]
【The invention's effect】
The present invention has the following effects.
(1) According to the first to seventh aspects of the present invention, when selecting a recognition candidate using the operating device, if a recognition target word that is likely to be erroneously recognized exists in the recognition candidate, the degree of coincidence of speech The selection from the candidates based on the above and the selection from the candidates easily misrecognized can be performed by independent operations. Therefore, a desired input can be performed with a simple operation.
(2) According to the invention of claim 2, since the operating direction of the operating device matches the moving direction of the selection on the display device, the recognition device can be easily operated.
(3) According to the invention of claim 3, since a force is applied to the operating device, it is possible to easily know that a candidate that is easily misrecognized by an operation in the second direction can be selected.
(4) According to the invention of claim 4, since a force is applied to the operating device, it is easy to know that there are no more recognition target words that are easily misrecognized even if the operation is performed in the second direction. Can do.
(5) According to the invention of claim 5, since the selection of the recognition target word is not changed only by applying force, when the recognition target word selected last is the word to be input, the input is confirmed as it is. Can do.
(6) According to the invention of claim 6, since the direction in which the operating device is operated and the direction in which the force is applied at that time are almost orthogonal to each other, the operator can easily recognize that the force has been applied.
(7) According to the invention of claim 7, since the display of the recognition target words that are easily misrecognized is limited, the display can be simplified.
[0008]
DETAILED DESCRIPTION OF THE INVENTION
<1. Overall configuration>
FIG. 1 is a diagram showing an overall configuration of a speech recognition apparatus according to an embodiment of the present invention.
The speech recognition device according to the first embodiment includes a microphone 101 that is a speech input device, a speaker 102, a signal processing unit 103, an input device 104 that is an operation device, and a display 105 that is a display device. The signal processing unit 103 includes an A / D converter 1031, a D / A converter 1032, an output amplifier 1033, a signal processing device 1034 that is a control device, and an external storage device 1035.
[0009]
The audio input via the microphone 101 is input to the A / D converter 1031 of the signal processing unit 103 as an audio signal. The A / D converter 1031 converts the input audio signal into a digital signal and outputs the digital signal to the signal processing device 1034. The signal processing device 1034 includes a CPU 1034a and a memory 1034b. The signal processing device 1034 calculates the degree of coincidence between the digital data of the recognition target word stored in the external storage device 1035 and the digital data of the input voice. The external storage device 1035 stores a plurality of recognition target words. When these recognition target words include recognition target words that are easily misrecognized, they are stored in association with each other.
[0010]
Selection of words that are easily misrecognized is performed as follows. For example, in the past data such as experiments, when the first recognition candidate having a high degree of coincidence with the voice is “Kanagawa”, the frequency that the actually input confirmed word is “Nagano” is high. To do. In this case, “Nagano” is stored in the external storage device 1035 in association with “Kanagawa” as a recognition target word that is easily misrecognized. Alternatively, “Kanagawa Prefecture” and “Nagano Prefecture” are stored in association with each other as being easily misrecognized.
[0011]
The D / A converter 1032 converts the digital data of the recognition target word into an analog signal and outputs it to the output amplifier 1033. The output amplifier 1033 amplifies the input analog signal, and the speaker 102 outputs this signal as sound.
[0012]
The display 105 is used to display recognition candidates for the input voice. The input device 104 detects an operator's voice recognition start request input, cancels input, recognizes candidate selection operation, and the like, and outputs them to the signal processing device 1034. The input device 104 has a joystick, for example. The joystick has a rotation operation in the direction of the arrow 104A (first direction) in FIG. 1, a rotation operation in the direction of the arrow 104B (second direction), and a push-in operation in the direction of the arrow 104C. Is possible. The rotation operation in the first direction and the second direction is performed to select a recognition target word displayed on the display 105. The pushing operation in the direction of the arrow 104C is performed in order to confirm the recognition target word selected by the turning operation.
[0013]
<2. Configuration of input device>
FIG. 2 is a block diagram illustrating a detailed configuration of the input device 104. The input device 104 includes a joystick 501, joystick drive motors 502 a and 502 b that are force generation devices, joystick position sensors 503 a and 503 b, a joystick control CPU 504, a communication device 505, and other switches 506.
[0014]
The joystick 501 can be operated in the vertical direction as the first direction, in the horizontal direction as the second direction, and pushed in. Here, the vertical direction is the Y-axis direction, and the operation in the Y-axis direction corresponds to selection of an icon in the vertical direction of the screen. The horizontal direction is the X-axis direction, and an operation in the X-axis direction corresponds to selection of an icon in the horizontal direction of the screen.
[0015]
Joystick drive motors 502a and 502b generate torque in the operation direction of the joystick. The joystick position sensors 503a and 503b detect the operation direction of the joystick 501. The joystick control CPU 504 outputs a torque control signal to the joystick drive motors 502a and 502b. The joystick control CPU 504 acquires position information detected by the joystick position sensors 503a and 503b. The communication device 505 outputs joystick position information input from the joystick control CPU 504 to the signal processing device 1034. The communication device 505 outputs the generated torque output from the signal processing device 1034 to the joystick control CPU 504 .
[0016]
FIG. 3 shows a specific structure of the joystick, drive motor, and position sensor of FIG. The joystick operation unit 601 is an operation unit of the joystick 501, and usually only this portion is visible to the user. The joystick rotation unit 602 has a spherical shape and is integrated with the joystick operation unit 601. The joystick operation unit 601 can be operated in a plurality of directions. In this embodiment, as shown in FIG. 1, the operation in only two directions of the X axis and the Y axis is enabled by the cross-shaped groove 104D. Drive motors 605 and 606 are motors for generating torque in two directions. Torques generated by the drive motors 605 and 606 are transmitted to the joystick operation unit 601 through the joystick rotation unit 602 by the torque transmission units 603 and 604, respectively. The drive motors 605 and 606 correspond to the drive motors 502a and 502b in FIG. 2, respectively.
[0017]
An electrical contact 609 is attached to the joystick rotating unit 602. By detecting the contact of the electrical contact with the outer electrical contact 607 or 608, it is determined in which direction the electrical contact is operated. Each outer electrical contact has a pair of outer electrical contacts (not shown) on the opposite side across the joystick rotating portion 602. The outer electrical contacts 608 and 607 correspond to the position sensors 503a and 503b in FIG. 2, respectively. The joystick push-in operation is also detected by an outer electrical contact (not shown) located below the joystick rotating unit 602.
[0018]
<3. Outline of processing>
FIG. 4 is a flowchart showing a procedure of processing performed by the voice recognition apparatus. This control is performed by the signal processing device 1034 of the signal processing unit 103.
[0019]
In step S <b> 201, when the operator operates the input device 104 and a signal for starting voice input is input to the signal processing device 1034, subsequent processing is started. In this case, the input device 104 functions as a speech switch, and the signal processing device 1034 inputs a voice recognition start request signal.
[0020]
In step S 202, a notification sound signal for notifying the operator that voice recognition processing is to be started is read from the external storage device 1035 and output to the D / A converter 1032. The notification sound signal analog-converted by the D / A converter 1032 is output as a notification sound from the speaker 102 via the output amplifier 1033. The operator listens to the notification sound emitted from the speaker 102 and starts voice input to the microphone 101. Here, an example in which the speech recognition apparatus according to the present invention is applied to a car navigation apparatus will be described. That is, the operator inputs the destination by voice. In order to facilitate the explanation, it is assumed here that the name of the destination prefecture is inputted by voice, and the name of the prefecture is stored in the external storage device 1035 as a recognition target word.
[0021]
In the next step S203, input of the input voice is started. The voice uttered by the operator toward the microphone 101 is converted into a digital signal by the A / D converter 1031 and then input to the signal processing device 1034. When power is supplied from a power source (not shown), the microphone 101 picks up surrounding sounds and outputs them to the A / D converter 1031 before the operator operates the input device 104 in step S201. The digital signal converted by the A / D converter 1031 is input to the signal processing device 1034.
[0022]
The signal processing device 1034 calculates the average power of the input digital signal until the operator operates the input device 104 and inputs the voice recognition start request signal in step S201. When the input device 104 is operated to input a voice recognition start request signal and then a voice is input, a digital signal having a power higher than the average power of the calculated digital signal is input. Therefore, the signal processing device 1034 determines that the operator has made a voice input to the microphone 101 when a digital signal having a power greater than a predetermined value from the calculated average power is input, and captures the voice. Start.
[0023]
When the audio capturing is started, the process proceeds to step S204. In step S204, the degree of coincidence between the captured voice and the recognition target word stored in the external storage device 1035 is calculated. The signal processing device 1034 identifies the start of the voice section issued by the operator based on the power of the signal among the digital signals of the voice that has been captured. The degree of similarity (degree of coincidence) between the digital signal after the start of the speech section and the digital signals of the plurality of recognition target words stored in the external storage device 1035 is always calculated and digitized. The larger the value of the degree of coincidence, the more similar the two being compared. Note that, while the matching degree is being calculated by the parallel processing, the voice is continuously captured.
[0024]
If the time during which the power of the digital signal of the voice being captured is less than or equal to a predetermined value continues for a predetermined time or longer, it is determined that the voice input by the operator has been completed, and the voice capture is terminated in step S205.
[0025]
In the next step S206, after completion of the coincidence calculation process, a predetermined number of recognition target words are extracted in descending order of coincidence to be recognition candidates. The predetermined number of recognition target words to be extracted can be determined in advance, for example, 10.
[0026]
When a predetermined number of recognition candidates are extracted, the process proceeds to step S207. In step S207, the operation signal of the input device 104 is received, and a desired recognition target word is selected and determined from the recognition candidates displayed on the display 105.
[0027]
That is, a joystick rotation operation or push-in operation is detected by the joystick position sensors 503a and 503b and the signal is sent to the joystick control CPU 504. Then, the signal is input to the signal processing apparatus 1034 via the communication device 505. When the signal processing device 1034 receives a joystick rotation operation signal, the signal processing device 1034 selects a recognition target word. This selection can be changed by a further turning operation. When a joystick push-in operation signal is received in a state where a desired recognition target word is selected, a recognition target word determination process is executed and this control is terminated.
[0028]
<4. Details of recognition candidate determination process>
FIG. 5 is a flowchart showing details of the processing in step S207 of FIG. First, recognition candidates extracted according to the degree of coincidence of voice are displayed on the display 105 in step S301.
[0029]
FIG. 6 is an example of a screen display of recognition candidates output in step S301 in FIG. The display of the coloring unit 701 indicates a selected state, and no recognition candidate is selected at this time. In FIG. 6, three recognition candidates are displayed in the vertical direction, which is the first direction, in descending order of coincidence. When the predetermined number of extracted recognition candidates is 10, there are further seven recognition candidates whose degree of matching is smaller than that of Saga Prefecture. Note that it is desirable to display the degree of coincidence along with the recognition candidates on the display 105.
[0030]
FIG. 7 shows the correspondence between the recognition candidates and the recognition candidates and the candidates that are easily misrecognized. The recognition candidates are arranged in the vertical direction in the order of acoustic closeness to the speech, and here are the order of Kanagawa Prefecture, Shiga Prefecture, and Saga Prefecture. If there is a candidate that is easily misrecognized with respect to a certain recognition candidate, it is arranged in the horizontal direction next to the recognition candidate. In the example of FIG. 7, there is one word that is easily misrecognized for Kanagawa Prefecture, which is Nagano Prefecture. For example, even if the speaker pronounces “Nagano Prefecture”, the candidate with the best matching voice may be “Kanagawa Prefecture” as a result of noise at the beginning of the word. In this case, “Nagano Prefecture” can be easily selected regardless of the degree of coincidence of speech.
[0031]
When the user operates the joystick downward (direction 610 in FIG. 3) from the state of FIG. 6 (step S302: YES), the electrical contact 609 in FIG. The operation direction is detected by contact. As a result, an attempt to select Kanagawa Prefecture 801 in FIG. 7 is transmitted to the signal processing unit 103. In response to this, the signal processing unit 103 changes the display so that Kanagawa Prefecture is in the selected state as indicated at 901 in FIG. 8 (step S303).
[0032]
When there is a word (Nagano Prefecture) that is easily misrecognized with respect to the selected “Kanagawa Prefecture” (S304: YES), the drive motor 606 generates torque. With this torque, the joystick operation unit 601 in FIG. 3 transmits a force in the right direction (direction 611 in FIG. 3) to the user. However, since the duration of this torque is extremely short, the electrical contact 609 does not newly contact another contact. At the same time, “Nagano Prefecture” 803, which is a candidate for erroneous recognition, is displayed at a position in the horizontal direction that is the second direction with respect to “Kanagawa Prefecture” (step S305). Based on this torque and the display of the misrecognition candidate, the user can know that a candidate that is easily mistaken for Kanagawa can be selected by operating the joystick in the right direction. In addition, after “Kanagawa Prefecture” as a recognition candidate is selected in this way, “Nagano Prefecture” that is easily misrecognized is displayed. Thereby, the display in step S301 can be simplified, and the display process can be speeded up.
[0033]
Here, when the user operates the joystick downward (step S306: “down”), the next recognition candidate (here, Shiga Prefecture) is selected (step S311), and the process returns to step S304. When the joystick is pushed in (step S306: “OK”), the input word is determined as “Kanagawa Prefecture”.
[0034]
When the user operates the joystick to the right, that is, operates the joystick in the direction 611 in FIG. 3 (step S306: “right”), the electrical contact 609 in FIG. 3 contacts the outer electrical contact on the opposite side of the outer electrical contact 608. By doing so, the operation direction is detected. When an operation is detected, the signal is transmitted to the signal processing unit 103. In response to this, the signal processing unit 103 changes the display so that Nagano Prefecture is selected as indicated by 1001 in FIG. 9 (step S307).
[0035]
If there is a word that is easily misrecognized as “Kanagawa prefecture” other than Nagano prefecture (step S308: NO), the process returns to step S306 to wait for the next operation. In this case, in step S306, the joystick may be operated to the right to select the right next erroneous recognition candidate (step S307), or the joystick may be operated to select the next recognition candidate (Shiga Prefecture). It is also possible (step S311), and the joystick may be pushed to determine the currently selected erroneous recognition candidate (Nagano Prefecture).
[0036]
If Nagano Prefecture is the last candidate among words that are easily misrecognized as “Kanagawa Prefecture” (step S308: YES), a signal is transmitted to the drive motor 605 to generate a downward torque indicated by 610 in FIG. To do. As a result, the drive motor 605 generates torque, and the joystick operation unit 601 in FIG. 3 transmits such torque that moves in the 610 direction to the user's hand (step S309). However, since the duration of this torque is extremely short, the electric contact 609 does not newly contact another contact by itself. Based on this torque, the user knows that there are no more candidates for erroneous recognition and that it is better to operate the joystick downward to select an unselected recognition target word. it can. In particular, whether or not there is a word that is likely to be misrecognized after the currently displayed word based on the torque even when many misrecognizable words cannot be displayed side by side in the display area of the display 105 at the same time. Can be easily known.
[0037]
Therefore, when the operator operates the joystick downward, that is, operates the joystick in the direction 610 in FIG. 3 (step S310: “down”), as shown in FIG. 10, the next recognition candidates extracted in order of the degree of coincidence with the speech. “Shiga Prefecture” is selected (step S311), and the process returns to step S304. When “Shiga Prefecture” is selected, the word “Nagano Prefecture”, which is easily misrecognized by the previous recognition candidate “Kanagawa Prefecture”, may be hidden in order to simplify the display. On the other hand, when the joystick is pushed down (step S310: “decision”), the input word is fixed at “Nagano prefecture” currently selected.
[0038]
In step S304, when a word that is easily misrecognized is not registered for the selected recognition candidate, such as “Shiga Prefecture” in FIG. 7 (step S304: NO), the next operation is awaited. If the joystick is pushed in (step S312: “OK”), the selected recognition candidate is confirmed. If the joystick is operated downward (step S312: “down”), the next recognition candidate is selected (step S313), and the process returns to step S304. Note that in step S312, not only “decision” or “down” but an upward operation may be permitted. In that case, the previous recognition candidate is selected, and the process returns to step S304. In step S312, when the operation is performed to the right or left, an error message (not shown) may be output, or the processing may be performed in the same manner as when the operation is performed downward or upward.
[0039]
The selected recognition candidate or misrecognized word is displayed in reverse on the display 105 (FIGS. 8 to 10), and at the same time, the speaker 102 informs the operator with synthesized speech. Thereby, the operator can know exactly what the selected word is.
[0040]
With the above operation, the user can efficiently search for a desired candidate by operations in the vertical direction and the horizontal direction.
[0041]
In the above-described embodiment, the case where there is the first recognition candidate that is easily misrecognized has been described. However, the order of recognition candidates that are easily misrecognized may be any order.
[0042]
In the present embodiment, a joystick has been described as an example of the input device 104. However, any other input device may be used as long as operations in two or more directions are possible. Desirably, if the operation can be performed in two or more directions with a single operation unit, such as a cross key or a trackball, the user can apply torque in the other direction when operating in one direction. I can tell you.
[0043]
In the above-described embodiment, an example in which the speech recognition apparatus according to the present invention is applied to a car navigation apparatus has been described, but the present invention can also be applied to apparatuses other than a car navigation apparatus.
[Brief description of the drawings]
FIG. 1 is a diagram showing an overall configuration of a voice recognition device according to an embodiment of the present invention. FIG. 2 is a block diagram showing a detailed configuration of an input device. FIG. 3 is a specific example of a joystick, a drive motor, and a position sensor in FIG. FIG. 4 is a flowchart showing a procedure of processing performed by the speech recognition apparatus. FIG. 5 is a flowchart showing details of processing in step S207 in FIG. 4. FIG. 6 is output in step S301 in FIG. FIG. 7 is a diagram showing an example of the screen display of recognition candidates. FIG. 7 is a diagram showing the alignment of recognition candidates and the correspondence between recognition candidates and misalignment candidates. FIG. 8 is a recognition output in step S303 of FIG. FIG. 9 is a view showing an example of a candidate screen display. FIG. 9 is a view showing an example of a recognition candidate screen display output in step S307 in FIG. 5. FIG. 10 is an output in step S311 in FIG. Of an example of a screen display of recognition candidates to be displayed 【Explanation of symbols】
101 ... Microphone (voice input device),
102 ... Speaker,
103 ... Signal processing unit,
1031 ... A / D converter,
1032 ... D / A converter,
1033: Output amplifier,
1034 ... Signal processing device (control device),
1034a ... CPU,
1034b ... memory,
1035 ... External storage device,
104 ... input device (operation device),
506 ... switch,
502a, 502b ... drive motor (force generator),
504 ... Joystick control CPU,
503a, 503b ... Joystick position sensors,
505 ... Communication device,
105. Display (display device)

Claims (5)

A voice input device for inputting voice;
A storage device that stores a plurality of recognition target words and stores correspondences between individual recognition target words and other recognition target words that are easily misrecognized.
A candidate extraction device that calculates a degree of coincidence between the input speech and each recognition target word, and sets a plurality of recognition target words as a plurality of recognition candidates based on the degree of coincidence;
A display device that displays the plurality of recognition candidates side by side in a first direction and displays other recognition target words that are easily misrecognized with respect to the recognition candidates in a second direction;
An operating device for performing a first operation corresponding to the first direction and a second operation corresponding to the second direction ;
Said the first operation and selects from the recognition candidates, selecting make a selection from the misrecognized susceptible another recognition target word against recognition candidate that is the selected by said second operating device,
And a force generation device that applies a force to the operation device in the second direction when a recognition candidate having a recognition target word that is likely to be erroneously recognized is selected by the first operation. Voice recognition device. A voice input device for inputting voice;
A storage device that stores a plurality of recognition target words and stores correspondences between individual recognition target words and other recognition target words that are easily misrecognized.
A candidate extraction device that calculates a degree of coincidence between the input speech and each recognition target word, and sets a plurality of recognition target words as a plurality of recognition candidates based on the degree of coincidence;
A display device that displays the plurality of recognition candidates side by side in a first direction and displays other recognition target words that are easily misrecognized with respect to the recognition candidates in a second direction;
An operating device for performing a first operation corresponding to the first direction and a second operation corresponding to the second direction;
A selection device that performs selection from the recognition candidates by the first operation, and performs selection from other recognition target words that are likely to be erroneously recognized by the second operation;
When a recognition target word that is easily misrecognized with respect to the recognition candidate selected by the first operation is selected to the end by the second operation, a force is applied to the operation device in the first direction. A voice recognition device comprising a force generation device. The speech recognition apparatus according to claim 1 or 2,
The speech recognition apparatus according to claim 1, wherein the selected recognition target words are the same before and after the force is applied by the force generation apparatus. The speech recognition apparatus according to any one of claims 1 to 3,
The speech recognition apparatus according to claim 1, wherein the first direction and the second direction are substantially orthogonal to each other. The speech recognition apparatus according to any one of claims 1 to 4,
The display device displays the plurality of recognition candidates and a recognition target word that is easily misrecognized with respect to a recognition candidate selected from the plurality of recognition candidates, and is not selected from the plurality of recognition candidates. A speech recognition apparatus characterized in that recognition target words that are easily misrecognized with respect to recognition candidates are not displayed.

Images