JPH11352985A - Voice recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent an unnecessary language element code outputted by voice recognition from being transferred to an application program. SOLUTION: This voice recognition device comprise a voice input 11 for inputting the voice of a speaker; a voice recognition source 12 for storing a plurality of words and phrases to be preliminarily recognized and a language element code corresponding to each word and phase, a voice recognition part 13 for recognizing the words and phrases from the voice inputted from the voice input 11 and extracting and outputting the language element code corresponding to each word and phrase from the recognition source 12 when the recognized words and phrases include the word and phrase to be preliminarily recognized, and a filter part 14 for removing all the language element codes when a necessary language element code is not included in a plurality of language element codes outputted from the voice recognition part 13 by one voice recognition.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech recognition apparatus for recognizing and coding words and phrases from input speech.

[0002]

2. Description of the Related Art As shown in FIG. 12, a conventional voice recognition apparatus has a voice input unit 1 having a microphone 1a for inputting voice and an A / D converter 1b for converting voice from the microphone into a digital signal. A speech recognition resource 2, which is a collection of language element codes defined for the phrase to be recognized;
Recognize words and phrases based on the output from the voice input unit 1,
The speech recognition unit 3 extracts a language element code corresponding to the phrase based on the speech recognition resource 2 and an application program 4 that uses a code string of the language element code extracted by the speech recognition unit 3 as speech recognition data. .

In such a device, when a speaker speaks toward the microphone 1a, the speech is converted into a digital signal by the speech input unit 1 and supplied to the speech recognition unit 3. The speech recognition unit 3 recognizes words and phrases defined in advance in the order in which they were uttered, and passes a code string of language element codes to the application program 4 as speech recognition data in that order.

[0004]

However, when such a voice recognition device performs voice recognition from, for example, a conversation between a customer and a clerk, the application program 4 does not recognize unnecessary words, such as greetings, small talk, and noise. Microphone 1
Since it is taken as a voice from "a", there is a possibility that it may be erroneously recognized as a necessary phrase, and the result may be passed to the application program 4. This causes a malfunction of the application program 4.

[0005] For this reason, the above-described speech recognition apparatus cannot be used as it is in a case where only necessary words and phrases are recognized in a conversation between speakers.

Accordingly, an object of the present invention is to provide a speech recognition apparatus which can prevent an unnecessary language element code output by speech recognition from being passed to an application program.

[0007]

According to a first aspect of the present invention, there is provided a voice input means for inputting a voice of a speaker, a voice recognition means for recognizing a phrase from a voice input from the voice input means, and A language element code storage unit for storing a plurality of words to be recognized and a language element code corresponding to each word, and a word corresponding to each word when the word recognized by the voice recognition unit includes a word to be recognized in advance. A language element code output means for extracting and outputting a language element code from the language element code storage means, and removing unnecessary language element codes from the language element code output from the language element code output means and passing necessary ones A speech recognition device comprising a filter unit.

According to a second aspect of the present invention, there is provided a voice input means for inputting a voice of a speaker, a voice recognition means for recognizing a phrase from the voice input from the voice input means, and a plurality of voice recognition means to be recognized in advance. A language element code storage unit that stores a word and a language element code corresponding to each phrase, and when the phrase recognized by the speech recognition unit includes a phrase to be recognized in advance,
Language element code output means for extracting and outputting a language element code corresponding to each phrase from the language element code storage means,
If one of the plurality of language element codes output from the language element code output means in one speech recognition process does not include any necessary language element code, all language element codes are removed, and When all the language element codes are available, the speech recognition apparatus is provided with filter means for passing the language element codes.

According to a third aspect of the present invention, there is provided a voice input means for inputting a voice of a speaker, a voice recognition means for recognizing a phrase from a voice input from the voice input means, and a plurality of voice recognition means to be recognized in advance. A language element code storing means for storing a language element code within a predetermined numerical range corresponding to each of the words and phrases; and, when the words recognized by the speech recognition means include words to be recognized in advance, the words correspond to the respective words. A language element code output means for extracting and outputting a language element code from the language element code storage means, and a language element if the language element code output from the language element code output means is out of a predetermined numerical range. And a filter means for removing the code and passing the language element code when the code is within a predetermined numerical range.

According to a fourth aspect of the present invention, there is provided a voice input means for inputting a voice of a speaker, a voice recognition means for recognizing a phrase from the voice input from the voice input means, and a voice as a deposit from a customer. Language element code storage means for storing words indicating a plurality of amounts to be recognized by the recognizing means and language element codes indicating the amounts corresponding to the respective words, and words to be recognized in advance by the speech recognition means When including
Language element code output means for extracting and outputting a language element code corresponding to each phrase from the language element code storage means,
When the price data is fetched from outside and the amount indicated by the language element code output from the language element code output means is lower than the price of the data fetched from outside, the language element code is removed and fetched from outside Filter means for passing the language element code when the price is equal to or more than the price of the data.

According to a fifth aspect of the present invention, there is provided a voice input unit for inputting a voice of a speaker, a voice recognition unit for recognizing a phrase from the voice input from the voice input unit, and a voice as a deposit from a customer. Language element code storage means for storing words indicating a plurality of amounts to be recognized by the recognizing means and language element codes indicating the amounts corresponding to the respective words, and words to be recognized in advance by the speech recognition means When including
Language element code output means for extracting and outputting a language element code corresponding to each phrase from the language element code storage means,
The price data is taken in from the outside, and if the amount indicated by the language element code output from the language element code output means does not satisfy the conditions derived based on the price of the data taken in from the outside, the language element code is removed. And a filter means for passing the language element code when the condition is satisfied.

According to a sixth aspect of the present invention, there is provided a voice input means for inputting a voice of a speaker, a voice recognition means for recognizing a phrase from the voice input from the voice input means, and a plurality of voice recognition means to be recognized in advance. Language element code storage means for storing words and language element codes corresponding to the respective words; product information storage means configured by collecting product information associated with the language element codes; and words recognized by the speech recognition means. A language element code output means for extracting and outputting a language element code corresponding to each word from the language element code storage means when including a word to be recognized in advance; and a language element code output from the language element code output means The product information is retrieved from the product information storage means using the key as a key. If there is no corresponding product information, the language element code is removed. A speech recognition apparatus characterized by comprising a filter means for passing the language elements code.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a first embodiment in which the present invention is applied to a business processing device such as an electronic cash register, a POS terminal or the like which performs merchandise sales code registration processing will be described with reference to FIGS. It will be described with reference to FIG.

FIG. 1 is a functional block diagram showing the configuration of a business processing device according to the present embodiment. This business processing device includes a microphone 11a for inputting audio as an analog signal.
And an audio input unit (audio input means) 11 including an A / D converter 11b for converting audio from the microphone 11a into a digital signal, a language element defined (associated) with a phrase to be recognized in advance. A speech recognition resource (language element code storage unit) 12 as a language element code storage unit, which is a set of codes, and a phrase corresponding to the input speech is recognized based on an output from the voice input unit 11 (speech recognition unit). ), A speech recognition unit 13 that extracts a language element code corresponding to the phrase from the speech recognition resource 12 and outputs (language element code output means). Among the language element codes extracted by the speech recognition unit 13, A filter unit (filter means) 14 for outputting only necessary language element codes based on the language element codes from the filter unit 14; It consists of Deployment program 15.

The speech recognition resource 12 is constituted by a storage device such as a hard disk device. Specifically, it is composed of words to be recognized as shown in FIG. 2 and language element codes corresponding to the words and phrases. The language element codes are all four-digit codes. Words to be voice-recognized are classified into "product names" and "quantities", and the codes are regularly associated with each classification. Here, as a predetermined rule, a case where the “product name” is expressed by the lower two digits and the upper two digits are set to 01, and the “number” is expressed by the lower two digits and all the upper two digits are set to 00 This is given as an example. For example, the language element code “0101” is associated with the phrase “product A”, and “1”
A language element code of “0001” is associated with the phrase “individual”.

Although not shown, the words to be speech-recognized are stored in advance in association with the speech characteristic data of the standard speaker (unspecified speaker-compatible type). However, the voice feature data actually uttered by the user may be associated (specific speaker correspondence type).

The voice recognition unit 13 and the filter unit 1
4 and the application program 15 are CPU
(Central processing unit) · ROM (read only memory)
-It is composed of a personal computer having a RAM (random access memory). The voice recognition unit 13, the filter unit 14, and the application program 15 are specifically configured by a software program stored in a storage device such as a hard disk device or a memory such as a ROM and readable by the CPU of the personal computer. Is done.

The speech recognition unit 13 defines (corresponds to) speech input based on the output from the speech input unit 11 and speech feature data in the speech recognition resource 12 in advance.
Detecting similarity / approximation with the word (for example, making the speech recognition resource 12 correspond to the same word with speech feature data of a plurality of types of phrases meaning the same word, and input based on this. The speech recognition is performed to identify the uttered phrase), and the speech recognition is performed. The language element code corresponding to the phrase obtained by the speech recognition is stored in the speech recognition resource 12.
Extract from and output.

The filter unit 14 removes unnecessary language element codes from the language element codes extracted by the speech recognition unit 13 based on a predetermined rule, and sends only necessary language element codes to the application program 15. To pass.

In the present embodiment, it is normal for the speech recognition unit 13 to output one language element code of "product name" and one "number" in one speech recognition process. When the language element code is output or when the language element code is insufficient, it is necessary not to output the language element code. That is, in this case, if only one of the “product name” and “number” language element codes is output from the voice recognition unit 13 in one voice recognition process, the language element code is output from the application program 15. Is unnecessary code, it is necessary to remove it.

Therefore, if one language element code having the number of digits assigned to each classification does not exist in one speech recognition process, the filter unit 14 determines that the required combination of language element codes is not enough. Then, the language element code removal processing is performed based on the rule of removing the language element code.

For example, in the present embodiment, 100 or more language element codes are assigned to “product name”, and 1 is assigned to “number”.
Since a language element code of less than 00 is assigned, as a combination of a plurality of language element codes from the speech recognition unit 13,
Unless there is one language element code of 100 or more and one language element code of less than 100, the language element code is removed. Thus, unnecessary language element codes are not passed to the application program 15.

The application program 15 is:
It is composed of a software program that performs predetermined business processing such as registration of a product sales code and calculation of a price based on the language element code that has passed through the filter unit 14. When the product name and the number are uttered, a language element code is created and output through the filter unit 14. The application program 15 selects and displays the corresponding product name on a screen such as a display using the language element code. , And to execute subsequent accounting processes such as registration of a product code and calculation of a price.

In the embodiment of the present invention having such a configuration, for example, in one speech recognition process, the user of the present apparatus utters "3 products A" as shown in FIG. Then, the voice is converted into a digital signal by the voice input unit 11 and supplied to the voice recognition unit 13. Then, the speech recognition unit 13 refers to the speech recognition resource 12 to detect the similarity / approximation between the input speech and a phrase defined in advance in the speech recognition resource. The language element code “0101” is output, and “3”
Is output as a language element code "0003". These language element codes are passed to the next filter unit 14. Then, the filter unit 14 determines that the speech recognition has been normally performed because there is one language element code of 100 or more and one language element code of less than 100, and these language element codes are passed to the application program 15.

On the other hand, in one speech recognition process, the user of the present apparatus uses "Product A" as shown in FIG.
Is XX (where XX is a word irrelevant to the “number”), this voice is converted into a digital signal by the voice input unit 11 and supplied to the voice recognition unit 13. Then, the speech recognition unit 13 refers to the speech recognition resource 12 to detect the similarity / approximation between the inputted speech and a phrase defined in advance in the speech recognition resource. 01
A language element code of “01” is output. Since “XX” is a word that is not defined in the speech recognition resource 12, no language element code is output.
Then, the filter unit 14 determines that the speech recognition was not performed normally because one language element code less than 100 is insufficient, and removes the language element code. Therefore, in this case, the language element code is not passed to the application program 15.

As described above, only when there is one language element code from the speech recognition unit 13 for each of "commodity" and "quantity", the filter unit 14 passes those language element codes to the application program 15. Therefore, even if speech recognition is performed from a conversation between a customer and a clerk, unnecessary words and phrases such as greetings, small talk, and noise are not recognized by the application program 15 even if an unnecessary language element code is erroneously recognized. Even if output from the recognition unit 13, it is removed by the filter unit 14 and is not passed to the application program 15, so that malfunction of the application program 15 can be prevented. Therefore, it is possible to provide a speech recognition device that can accurately code speech even when used in a case where only necessary words and phrases are recognized from a speaker's conversation.

The speech recognition unit 13 selects the closest phrase from the predefined words to be recognized, but the filter unit 14 can remove unnecessary language element codes. Therefore, it is possible to prevent the application program 15 from malfunctioning without defining all of the user's characteristic wording of the phrase to be speech-recognized in the speech recognition resource 12. In the following embodiments, unnecessary language element codes are also added to the filter unit 1.
4, the effect can be obtained.

Next, a second embodiment in which the present invention is applied to a business processing device such as an electronic cash register, a POS terminal or the like which performs a merchandise sales code registration process will be described with reference to FIG. The functional block diagram of the business processing device and the configuration diagram of the speech recognition resource 12 according to the present embodiment are the same as those shown in FIGS. 1 and 2, respectively, and therefore, detailed description thereof will be omitted.

The filter section 14 in the present embodiment is
When the language element code from the voice recognition unit 13 is out of the numerical range assigned in advance, the language element code is removed.
And different.

That is, in the speech recognition resource 12,
The numerical range of the language element code to be used is predetermined (for example, in the case of the one shown in FIG. 2, the numerical range of “product” is “0101” to “0103”,
The numerical range for “number” is “0001” to “000”.
3 "), when such a language element code out of the numerical range is output due to an error of the voice recognition unit 13 or the like, the unnecessary language element code is removed so as not to be passed to the application program 15. Is what you do.

In the embodiment of the present invention having such a configuration, for example, when the user of the apparatus utters "3 products A" as shown in FIG. At 11, it is converted into a digital signal and supplied to the voice recognition unit 13. Then, the speech recognition unit 13 refers to the speech recognition resource 12, and detects the similarity / approximation between the input speech and a phrase defined in advance in the speech recognition resource,
A language element code “0101” is output for “product A”, and a language element code “0003” is output for “3”. These language element codes are passed to the next filter unit 14. Then, the filter unit 14
In these cases, it is determined that these language element codes are in a numerical range assigned in advance, and the application program 1
Handed over to 5.

On the other hand, the user of this apparatus
As shown in (b), when "xx is three" (xx is "product" not registered in the voice recognition resource), this voice is converted into a digital signal by the voice input unit 11 and voice recognition is performed. It is supplied to the unit 13. Then, the speech recognition unit 13 refers to the speech recognition resource 12 to detect the similarity / approximation between the input speech and a phrase defined in advance in the speech recognition resource. 0003 "is output, and" xx "is a word that is not defined in the speech recognition resource 12, so that" 010 "is undefined due to erroneous recognition by the speech recognition unit 13, for example.
Assuming that the language element code “5” is output, the filter unit 14 outputs “0” outside the numerical range that is assigned in advance.
The language element code of “105” is removed, and only the language element code of “0003” within the numerical range that is assigned in advance is passed to the application program 15.

As described above, only when there is one language element code from the speech recognition unit 13 for "product" and "quantity", the filter unit 14 passes those language element codes to the application program 15. Therefore, even if speech recognition is performed from a conversation between a customer and a clerk, unnecessary words and phrases such as greetings, small talk, and noise are not recognized by the application program 4 even if they are erroneously recognized due to an error. Even if the code is output from the speech recognition unit 13, the code is removed by the filter unit 14 and is not passed to the application program 15, so that the malfunction of the application program 15 can be prevented. Therefore, similarly to the first embodiment, a speech recognition apparatus capable of accurately encoding speech even when used for speech recognition of only a necessary phrase from a speaker's conversation is provided. be able to.

Next, a third embodiment in which the present invention is applied to a business processing apparatus for processing a deposit from a customer using an electronic cash register, a POS terminal, or the like will be described with reference to FIGS. This will be described with reference to FIG. In addition, the first
The same reference numerals are given to the same portions as those of the embodiment, and the detailed description is omitted.

As shown in FIG. 5, the business processing device according to the present embodiment includes a voice input unit 11 and a voice recognition resource 1.
2 ', a speech recognition unit 13, a filter unit 14, and an application program 15 using language element codes passed through the filter unit 14. The difference from the one shown in FIG. The point is that data of the total price (price) of the product can be received.

Further, the speech recognition resource 12 'in the present embodiment differs from the first embodiment in that the speech recognition resource 12' is constituted by money data as shown in FIG. For example, a language element code of “1000” is associated with a phrase of “1,000 yen”, and a language element code of “3000” is associated with a word of “3,000 yen” and stored.

By the way, in a business processing device for processing a deposit from a customer, the application program 15
It has data on the total amount of goods already sold, and it is not usually conceivable to deposit less than this total amount from customers. Therefore, when the language element code of the deposit less than the total price of the product is output from the voice recognition unit 13, it is determined that the language element code is erroneously recognized, and the unnecessary language element code is removed by removing the language element code. To the application program 15 can be prevented.

Based on such a principle, the filter unit 14 in the present embodiment compares the total price of the product received from the application program 15 with the language element code, and when the language element code is less than the total price of the product, , The language element code is removed so as not to be passed to the application program 15.

In the embodiment of the present invention having such a configuration, for example, when the user of the present apparatus utters "3,000 yen" as shown in FIG.
Is converted into a digital signal and supplied to the voice recognition unit 13. Then, the speech recognition unit 13 outputs a speech recognition resource 12 ′.
Is detected, the similarity / approximation between the input speech and the phrase defined in advance in the speech recognition resource is detected, and a language element code of “3000” is output for “3,000 yen”.

On the other hand, in the filter unit 14, the total price of the product of “2500” yen from the application program 15 and the language element code “300” from the speech recognition unit 13
0 "is compared. In this case, since the language element code is equal to or more than the total price of the product, it is determined that the speech has been normally recognized, and the language element code is passed to the application program 15.

On the other hand, for example, when the user of the present apparatus utters “XXX” as shown in FIG. 7B, this voice is converted into a digital signal by the voice input unit 11 and 13 is supplied. Then, the speech recognition unit 13 refers to the speech recognition resource 12 ′, detects the similarity / approximation between the input speech and the phrase defined in advance in the speech recognition resource, and It is assumed that a language element code “2000” is output due to erroneous recognition.

On the other hand, in the filter unit 14, the total price of the product of “2500” yen from the application program 15 and the language element code “200” from the voice recognition unit 13
0 "is compared. In this case, since the language element code is less than the total price of the product, it is determined that speech recognition has not been normally performed, and the language element code is not passed to the application program 15.

As described above, the language element code from the voice recognition unit 13 is compared with the total price of the product from the application program 15, and only when the language element code is not less than the total price, the language element code is compared with the application program. Since the filter unit 14 is provided so as to be passed to a customer, even if speech recognition is performed from a conversation between a customer and a clerk,
Unnecessary words in the application program 15, for example, voices of deposit amounts, greetings, small talks, noises, etc., which are not normally considered from the viewpoint of the total price of products, are recognized as erroneous language element codes. Even if output from the recognition unit 13, it is removed by the filter unit 14 and is not passed to the application program 15, so that malfunction of the application program 15 can be prevented. Therefore, it is possible to provide a speech recognition device that can accurately code speech even when used in a case where only necessary words and phrases are recognized from a speaker's conversation.

Next, with reference to FIG. 8, a fourth embodiment in which the present invention is applied to a business processing device for processing a deposit from a customer using an electronic cash register, a POS terminal, or the like. Will be explained. Note that the functional block diagram of the business processing device in the present embodiment, the speech recognition resource 1
2 are the same as those shown in FIGS. 5 and 6, respectively, and therefore, detailed description thereof will be omitted.

The difference between the filter unit 14 of the present embodiment and the third embodiment is that the language element code obtained by voice-recognizing the deposit amount from the customer is viewed from the total price of the products from the application program 15. Instead of judging whether or not it is something that cannot be considered normally based on the magnitude relationship, a condition that is usually satisfied from the viewpoint of the total price of the product is determined, and judgment is made based on this condition.

For example, when the total price of the product is a multiple of 5 such as 255 yen, the deposit amount is usually 1
Although it is conceivable that it is a multiple of 5 such as 005 yen or 500 yen, it is not considered that it is 1006 yen. In consideration of this, the condition set by the filter unit 14 may be, for example, that the language element code (deposit amount) is a multiple of five. Here, the condition is further narrowed down, and a condition is set in advance that the value obtained by subtracting the total price of the product from the language element code (deposit amount) is also a multiple of 5. A plurality of conditions may be prepared in advance, and a necessary condition may be selected according to the total amount.

In the embodiment of the present invention having such a configuration, for example, when the user of the present apparatus utters “1000 yen” as shown in FIG.
Is converted into a digital signal and supplied to the voice recognition unit 13. Then, the speech recognition unit 13 refers to the speech recognition resource 12 to detect the similarity / approximation between the input speech and a phrase defined in advance in the speech recognition resource. The language element code “1005” is output.

On the other hand, in the filter unit 14, the above-described condition is selected based on the total amount of the product of “255” yen from the application program 15, and the speech recognition unit 13
It is determined whether the language element code “1005” from satisfies the above-described condition. In this case, since the condition is satisfied, it is determined that speech recognition has been normally performed, and the language element code is passed to the application program 15.

Also, for example, the user of the present apparatus is shown in FIG.
When "500 yen" is uttered as shown in FIG. 7, this voice is converted into a digital signal by the voice input unit 11 and
Supplied to Then, the speech recognition unit 13 refers to the speech recognition resource 12 to detect the similarity / approximation between the input speech and a phrase defined in advance in the speech recognition resource. The language element code "500" is output.

On the other hand, in the filter unit 14, the above-described condition is selected based on the total price of the product of “255” yen from the application program 15, and the voice recognition unit 13
It is determined whether or not the language element code “500” from satisfies the above condition. In this case, since the condition is satisfied, it is determined that speech recognition has been normally performed, and the language element code is passed to the application program 15.

On the other hand, for example, when the user of this apparatus utters “XXX” as shown in FIG. 8C, this voice is converted into a digital signal by the voice input unit 11 and 13 is supplied. Then, the speech recognition unit 13 refers to the speech recognition resource 12 and detects the similarity / approximation between the input speech and a phrase defined in advance in the speech recognition resource. It is assumed that a language element code “1006” is output by the recognition.

On the other hand, in the filter unit 14, the above-described conditions are selected based on the total price of the product of “255” yen from the application program 15, and the speech recognition unit 13
It is determined whether the language element code “1006” from satisfies the above-described condition. In this case, since the condition is not satisfied, it is determined that the speech has not been normally recognized, and the language element code is not passed to the application program 15.

As described above, only when the language element code from the voice recognition unit 13 satisfies the predetermined condition determined based on the total price of the product from the application program 15, the language element code is Since the filter unit 14 is provided so that it can be passed to the application program 15, even if speech recognition is performed from a conversation between the customer and the clerk, the application program 15 does not normally think of unnecessary words and phrases, for example, from the viewpoint of the total price of goods. Regarding the voice of the deposit amount, greetings, small talk, noise, etc., even if an erroneous recognition and unnecessary language element code
Even if it is output from 3, the filter is removed by the filter unit 14 and is not passed to the application program 15, so that malfunction of the application program 15 can be prevented. Therefore, similarly to the first embodiment, a speech recognition apparatus capable of accurately encoding speech even when used for speech recognition of only a necessary phrase from a speaker's conversation is provided. be able to.

The filter unit 1 according to the present embodiment
The condition set in 4 is that if the last digit of the total price of the product is 0, for example, 1550 yen, the language element code (deposit amount) is a multiple of 50 and the language element code (deposit amount) The condition that the value obtained by subtracting the total price of the product from the price) is also a multiple of 50 may be set in advance.
This makes it possible to remove 1560 yen, which is not normally considered as a deposit amount, as compared with the case of “multiple of 5”.

Next, a fifth embodiment in which the present invention is applied to a business processing device for processing a deposit from a customer using an electronic cash register, a POS terminal, or the like will be described with reference to FIGS. 9 to 11. This will be described with reference to FIG. Note that, in the present embodiment, the same portions as those in the first embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted.

As shown in FIG. 9, the business processing device according to the present embodiment includes a voice input unit 11 and a voice recognition resource 1.
2. Speech recognition unit 13, filter unit 14, application program 15 using the language element code from filter unit 14, and speech recognition unit 13
And a product information database (product information storage means) 21 which collects product information of the handled products to be referred to when determining whether to remove the language element code from the product.

The voice recognition unit 13, the filter unit 14, and the application program 15 in the present embodiment
Is composed of a personal computer having a CPU, ROM and RAM. These voice recognition unit 13, filter unit 14, and application program 15
Is, for example, a software program stored in a storage device such as a hard disk device or a memory such as a ROM and readable by the CPU of the personal computer.

The speech recognition resource 12 is constituted by a storage device such as a hard disk device. The specific configuration is the same as that shown in FIG. 2, and a detailed description thereof will be omitted.

The product information database 21 is constituted by a storage device such as a hard disk device. Specifically, it is configured by collecting merchandise information in which the merchandise names, language element codes, unit prices, and stock quantities of the handled merchandise as shown in FIG. 10 are associated with each other. For example, if the product name is "Product A",
A language element code of “101”, a unit price of “100” yen, and a stock quantity of “1000” are associated.
The product information in the product information database 21 is updated by, for example, a sales process of the product.

The filter section 14 in the present embodiment is
The product information is retrieved from the product information database 21 using the language element code from the voice recognition unit 13 as a key. When there is no corresponding product information, the language element code is removed, and when there is the corresponding product information, the language element code is passed to the application program 15.

That is, when the speech recognition unit 13 misrecognizes the language element code, it is highly likely that the output language element code does not actually include product information. Therefore, the language element code from the voice recognition unit 13 is stored in the product information database 21.
By determining whether or not the product information is a language element code in which the product information exists, unnecessary language element codes in which the product information does not exist are removed so as not to be passed to the application program 15.

In the embodiment of the present invention having such a configuration, for example, when a user of the present apparatus utters “product A” as shown in FIG.
The signal is converted into a digital signal at 1 and supplied to the voice recognition unit 13. Then, the voice recognition resource 13
Is detected, the similarity / approximation between the input voice and the phrase defined in advance in the voice recognition resource is detected, and the language element code “0101” is output for “product A”. This language element code is passed to the next filter unit 14. Then, the filter unit 14 searches the product information database 21 for product information using the language element code as a key. In this case, since there is corresponding product information, the language element code is passed to the application program 15.

On the other hand, the user of this apparatus
When “XXX” is uttered as shown in (b), this voice is converted into a digital signal by the voice input unit 11 and supplied to the voice recognition unit 13. Then, the speech recognition unit 13 refers to the speech recognition resource 12, detects the similarity / approximation between the input speech and a phrase defined in advance in the speech recognition resource, and incorrectly recognizes “xx”. For example, assume that the speech recognition unit 13 outputs an undefined language element code of “0105”.

Then, the filter unit 14 searches the product information database 21 for product information using the language element code as a key. In this case, since there is no corresponding product information, the language element code is removed and is not passed to the application program 15.

As described above, the product information is searched from the product information database using the language element code from the voice recognition unit 13 as a key, and the language element code is passed to the application program 15 only when there is the corresponding product information. Since the filter unit 14 is provided, even if speech recognition is performed from a conversation between a customer and a clerk, unnecessary words and phrases such as greetings, small talk, and noise are not recognized by the application program 15 even if the unnecessary language is recognized. Even if the element code is output from the voice recognition unit 13, it is removed by the filter unit 14 and the application program 1
5 is not passed to application program 1
5 can be prevented from malfunctioning. Therefore, the first
As in the case of the first embodiment, it is possible to provide a speech recognition apparatus capable of accurately encoding speech even when used in a case where only necessary words and phrases are recognized in a conversation between speakers.

The filter unit 1 according to the present embodiment
In step 4, when the product information is retrieved from the product information database 21 using the language element code from the voice recognition unit 13 as a key, if the stock quantity is 0, the language element code is removed. You may make it. As a result, it is possible to prevent the language element code of the product whose stock quantity is 0 from being passed to the application program.

Although the above first to fifth embodiments have been described with respect to the case where they are applied separately, these embodiments may be applied in combination.

[0068]

As described above in detail, according to the present invention, even when speech recognition is performed from a conversation between a customer and a clerk, unnecessary words and phrases such as greetings and small talks in an application program can be obtained.
Regarding noise, even if an unnecessary language element code is erroneously recognized and output from the speech recognition means, it is removed by the filter means and is not passed to the application program, so that malfunction of the application program can be prevented. it can. Therefore, it is possible to provide a speech recognition device that can accurately code speech even when used in a case where only necessary words and phrases are recognized from a speaker's conversation.

The speech recognition means selects the closest one of the words to be recognized in advance, but unnecessary language element codes can be removed by the filter means. The malfunction of the application program can be prevented without defining all of the user's characteristic wording of the phrase to be speech-recognized in the speech recognition resource.

[Brief description of the drawings]

FIG. 1 is a functional block diagram showing a configuration of a business processing device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of a speech recognition resource shown in FIG.

FIG. 3 is a diagram illustrating an operation of the embodiment.

FIG. 4 is an exemplary view for explaining the operation of a business processing device according to a second embodiment of the present invention;

FIG. 5 is a functional block diagram showing a configuration of a business processing device according to a third embodiment of the present invention.

FIG. 6 is a diagram showing a configuration of a speech recognition resource shown in FIG. 5;

FIG. 7 is a diagram illustrating the operation of the present embodiment.

FIG. 8 is a functional block diagram showing a configuration of a business processing device according to a fourth embodiment of the present invention.

FIG. 9 is a functional block diagram showing a configuration of a business processing device according to a fifth embodiment of the present invention.

FIG. 10 is a diagram showing a configuration of a speech recognition resource shown in FIG. 9;

FIG. 11 is a diagram illustrating an operation of the present embodiment.

FIG. 12 is a functional block diagram showing a configuration of a business processing device to which a conventional voice recognition device is applied.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 11 ... Speech input part 11a ... Microphone 11b ... A / D converter 12 ... Speech recognition resource 13 ... Speech recognition part 14 ... Filter part 15 ... Application program 21 ... Product information database

Claims (6)

[Claims] 1. A speech input means for inputting a voice of a speaker, a speech recognition means for recognizing a phrase from speech inputted from the speech input means, and a plurality of phrases to be recognized in advance and corresponding to each phrase. Language element code storage means for storing the obtained language element code, and when the phrase recognized by the speech recognition means includes a phrase to be recognized in advance, a language element code corresponding to each phrase is stored from the language element code storage means. Language element code output means for extracting and outputting, and filter means for removing unnecessary ones of language element codes output from the language element code output means and passing necessary ones, Voice recognition device. 2. A speech input means for inputting a voice of a speaker, a speech recognition means for recognizing a phrase from speech inputted from the speech input means, and a plurality of phrases to be recognized in advance and corresponding to each phrase. Language element code storage means for storing the obtained language element code, and when the phrase recognized by the speech recognition means includes a phrase to be recognized in advance, a language element code corresponding to each phrase is stored from the language element code storage means. A language element code output means for extracting and outputting, and at least one necessary language element code among a plurality of language element codes output from the language element code output means in one speech recognition process is not included. If necessary, filter means to remove all language element codes and, if all necessary language element codes are available, to pass those language element codes Characteristic speech recognition device. 3. A voice input means for inputting a voice of a speaker, a voice recognition means for recognizing a phrase from a voice inputted from the voice input means, and a plurality of words to be recognized in advance and a plurality of words. Language element code storage means for storing a language element code within a predetermined numerical range, and when the words recognized by the voice recognition means include words to be recognized in advance, the language element code corresponding to each word is stored in the language element code. A language element code output means for extracting and outputting the language element code from the language element code storage means, and removing the language element code if the language element code output from the language element code output means is out of a predetermined numerical range. And a filter means for passing the language element code when the value is within a predetermined numerical range. 4. A voice input means for inputting a voice of a speaker, a voice recognition means for recognizing a phrase from a voice input from the voice input means, and a voice recognition means for recognizing a deposit from a customer. Language element code storage means for storing a plurality of words to be indicated and a language element code indicating an amount of money corresponding to each word, and when the words recognized by the voice recognition means include words to be recognized in advance, A language element code output means for extracting and outputting a language element code corresponding to each word from the language element code storage means, and a language element code output from the language element code output means which takes in data of the price from the outside. If the amount shown is lower than the price of the data imported from the outside, the language element code is removed, and if the amount is equal to or more than the price of the data imported from the outside, the price is reduced. And a filter means for passing the language element code. 5. A voice input means for inputting a voice of a speaker, a voice recognition means for recognizing a phrase from a voice inputted from the voice input means, and a voice recognition means for recognizing a deposit from a customer. Language element code storage means for storing a plurality of words to be indicated and a language element code indicating an amount of money corresponding to each word, and when the words recognized by the voice recognition means include words to be recognized in advance, A language element code output means for extracting and outputting a language element code corresponding to each word from the language element code storage means, and a language element code output from the language element code output means which takes in data of the price from the outside. If the indicated amount does not satisfy the condition derived based on the price of the data taken from the outside, the language element code is removed, and if the condition is satisfied, And a filter means for passing the language element code. 6. A voice input means for inputting a voice of a speaker, a voice recognition means for recognizing a phrase from a voice input from the voice input means, and a plurality of words to be recognized in advance and corresponding to each word. Language element code storage means for storing the obtained language element code, product information storage means configured by collecting product information associated with the language element code, and the phrase recognized by the voice recognition means should be recognized in advance. A language element code output means for extracting and outputting a language element code corresponding to each word from the language element code storage means, and a language element code output from the language element code output means as a key The product information is retrieved from the product information storage means. If there is no corresponding product information, the language element code is removed. A speech recognition device comprising: a filter means for passing a raw code.