JP3192324B2 - Word speaker for specific speaker - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a word-speech recognition apparatus, and more particularly to a word-speech recognition apparatus for a specific speaker using a specific-speaker voice recognition technique for recognizing a specific speaker's voice.

[0002]

2. Description of the Related Art Conventionally, word-speech recognition devices for specific speakers have
The word to be recognized is registered in advance by the user's own voice, and at the time of recognition, one of the words is selected and uttered, and the feature amount of the uttered word voice and the registered word The feature amount of the voice is compared, and the word that is most similar is selected. Although the speaker recognition for specific speakers requires registration work, it is possible to give a degree of freedom to the recognition vocabulary, and it is advantageous in recognition performance compared to speech recognition for unspecified speakers. have.

[0003] One of the causes of a decrease in speech recognition performance is that a similar word exists in a registered vocabulary, and the similar word is erroneously output as a recognition result. In order to solve such a problem, when registering a person's name by voice in a voice dialing device that generates a name of the other party instead of dialing and makes a call,
Japanese Patent Application Laid-Open No. 3-123249 discloses a method in which, when a similar pattern exists in a registered name, a user is notified of the fact and the similar voice is changed or deleted.

[0004]

The vocabulary of the registered voice is left to the user freely with respect to the name of the other party of the above-mentioned voice dialing device, and therefore, the echo back of the recognition result and the output of the guidance are performed by the user. The synthesis standard pattern created from the feature amount of the uttered voice is used. On the other hand, depending on the use of the recognition device, the recognition vocabulary is fixed, and voices for echo back and guidance are prepared in advance in the memory of the recognition device. If the recognition vocabulary is paraphrased, the voice for the echo back and the guidance will be different from the registered voice of the user, and the device may not recognize at all even if the user intends to pronounce correctly. A similar problem occurs when the recognition result or vocabulary is displayed on a panel or the like.

[0005] For example, in the case of a voice recognition device that needs to register a single-digit number of 0 to 9, a user is required to register “Rei”.
"Ichi", "Ni", "Ku", etc. are requested. In response to this, the user generates a number voice, including “1 (one)” and “7 (schi)”, and “6 (roku)” and “9 (ku)”. Contains words. From the beginning, it is only necessary to select such a confusing word in the recognition vocabulary, but there are cases where similar words are unavoidably included due to human habits. Further, since the utterance method is different for each individual, even if words are not similar to each other for a standard person, there are some people who have similar words. The user changes “7” to “Nana”
It is highly likely that the problem of similar words can be avoided by registering in other words, but the echo back and the guidance voice are “Shi”, so they differ from the registered voice of the user, and the user If the user does not remember what utterance was registered, in other words, the registered vocabulary cannot be recognized. If the guidance voice remains “Shi”, the user is likely to utter “Nana” and forget to register with the passage of time from the registration point, and will likely say “Shi” when recognizing. Because.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem, and a similar word can be registered without discrepancy between the registered voice of the user and the display of the echo back, the guidance voice or the recognition result. It is an object of the present invention to provide a specific-speaker word-speech recognition device that achieves high avoidance recognition performance.

[0007]

SUMMARY OF THE INVENTION According to the present invention, after a user registers in advance his / her own voices of words representing a plurality of objects,
When any of the words registered during use is selected and uttered, the uttered voice pattern is compared with the already registered voice pattern to recognize the uttered word and output the result. A word-for-speech recognition device for a specific speaker, which identifies the plurality of objects by presenting the word to prompt a user to utter a word to be registered; Input means for inputting the voice of the input, a feature value extracting means for analyzing the input voice and extracting the feature value, a storage means for storing the extracted feature value, and a feature value of the voice of the input word. Calculating means for calculating a distance between a speech feature of a registered word, and a distance between the feature of the newly input word and the registered word, Large enough to identify If it is determined that can not be ensured is,
Control means for controlling so that another word prepared as the next candidate is presented to the user one word at a time until a size sufficient to identify the plurality of objects can be secured; The means presents to the user another word prepared as the next candidate of the presented word, or presents another word prepared as the next candidate of the already registered word to the user. or together with the can be selected,
A word prepared as a next candidate of the presented word
The present invention provides a specific-speaker word speech recognition apparatus characterized by prompting the user to utter an arbitrary word when all the words have run out .

The present invention is characterized and output means for outputting the guide voice for prompting a voice input of a word by the user, that the distance between the feature amount is further comprising a notification means for notifying the words less than the predetermined value Word Recognition for Specific Speaker
Provide equipment.

[0009] After the user registers in advance his / her own voices of words representing a plurality of objects and selects and utters one of the words registered at the time of use, the uttered voice pattern and A word recognition apparatus for a specific speaker that identifies a plurality of objects by recognizing a uttered word by comparing with an already registered voice pattern and outputting the result, wherein a user recognizes a voice. An input unit for inputting, a feature amount extracting unit for analyzing the input voice and extracting a feature amount, a storage unit for storing the extracted feature amount,
Calculating means for calculating a distance between the voice feature of the input word and the voice feature of the already registered word; and a feature of the newly input word and the registered word. If it is determined that the distance between the quantities can not secure a sufficient size to identify the plurality of objects, until a sufficient size can be secured to identify the plurality of objects,
And control means for controlling so as to present the other words that are provided as the next candidate in one word by one user, by the user
An output unit that outputs a guide voice prompting a voice input of a word
And a word whose distance between the feature amounts is not more than a predetermined value.
The controlling means, when the word prepared as the next candidate runs out, prompts the user to utter an arbitrary word. provide.

According to the present invention, there is provided a word-for-speech recognition apparatus for a specific speaker, comprising voice output means or display means for outputting what word set is currently registered.
provide.

[0011]

According to the present invention, when the user inputs the voice of the word presented by the input means, the input voice is analyzed by the feature quantity extracting means to extract the feature quantity, and the feature quantity is extracted by the storage means. The feature amount is stored, and a distance between the feature amount of the speech input by the arithmetic means and the feature amount of the already registered word speech is calculated. If the distance is not large enough, the control means may use the other words prepared as next candidates one by one until the user can secure a sufficient size to identify the plurality of objects. Is controlled to be presented. Further, according to the present invention, the control means presents another word prepared as a next candidate of the presented word to the user or is prepared as a next candidate of a registered word. It is possible to select whether to present another word to the user. As a result, it is possible to reliably identify a plurality of targets by using a word having a sufficiently large distance between the feature amounts of speech. Sa
Further, according to the present invention, the unit prepared as the next candidate is
When a word is exhausted, the control means generates an arbitrary word.
Voice is prompted by the user. This allows you to register similar words
Avoid high recognition performance.

Further , according to the present invention, the output means outputs a guide voice prompting the user to input a voice, and the user confirms the guide voice and inputs a voice. Furthermore, a word whose distance between the feature amounts is not more than a predetermined value is notified by the notifying means, and the user can select voice registration of another word. Accordingly, registration of similar words can be avoided and high recognition performance can be realized without any discrepancy between the registered voice of the user and the display of the echo back, the guidance voice, or the recognition result.

Further, according to the present invention, when the user inputs the speech of the word presented by the input means, the input speech is analyzed by the feature quantity extraction means, and the feature quantity is extracted. The extracted feature amount is stored, and a distance between the feature amount of the voice input by the arithmetic unit and the feature amount of the already registered word voice is calculated,
In the case where the distance between the feature amounts cannot be secured to a sufficient size in the words prepared as the initial recognition vocabulary, the control unit sets the next candidate as a next candidate until a sufficient size for identifying the plurality of objects can be secured. Control is performed so that the other prepared words are presented to the user one word at a time. Further
In addition, according to the present invention, the user's voice input by the output means
Is output, and the user receives this guide audio.
Check and input audio. In addition, special
Words whose distance between collections is not more than a predetermined value are reported and used
The user can select voice registration of another word. This
As a result, the registered voice of the user and echo back and guidance
There is no discrepancy in the display of the speech or recognition result
In addition, avoiding registration of similar words and achieving high recognition performance
Can be. Further, according to the present invention, when the word prepared as the next candidate runs out, the control unit prompts the user to utter an arbitrary word. Thereby, registration of similar words is avoided and high recognition performance is realized.

According to the present invention, what kind of word set is currently registered is output by the voice output means or the display means.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of a specific-speaker word speech recognition apparatus according to the present invention will be described below with reference to the drawings.

As shown in FIG. 1, the specific-speaker word-speech recognition apparatus of this embodiment outputs, as shown in FIG. 1, a guide sound for prompting a user to input a voice, and what word set is currently registered. , A microphone 2 as an input unit for inputting a user's voice, and word registration as a feature extraction unit for analyzing the input speech and extracting a feature. / Recognition unit 3
A word recognition / reproduction pattern memory 4 as storage means for storing and reproducing the voice uttered by the user, and a control unit 5 for controlling each unit; And the like, a display panel 7 as display means for outputting what word set is currently registered, and a guide voice memory 8 for storing guide voice.

As shown in FIG. 2, the control unit 5 includes a feature amount of the input voice extracted by the word registration / recognition unit 3 and a feature amount of the word voice already registered in the word recognition / reproduction pattern memory 4. Calculating means 9 for calculating the distance between the two, and a notifying means 10 for notifying a word in which the distance between the feature amounts is not a predetermined value or more, and a word prepared as an initial recognition vocabulary when a user registers a voice. If the distance between the feature values cannot be sufficiently large, the control unit 11 controls to register the voice of another word prepared as the next candidate.
Is provided.

As an application example of the specific-speaker word speech recognition apparatus, for example, in home automation,
A device for operating the indoor lighting and the tape recorder by voice is considered, and a vocabulary as shown in FIG. 3 is assumed as a recognition vocabulary. As can be seen from FIG. 3, several candidate words are prepared in advance for one operation.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG. In addition, since the operation of the specific-speaker word speech recognition device is performed using the microphone 2 and the operation panel 6, in order to use the specific-speaker word speech recognition device, words are first spoken by the voice of the user. You need to register.

When the operation panel 6 is operated to start registration (step S1), the voice synthesizer 1 utters a guide voice "Please say the word to be said" stored in the guide voice memory 8. After that (step S2), for example, when the recognized vocabulary is as shown in FIG. 3, a guide voice "power" is played first.
Then, it is determined whether or not there is a next candidate selection operation, that is, whether or not an operation has been performed from the operation panel 6 (step S3). If you do not perform the next candidate selection operation,
The user listens to the guide voice “Power” and “Power”
(Step S4). The voice registered by the user is analyzed by the word registration / recognition unit 3, and the feature amount is extracted. Then, the distance between each word already registered in the word recognition / reproduction pattern memory 4 and the feature amount of the voice pattern is calculated by the calculating means 9 using a technique such as DP matching (step S5).

The control means 11 determines whether or not a combination in which the distance is equal to or less than the set threshold value, that is, a similar pattern exists.
(Step S6), and when there is no similar pattern, the feature amount of the voice "power" just uttered by the user is stored in the memory 4 as a storage pattern (step S7). Further, when it is determined that a similar pattern exists, for example, after the voice of “power” is registered, when the registration of the next vocabulary “light” is advanced from step S1, the “power” and “light” Are determined to be similar words when the inter-pattern distance becomes equal to or less than the set threshold. Then, a warning such as "The power supply and the light are similar. Please register either one again." Is issued to the user through the voice synthesizing unit 1 by the notification means 10 (step S8). .

Then, when the user receives such a warning, it is determined whether the user has selected any of re-registering, re-registering, re-registering, or re-registering similar words (step S9). If the registration operation is performed ignoring the warning, the process moves to step S7, the voice of the "light" just uttered is stored in the memory 4, and the process can proceed to the registration of the next word. However, when the user wishes to improve the recognition performance and re-registers a word for which a warning has been received, for example, when re-registering “light”, the user performs a registration cancel operation from the operation panel 6. After the registration is canceled (step S10), the process returns to step S2 again, and a guidance prompting utterance flows. As shown in FIG. 3, “lighting” prepared as the next candidate of “electric light” is shown in FIG. Is output. Here, if "utterance" is uttered, the distance between each word already registered in step S5 and "lighting" is calculated in the same manner as when the "light" is uttered. If there is, step S8
Will give a warning. If there is no combination of similar words, or if "illumination" is registered ignoring the warning, "illumination" is used instead of "electric light" for echo back of the recognition result. This is made possible by storing the number of candidate voices registered together with the voice feature amount. “1” is stored together with the feature amount of the voice when registered with “lighting” and “2” when registered with “lighting”. As shown in FIG. 5, the storage format includes a registration data start address, a data length, and a candidate number.

In step S3, when an operation is performed from the operation panel 6, for example, a guide voice "illumination" is issued, and if the user does not want to register by uttering "illumination", the voice is output. Operation panel 6 without speaking
Is operated, "light" is prepared as the next candidate word of "lighting" as shown in FIG. 3 (step S1).
1) A guide voice “light” is output (step S12), and the process returns to step S3. Step S3
Needless to say, the transition from to the step S11 is possible even if the registration is not redone. Step S
Returning to 3, if the user does not want to register even "light", the same operation as described above is performed from the operation panel 6, and since there is no next candidate for "light", the operation returns to the first candidate "electric light". That is, each time the process passes from step S11 to step S12, the candidate word is shifted.

In step S11, if there is no guide voice for the next candidate word for registration, for example, since there is no next candidate for "light", guidance is given such as "enter by arbitrary utterance" (step S13). ), Avoiding registration of similar words by making the user utter any word. Here, when a voice is registered, the voice is used not only for recognition but also for voice output, and thereafter, the voice uttered by the user is used for the echo back of the recognition result, the guide voice, and the like.

When an arbitrary utterance of the user is registered, "0" or the like is recorded in the candidate number column of FIG. 5 to distinguish it from the voice prepared in advance. The echo back of the registered voice of the user uses the sampling data at the time of registration as word recognition and
This can be achieved by storing the data in the reproduction pattern memory 4 and performing D / A conversion. However, in order to save the memory for echo back, there is also a method of synthesizing from the feature amounts of words registered for recognition. .

On the other hand, "Please input by arbitrary utterance"
The guide voice such as "power", "electric light", "light", and the like for the registration words shown in FIG. 3 do not need to be rewritten, and can be realized by a ROM that is cheaper than a RAM. The guide voice memory 8 in FIG. 1 stores these guide voice data, and the voice synthesizing unit 1 receives a command from the control unit 5, reads out the data from the guide voice memory 8, and reproduces the guide voice. At the point in time when an arbitrary utterance is requested, if no further utterance is selected and the next candidate is selected, the process returns to the first candidate and a guide sound of "light" is output again.

By the way, in the above step S8,
If a warning "Power supply and light are similar. Please register either one again" is received, step S9
When re-registration of a registered similar word is selected, the voice pattern of "electric light" is stored in the memory 4, and "power" is registered again (step S14). In this example, the operation panel 6 is operated in step S9, and the registration of “power” is accepted, and the next candidate word of “power” is selected and uttered in step S3. As a result of re-registering, the distance to another word already registered may be short. Of course, even in this case, a warning guidance is played, so that the operation panel 6 is operated and the corresponding word voice can be registered again.

As a registration algorithm, not only a method of issuing a warning when one word is registered as described above, but also a distance between words becomes equal to or less than a set threshold when all words are registered. It is also conceivable to check whether there is a combination and warn the corresponding one. Regardless of the registration algorithm, if there are multiple combinations of words whose distance is equal to or less than the set threshold,
All of them may be notified, or only the combination with the smallest distance may be notified. In addition, it seems that it is necessary to provide a function for confirming a combination of words having an insufficient distance at any time by operating the operation panel 6, but this can be achieved only by using the distance check function at the time of registration.

The registration and recognition of words is performed by the word registration / recognition unit 3
Receives an instruction from the control unit 5, analyzes the input from the microphone 2, and stores the characteristic amount in the word recognition / reproduction pattern memory 4 or compares it with the characteristic amount of another word already stored. This is done by: The control unit 5 issues a command to the speech synthesis unit 1 and the word registration / recognition unit 3 according to the input from the operation panel 6 and the state of the device at that time. The display panel 7 is required when the recognition vocabulary is confirmed by character display as described later. When the recognition vocabulary is confirmed by voice, the display panel 7 is not required. In the above embodiment, the recognition vocabulary displayed in FIG. 3 has been described as an example, but the present invention is not limited to the types of recognition vocabulary.

In the case of the recognition device having the above-described configuration, since the recognition vocabulary is not uniquely fixed, a method for confirming what vocabulary is registered at the present time is required. First, there is a method of confirming by voice.
When a specific operation is performed from the operation panel 6 of the recognition device shown in FIG. 1, sounds of registered words are sequentially output. As the confirmation voice, there are a method of outputting what is prepared as a guide voice for registration and a method of synthesizing and outputting the registered voice of the user. Although the use of the guide voice can output a clearer voice, the guide voice is not prepared for words registered by arbitrary utterances, so that a method of synthesizing and outputting the registered voice of the user is used.

As a second method, a method of using a display panel 7 such as a liquid crystal to display registered words in characters can be considered. A speech recognition apparatus provided with a recognition vocabulary confirmation unit using a display panel is configured as shown in FIG. If the total number of recognized words is small or the display panel is large enough, it is possible to display all words on one side,
If not, a function of operating the operation panel 31 to switch the screen of the recognized vocabulary display unit 32 or scrolling the screen is required. Since a character string is not prepared for a word in which the user has registered an arbitrary voice, a method of inputting a character string from the character input unit 33 of the operation panel at the time of voice registration may be considered. For example,
“Stop”, which is the seventh recognized vocabulary in the recognized vocabulary display unit 32 in FIG. 6, is not prepared in the word candidates in FIG. This is a word that the user thought and registered himself, because “stop” or “stop” cannot secure a sufficient distance from other recognized vocabulary. Since the character string “STOP” is not prepared, the user operates the character input unit 33 to input the character string.

The character string data is prepared in advance in the guide voice memory 8 for the word candidates in FIG. 3, and is stored in the word recognition / reproduction pattern memory 4 for the data arbitrarily input by the user. You can use the method of remembering. In addition,
Reference numeral 34 in the figure denotes a microphone.

[0033]

According to the present invention, when the user inputs the voice of the word presented by the input means, the input voice is analyzed by the feature extracting means, and the feature is extracted.
The storage means stores the extracted feature quantity, and calculates the distance between the speech feature quantity input by the calculation means and the feature quantity of the already registered word speech, and prepares the distance as an initial recognition vocabulary. If the distance between the feature values cannot be sufficiently large for the word that is present, another word prepared as the next candidate is kept until the control means can secure a sufficient size for identifying the plurality of targets. Is presented to the user one word at a time, and the control means presents to the user another word prepared as the next candidate of the presented word, or the registered word. The user can select whether to present another word prepared as the next candidate to the user, so that a plurality of objects can be reliably identified by a word having a sufficiently large distance between the speech feature amounts. This Can. In addition, as the next candidate
When the intended word is exhausted, the control means
It was designed to encourage users to utter arbitrary words
Thus, high recognition performance is realized by avoiding registration of similar words.

Further, according to the present invention, the output means outputs a guide voice prompting the user to input a voice, and the user confirms the guide voice and inputs the voice. In addition, the notification unit notifies the user of a word whose distance between the feature amounts is not more than a predetermined value, and the user can select voice registration of another word. It is possible to avoid registration of similar words and achieve high recognition performance without causing any discrepancies in the display of the recognition voice, the guidance voice, or the recognition result .

Further, according to the present invention, when the user inputs the voice of the word presented by the input means, the input voice is analyzed by the feature quantity extracting means, and the feature quantity is extracted. The extracted feature amount is stored, and a distance between the feature amount of the voice input by the arithmetic unit and the feature amount of the already registered word voice is calculated,
In the case where the distance between the feature amounts cannot be secured to a sufficient size in the words prepared as the initial recognition vocabulary, the control unit sets the next candidate as a next candidate until a sufficient size for identifying the plurality of objects can be secured. Control is performed so that the other prepared words are presented to the user one word at a time. Further
In addition, a guide voice that prompts the user to input voice by output means
Is output, and the user confirms this guide sound and outputs the sound.
input. In addition, the distance between the feature values is
Words that are not equal to or greater than the fixed value are notified, and the user can
Since we have configured so that you can select voice registration,
There is a registered voice of the user and an echo back or guidance voice
Display of the recognition result, etc.
High recognition performance can be realized by avoiding word registration.
Furthermore, when the word prepared as the next candidate runs out, the control means is configured to prompt the user to utter an arbitrary word, thereby avoiding registration of a similar word and realizing high recognition performance.

Further , according to the present invention, the speech output means or the display means is configured to output what word set is currently registered. Therefore, what kind of vocabulary is currently registered is determined. Can be easily confirmed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of a specific-speaker word speech recognition apparatus according to the present invention.

FIG. 2 is a block diagram showing a configuration of a control unit of the specific-speaker word speech recognition apparatus of the present invention.

FIG. 3 is a diagram showing recognized vocabulary candidates.

FIG. 4 is a flowchart showing the operation of the present invention.

FIG. 5 is a diagram showing a storage format of registered voice information.

FIG. 6 is a diagram showing a recognized vocabulary checking device using a display panel.

[Explanation of symbols]

 1 Voice synthesis unit 2 Microphone 3 Word registration / recognition unit 4 Word recognition / playback pattern memory 5 Control unit 6 Operation panel 7 Display panel 8 Guide voice memory 9 Computation unit 10 Notification unit 11 Control unit 31 Operation panel 32 Recognized vocabulary display unit 33 Character input section 34 Microphone

Continuation of the front page (56) References JP-A-57-129497 (JP, A) JP-A-58-76944 (JP, A) JP-A-60-218698 (JP, A) JP-A-3-123249 (JP) JP-A-63-70296 (JP, A) JP-A-62-260194 (JP, A) JP-A-1-285997 (JP, A) JP-A-63-38994 (JP, A) JP-A-63-294600 (JP, A) JP-A-3-10298 (JP, A) JP-A-2-23399 (JP, A) JP-A-58-178396 (JP, A) A) JP-A-2-141154 (JP, A) JP-B 5-74837 (JP, B2) JP-B 61-26677 (JP, B2) JP-B 2-15080 (JP, B2) JP-B 5 -38958 (JP, B2) JP 6-7347 (JP, B2) JP 4-62595 (JP, B2) (58) Fields studied (Int. Cl. ⁷ , DB name) G10L 15/06 G10L 15/28 JICST file (JOIS)

Claims (4)

(57) [Claims] When a user registers in advance his / her voice of a word representing a plurality of objects with his / her own voice and selects and utters one of the words registered at the time of use, the uttered voice A word recognition apparatus for a specific speaker, which identifies a plurality of objects by recognizing a spoken word by comparing a pattern with an already registered voice pattern and outputting a result thereof, Means for presenting the word to prompt the utterance of the word to be registered; input means for the user to input the voice of the presented word; and feature quantity for analyzing the input voice and extracting the feature quantity Extracting means; storing means for storing the extracted feature quantity; calculating means for calculating a distance between the speech feature quantity of the input word and the speech feature quantity of the already registered word; Newly entered word and registration If it is determined that the distance between the feature amount and the recorded word cannot be large enough to identify the plurality of objects, a sufficient amount is secured to identify the plurality of objects. And control means for presenting another word prepared as a next candidate to the user one word at a time, wherein said control means is prepared as a next candidate of said presented word. or presents other words are in user, with is already possible choose to present the other words that are provided as the next candidate of the words registered in the user, the
The word prepared as the next candidate of the presented word is exhausted.
A word-recognition apparatus for a specific speaker, which prompts a user to utter an arbitrary word when the word comes . 2. The apparatus according to claim 1, further comprising: an output unit that outputs a guide voice that prompts a user to input a word by voice; and a notifying unit that notifies a word that the distance between the feature amounts is not more than a predetermined value. A word-speech recognition device for specific speakers. 3. A voice uttered when a user pre-registers voices of words representing a plurality of objects with his / her own voice and selects and utters any of the words registered at the time of use. A word recognition apparatus for a specific speaker that identifies a plurality of objects by recognizing a spoken word by comparing a pattern with an already registered speech pattern and outputting a result, wherein a user An input unit for inputting a voice, a feature amount extracting unit for analyzing the input voice and extracting a feature amount, a storage unit for storing the extracted feature amount, a feature amount of a voice of the input word, Calculating means for calculating a distance between a speech feature of a word already registered; and a distance between a feature of the newly input word and the registered word identifying the plurality of targets. Be large enough to If it is determined that it can not, until enough large to identify the plurality of object, and control means for controlling so as to present the other words that are provided as the next candidate in one word by one user Output a guide voice prompting the user to input words
Output means, and a word whose distance between the feature amounts is not more than a predetermined value.
And a notifying unit, wherein the control unit prompts the user to utter an arbitrary word when the word prepared as the next candidate runs out. 4. The word recognition apparatus for a specific speaker according to claim 1, further comprising a voice output unit or a display unit for outputting what word set is currently registered. .