JPH05265482A - Information processor - Google Patents

Abstract

PURPOSE:To improve the speech recognition rate of the input speech from an unspecified speaker. CONSTITUTION:When a recognizing means 5 recognizes the speech input from an input means 4, the speech of the unspecified speaker is recognized by using standard speech information stored in an information storage means 6 and information in a difference information storage means 10 stored with difference information between the standard speech information in the information storage means 6 and speech functions characteristic to respective individuals, so the speech recognition is performed by utilizing the features of the vocalization of each unspecified speaker to provide a speech recognition device which is effective to service wherein a high recognition rate is required.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information processing apparatus that applies a corresponding service by recognizing a voice input instruction for an unspecified speaker.

[0002]

2. Description of the Related Art In recent years, a voice recognition device by voice input using a telephone line has been introduced in the financial industry and the distribution industry. Using these voice recognition devices, services such as bank deposit balance inquiry and order entry for various home deliveries are provided to users. Recently, voice recognition devices have been developed for home electric appliances and in-vehicle products, and have been applied to car audio and video reservations, telephones and the like.

An information processing apparatus having a conventional voice recognition function will be described below. FIG. 4 shows the configuration of a conventional information processing apparatus. In FIG. 4, reference numeral 51 is an information processing apparatus described here. As a component, 54 is an input means, which receives an input including a voice input. A recognition unit 55 performs voice recognition. An information storage unit 56 stores standard voice information necessary for voice recognition. 5
An output unit 8 outputs a message for prompting voice input, a voice recognition result, and the like.

The operation of the constituent elements related to each other will be described below. First, the user performs voice input according to an input prompting message (guidance by voice output or guidance display by a display device) from the output means 58 of the information processing device 51. The input means 54 accepts this voice input and performs the processing necessary for the recognition processing by the next recognition means 55.

First, the input analog signal is converted into a digital signal. Normally, 8 to 10 kHz is used as the sampling frequency, and at each sampling time,
It is quantized with a value of 8 to 16 bits. Recognition means 55
Is pattern-matched (hereinafter abbreviated as matching) with the input signal and the voice information in the information storage means 56 to select the most similar information above a certain threshold. Then, the speech recognition result matched as the uttered word is returned. Based on the voice recognition result, the content of the next input candidate to be requested is determined, and the content to be output from the output means 58 is determined.

When the information processing apparatus 51 is applied to a telephone, it is necessary to utter a number of about 10 digits in order to input the area code. At least a number from 0 (pronounced "zero") to 9 (pronounced "kyu") is uttered in sequence, the content is recognized, converted into dial pulse or tone signal, and a call is made. The recognition result is
The user can recognize the recognized result by using the display device of the output means 58 attached to the telephone or the voice output device (Japanese Patent Laid-Open No. 63-33796). The same applies to the input of a password other than the telephone number.

In order to shorten the time required to input a 10-digit number by voice as much as possible, one number is uttered, recognition is performed, and the result is displayed, while waiting for the next input. If it is recognized, it is converted into the corresponding numerical information and waits for the next input. When the voice cannot be recognized due to a small amount of utterance, a request for rewording is made and the system waits for the next input. As you become accustomed to voice input, you will be able to input without looking at information on the display device. Therefore, it is possible to input the telephone number by speaking in order without removing the line of sight from the corresponding number string in the telephone directory. The input result may be confirmed after the completion of all utterances, and it is not necessary to look at the display device or the like for each input.

[0008] In order to perform voice recognition in a device used by everyone, such as a public telephone, some conditions are required. One of the conditions is that it is necessary for any person's voice to be able to be voice-recognized without the user's voice registration work. This is because if the vocabulary is different for each device, it is necessary to register the voice every time, which may be a major factor that hinders the spread of the device. Also, the way the user speaks and the way the same person speaks depends on the environment in which the device is installed. Furthermore, the amount of utterance and the voicing frequency have large individual differences. Therefore, what is recognized by anyone's voice without specifying a user is defined as unspecified speaker recognition.

In unspecified speaker recognition, in order to create standard speech information for speech recognition, a plurality of people (several people to several hundred people) utter a predetermined vocabulary and the like. Is recorded and statistically processed. Then, information suitable for performing voice recognition is extracted, and the information is stored in the information storage means 56 as standard voice information. By matching the standard voice information stored in the information storage unit 56 with the voice input from the input unit 54, the vocal vocabulary is recognized by voice. In order to improve the voice recognition rate of unspecified speakers,
It is necessary to record more and different (gender, age, address) utterances and perform voice information processing.

Here, the recognition rate is defined as follows. It is assumed that the input voice within the threshold of voice recognition is matched from the specified word group, and the information indicating the same vocabulary as the input specified vocabulary is the probability obtained as the recognition result.
Therefore, the input means 54 is out of order, or the input voice is too small for the voice recognition to be excluded from matching. The probability is that no incorrect matching is performed.

In the unspecified speaker recognition, the recognition rate is about 95%, and the utterances of all persons cannot necessarily be recognized. This is because voice information processing is statistical processing,
This is because the average and variance of the recorded voices are calculated and the value of the standard deviation or the like is used to use voice information within a certain range. There is a condition that the amount of information is restricted in order to enable high-speed calculation of voice recognition. Therefore, it is difficult to have the amount of information that achieves the recognition rate of 100%. At the research level,
The calculation time for matching the voice recognition is not so much related, but at the practical level, it cannot be used if it takes several tens of seconds or more after the end of utterance. If the recognition result is not output within a few seconds at the longest after the utterance ends, it may cause the user to feel inconvenience.

However, since the unspecified speaker recognition can be recognized only by the user uttering, application to home electric appliances is under study. Voice reservation remote controllers for video equipment have been put to practical use. The main components of the voice signal processing of this voice reservation remote controller are an LSI for voice synthesis and an A
An IC for performing D / D conversion, a DSP (digital signal processing processor) for performing high-speed calculation for voice recognition, and a ROM of a memory storing standard voice information are stored. In the case of this device, the recognition result is output within 1 second after the end of utterance.

For example, in the voice reservation remote controller, the corresponding channel number, the reserved day of the week, the reservation start time and the reservation time are designated by voice according to the guidance output by the synthesized voice. The voice recognition result is output to the LCD display device which is the output means 58. If the voice recognition result matches the utterance content, the confirmation button switch is pressed, and if wrong, the voice is uttered again (until correctly recognized). is there.

On the other hand, there is also a voice recognition method in which the speaker is limited to a specific individual and the voice that the individual speaker wants to operate is registered in the information storage means 56 to compare the registered content with his or her own voice. .. This is defined as specific speaker recognition. In the specific speaker recognition, voice recognition is possible without collecting the vocalization information of many people. However, it is always necessary to utter a few times and register the voice first. In addition, there is a problem that only the registered person can use the device. Therefore, application of the specific speaker recognition is being considered for car audio for automobiles, automobile telephones, etc., which are fields in which users are limited in advance.

In general, the recognition rate of specific speaker recognition by the same vocabulary of the same person is higher than the recognition rate of the same vocabulary by unspecified speaker recognition. In this case, the specific speaker recognition can perform the recognition by making better use of the characteristics of the utterance. Therefore, since it is possible to calculate only the characteristic parameters of the individual speaker, the calculation amount is often smaller than that of the unspecified speaker.

[0016]

However, in the conventional configuration, when the voice recognition is applied to a device (public telephone, ticket vending machine, rental device, etc.) that does not limit the user (recognition of unspecified speaker), the voice recognition rate. However, there is a problem that the speech content of a specific person, which is quite different from the standard speech, cannot be recognized (misrecognized).

On the other hand, in order to use the specific speaker recognition technique,
Since the voice registration work has to be performed, there is also a problem that it is difficult to apply the device to the above-mentioned devices that do not limit users.

The present invention has been made in view of the above problems, and an object of the present invention is to provide an information processing apparatus having an increased voice recognition rate of an unspecified speaker.

[0019]

In order to achieve this object, an information processing apparatus of the present invention comprises an input means for receiving a voice input, a voice recognition recognition means, an information storage means referred to at the time of voice recognition, and A difference information storage means for storing difference information from information stored in the information storage means, and a control means for performing voice recognition using the information storage means and the information stored in the difference information storage means. I have.

[0020]

In the information processing apparatus of the present invention having the above structure, the difference information storing means stores the difference information content unique to each individual in the storage medium, and when the voice recognition is performed, the standard voice of the information storing means is controlled by the control means. By performing the unspecified speaker recognition with reference to the information and the difference information in the difference information storage means, it is possible to provide a device effective for a service requiring a high recognition rate.

[0021]

【Example】

(Embodiment 1) An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an information processing apparatus according to the first embodiment of the present invention. As shown in FIG. 1, as a component, 1 is an information processing device described here, and 2 is a voice recognition unit that performs voice information processing. An input unit 4 receives an input including voice. An information storage unit 6 stores voice recognition vocabulary information and the like. Reference numeral 5 is a recognizing means for performing a voice recognizing operation on the voice signal input from the input means 4 by referring to the information in the information storing means 6. An information control unit 7 controls information other than the information stored in the information storage unit 6. Reference numeral 10 is a difference information storage means for storing difference information with respect to the information stored in the information storage means 6, and reference numeral 3 is a storage medium as a storage device having the difference information storage means 10.

With respect to the information processing apparatus 1 configured as described above, the mutual operation of the components will be described.
As an assumption, an example in which the storage medium 3 is removable from the voice recognition unit 2 and is a portable medium will be described. For the embodiment of the storage medium 3 which is not removable, refer to Embodiment 2
Described in.

First, the user loads the storage medium 3 into the information processing device 1
To insert. The information processing device 1 detects that the storage medium 3 has been inserted and couples it with the voice recognition unit 2. Next, the information control means 7 determines whether the coupled storage medium 3 is a usable medium. If it is determined that it is available, the difference information in the storage medium 3 is acquired from the difference information storage means 10.

Next, the output means (not shown) outputs a message or the like for prompting the user to input information necessary for starting the service. For example, if you want to receive a service such as a cash withdrawal service, enter the applicable service name or the like and output it from the display device or the voice output device. The user relies on the output message to input using voice.

The service name input from the input means 4,
The input voice such as a personal identification number is matched based on the standard voice information stored in the information storage means 6 and the difference information acquired by the information control means 7 to perform voice recognition. Various information processing is performed by performing voice recognition, converting the relevant information from the voice signal into information such as appropriate numeral and character information. The output unit is configured not only to output an input request but also to apply various services, by configuring a display device and a voice output device individually or in combination.

The matched voice information is stored as follows. The vocabulary that can be recognized by voice is divided into groups and stored for each usage type.
For example, the first group includes service application names (eg, deposit,
Withdrawals, balances, credits, etc.), the second group is numbers used for inputting personal identification numbers (“Zero” to “Kyu”), and the third group is vocabulary for service confirmation instructions (cancellation, correction, Confirmation, confirmation, etc.) etc.

By grouping the vocabulary information in accordance with the purpose of use, work is performed to find the one that is closest to the most vocalized voice information in the designated group. Of course, each information is coded into information that facilitates matching in a short time. Further, a threshold value is provided for this matching, and the recognition means 5 has a threshold value judgment means for outputting a voice recognition result that there is no matching vocabulary unless it is a certain value or more.

When the input voice exceeds the threshold of a specific input level, the start of utterance having vocabulary information is detected, and when the input voice falls below the threshold, the end of utterance is confirmed. Input voice in this vocal section is several kHz (8kHz to 10kHz)
Are sampled, and at each sampling time, about 8 to 12 bits are quantized and digitized.

For speech recognition, for example, an LPC cepstrum coefficient is obtained by using this digital value, and this is used as a characteristic parameter and compared with the parameter amount of the standard vocabulary to give the closest one by its group number. Find out from the given vocabulary. This is matched throughout the vocalization section, and finally the closest one is recognized as the vocalized vocabulary. Although there are various methods other than the LPC cepstrum coefficient, a method is generally used in which the digital value obtained by the above-described sampling is subjected to various signal processings to extract characteristic parameters and matching is performed.

The information storage means 6 is subjected to statistical processing from utterances of several people to several hundred people, and has LPC cepstrum coefficient which is a characteristic parameter, a voice power value (a voice level of a specific frequency band, etc.), The utterance time, the matching search coefficient, etc. are stored. However, as explained in the conventional example, due to constraints such as arithmetic processing performance, the values obtained from the average, variance, standard deviation, etc. are used, so that there is a tendency to recognize a vocal vocabulary that is far from the standard. Is difficult to do accurately.

Therefore, the difference information storage means 1 in the storage medium 3
The difference information in which the voice information of the user is different from the average information is stored in 0, and the difference information is subtracted from the standard information of each feature parameter to perform the recognition process. Thereby, the information peculiar to the individual is reflected, and the recognition rate is improved.

Only the difference information from the standard voice information is stored in the storage medium 3
Therefore, it is not necessary to retain all the voice information necessary for voice recognition. Further, the information storage means 6 of the voice recognition unit 2 need only hold standard voice information. A sufficient recognition rate can be raised by subtracting the correction amount due to the difference. Therefore, when collecting the standard voice information, there is an advantage that the standard voice information can be easily created with a smaller number of voice information than before.

Typical examples of the voice recognition include phoneme recognition, word recognition, and sentence recognition as recognition units. The phoneme recognition is mainly for recognizing voices in units of consonants and vowels. In word recognition, voice recognition is performed in units of one word (a few seconds at the longest). Sentence recognition is for recognizing a grammar and the like in units of clauses and sentences consisting of a plurality of words and performing speech recognition.

For the phoneme recognition in which the voice recognition is carried out in units of phonemes, standard voice information is created for each phoneme information, and
Create difference information. The difference information in this case is a value for each phoneme. On the other hand, for word recognition performed on a word-by-word basis, standard voice information is created for each word information, and the difference information is also information for utterance of each word. Alternatively, word recognition can be performed using phoneme information that constitutes a word. The storage medium 3 includes difference information for phonemes, and the information storage unit 6 includes phoneme-based speech standard information and vocabulary information in which phonemes are combined. The sentence unit is a combination of multiple word recognitions.

Registration of the difference information from the standard voice information is performed as follows. Ask the user to utter an applicable vocabulary including phoneme information, and for example, a service application company or the like may use the DAT (digital audio tape) to describe the user's utterance content.
Record on a high-performance recording device such as. A service application company or the like processes the DAT tape by voice processing to obtain difference information,
Write to the difference information storage means 10 of the storage medium 3. In addition, it is possible to use a telephone or the like. The current voice band that can be used in a telephone is from several tens of Hz to 3.4 kHz, but even information in this band includes information that allows voice recognition. However, since various noise components are also included, even if correction processing is performed, it is not possible to expect to collect voices with high sound quality. If digital telephones using ISDN (Integrated Services Digital Network) become widespread, the amount of information that can be transmitted will dramatically increase and a wide band can be used, so it will be possible to collect high-quality voice from home telephones. Become. It is also effective to collect voices using a home recorder.

The removable storage medium 3 is composed of the following medium. Card-shaped storage medium 3 such as a magnetic card, optical card, IC memory card, IC card, CD-
There is a disk-shaped storage medium 3 such as a ROM or a magneto-optical disk. For example, magnetic card, optical card, CD-ROM
In the storage medium 3 having only storage means such as the above, the content of the difference information storage means 10 is read into the memory included in the information control means 7, and the recognition means 5 refers to it. There is also a non-contact type storage medium 3 as described above. Taking the form of communication via electromagnetic induction, radio waves, light, etc.,
The information stored in the difference information storage means 10 is acquired.

On the other hand, there is a storage medium 3 such as an IC memory card or an IC card, in which the voice recognition unit 2 can be used as a memory means in addition to the storage means. In such a case, the information control means 7 has a communication function with the storage medium 3 and can be configured to perform voice recognition using the memory means of the difference information storage means 10.

The following items are examples of the difference information.
For example, there is a case where difference information that the voice section is short is stored. Examples of such information include difference information from the standard value at each sampling time, voicing section length information, total or partial voicing power magnitude information, beginning and ending utterance strength feature information, and voice pitch. Information, setting path information for time setting when matching with each sampling value, and the like. The voice recognition rate is increased by acquiring these pieces of information from the difference information and performing voice recognition after knowing the characteristics of the user.

The calculation method using the difference information includes the following examples. For example, a method of subtracting the difference value that is far from the standard value from the input voice and matching with the standard value, a method of changing the calculation limit by the standard value by the difference value (changing the threshold value), and inputting the standard value of the difference value Method in which the corrected standard value and the input voice are matched in advance,
After matching with standard voice information, there is a method of further matching using difference information.

The recognition rate is as defined in the conventional example,
This is the probability of selecting the correct vocabulary during the speech recognition process for inputs within the threshold. The vocalization is too small, etc.
Inputs other than the threshold are not included in the recognition rate. However, from the user's perspective, it is likely that a recognition error has occurred even when the input is too small to be recognized. Therefore, it is also necessary to increase the recognition rate in a broad sense, that is, the probability that a correct recognition result can be obtained while uttering a voice (even if the voice is small).

When the use of the storage medium 3 is started, it is necessary to confirm the validity of the storage medium 3 of the information control means 7. In the storage medium 3 such as a magnetic card, in addition to the difference information stored in the storage medium 3, inspection information such as error detection and error correction and registration number information are stored, and the information control means 7 reads out the respective information. Then, the relevance of each information such as the difference information and the inspection information is confirmed, and the validity is confirmed. It may be configured to store each information by using a means for encrypting each information as needed.

On the other hand, when the storage medium 3 is composed of an IC card, the function of the IC card (the inside cannot be referred unless the corresponding password such as the personal identification number is collated) is used as the validity confirmation means. In the case of an IC card, the validity of the user can be confirmed by inputting a personal identification number or the like, but it is also possible to confirm the validity of the voice recognition unit 2 in the same manner. It is also possible to hold the model number information of the voice recognition unit 2 and the information processing apparatus 1 that can be used in the IC card and use it only when the permitted model number information is obtained.

Further, in order to confirm the legitimacy of the IC card side, necessary information is given to the IC card, a calculation is performed based on the information, and the result is returned to the information control means 7. On the other hand, it is possible to confirm the validity of the IC card by performing a calculation also on the information control means 7 side and comparing it with the information obtained from the IC card side. If the information given to this IC card and the calculation means are changed each time, it becomes difficult to be illegally used, and it can be an effective means even in applications requiring high security.

Depending on the contents of use, especially for contents that do not require security, when using the difference information, even if the stored information is copied to another person, the voice of the individual user Because it is information, it is of little value to others. In the commonly used one, this difference information serves only the purpose for the convenience of the user. Further, since the difference information is valid only for the user, it only has to have a function of determining whether the storage medium 3 is correct. However, this does not apply to services that require high security.

When information such as ambient noise is mixed and the utterance section is erroneous, the voice recognition may not be performed correctly. Therefore, the input means 4 has a function of periodically acquiring the ambient noise level and correcting the input level, and also displays to the user a message that the user wants to speak with a slightly louder voice. Or an output unit (not shown) such as a loudspeaker. On the other hand, if the previous utterance was too loud and could not be matched with a suitable vocabulary and the corresponding vocabulary could not be selected, the gain of the input level is adjusted and an output indicating that a little voice is desired is output. In the case of uttering a plurality of vocabularies continuously, the above gain adjustment is effective.

Depending on the content of the applied service, it is determined whether the vocal vocabulary is in the form of being selected from the menu or in the form of showing only the input prompting message and not displaying the selected vocabulary. Further, since the recognition vocabulary is in the form of the information storage means 6 and is separated from the recognition means 5, the information storage means 6 can be composed of various memories. By configuring a memory that stores various vocabularies, it is possible to provide a wide variety of information processing devices 1 by selecting and using vocabulary information according to the purpose.

Further, depending on the form of use, the difference information of the voices of a small number of people (family etc.) can be obtained and applied to the voice recognition of the information processing apparatus used at home. The voice information stored in the storage medium 3 is not limited to an individual, and may be a specific group. In such a case, when the difference information storage means 10 for a plurality of persons is prepared for the number of persons, there is a configuration in which the plurality of persons are averaged and the difference information storage means 10 is integrated. The latter is expected to have a lower recognition rate than the former, but in the case of family members, etc., there are similar points in vocalization and intonation. Information processing device 1 for recognition
Can be configured.

According to the present invention, the voice recognition is performed by using the voice input and the feature of the individual contained in the voice is taken into account, so that a high recognition rate can be obtained even in the device which does not limit the user. It is possible to provide a device effective for speech recognition of various devices such as various information devices and home electric appliances.

(Second Embodiment) A second embodiment of the present invention will be described below. The configuration is the same as that of the first embodiment,
An example in which the storage medium 3 is fixed, not removable, will be described. The fixing means is fixed to the information processing apparatus 1 from the beginning, or is fixed so that the user does not remove it after the user has installed it according to the purpose unless it is disassembled and repaired. Two examples will be described.

First, the case of being fixed from the first beginning will be described. The information storage means 6 is realized by a storage means such as an IC memory card. Next, the service provider records the voice information of the user and stores the difference information in the storage medium 3.
It is possible to adopt a configuration in which the service provider sends a medium containing both of these pieces of information to the user, and only the person who has the medium can use the information processing apparatus 1.

The voice information at this time is not limited to word information,
Storing difference information between consonant and vowel phonemes enables effective speech recognition by comparing the phoneme information of consonants and vowels contained in those words, no matter what word information is included in the standard speech information. You can Of course, phoneme information may be used as the standard speech information, and the information may be configured to have direct difference information with respect to each phoneme information.

Since the voice information for standard use also has a different vocabulary for use depending on the type of service, it is necessary to prepare different standard voice information. However, the difference information is
It is possible to configure services that do not need to be changed even if the types of services are different. In this example, the case where the medium containing the audio information and the other part are separated has been described. However, this is a form within the manufacturing and sales company, and in the case where it actually covers users, the storage medium 3 is fixed to the information processing device 1 from the beginning. At the stage of reaching the user's hand, it is possible to configure the information processing apparatus 1 that performs voice recognition reflecting individual characteristics without registering voice.

Next, an example in which the second user fixes the storage medium 3 will be described. Assume a device that recognizes an unspecified speaker, such as a home electric appliance described in the conventional example. In the normal case, it can be recognized by the standard voice information prepared from the beginning. However, it is preferable to be able to provide a service similar to that of a person without a disability to a person who has a damaged vocal cord or a specific person who has difficulty in voice recognition because he / she has old teeth and does not have solid teeth.

In such a case, by attaching the storage medium 3 storing the difference information of the person as an option of the main body device, an information processing apparatus capable of providing the same service as a person who makes an average voice is constructed. be able to.

For information processing devices having different services and different vocabulary, a plurality of storage media 3 of the voice recognition device are prepared, and the contents of the difference information storage means 10 as basic information are copied and attached to each device. .. Since voice recognition can be used in any attached device, it is much more convenient to use than a specific speaker voice recognition device that registers voices for each device.

When the specific speaker recognition is used, there is an inconvenience that registration work is involved, but since it is not suitable for voice recognition of a plurality of people such as family members, it is a service for only one user, and everyone can use it. If you try to use it in a usable device, it will be expensive. Therefore, with the configuration of the present invention,
Since the voice information of a plurality of persons can be answered, it can be used as an information processing apparatus for purchasing in units of family.

As described above, according to this embodiment, even if the storage medium 3 containing the difference information is fixed, it is convenient for the user and has a high recognition rate without impairing the characteristics of the unspecified speaker voice recognition. An expensive information processing device can be provided.

(Embodiment 3) A third embodiment of the present invention will be described below with reference to the drawings. FIG. 2 shows the configuration of the information processing apparatus in the third embodiment.
As shown in FIG. 2, as the components, the voice recognition unit 2, the storage medium 3, the input unit 4, the recognition unit 5, the information storage unit 6, the information control unit 7, and the difference information storage unit 10 are the same as those in the first embodiment. Is. Reference numeral 8 is an output means.

With respect to the information processing apparatus 1 configured as described above, the mutual operation of the constituent elements will be described.
First, the user inserts the storage medium 3 into the information processing device 1.
The information processing device 1 detects that the storage medium 3 has been inserted, and the information control means 7 determines whether or not the combined storage medium 3 is an available medium. If it is determined that it is available, the difference information in the storage medium 3 is acquired from the difference information storage means 10.

Next, the output means 8 outputs a message or the like for prompting the user to input information necessary for starting the service. The user relies on the output content of the message to input using voice. The voice input from the input unit 4 is matched based on the voice information stored in the information storage unit 6 and the difference information acquired by the information control unit 7, and voice recognition is performed. Various information processing is performed by performing voice recognition, converting the relevant information from the voice signal into information such as appropriate numeral and character information. Output means 8
Is used not only to output an input request but also to apply various services by configuring a display device and a voice output device individually or in combination.

When the information control means 7 acquires the difference information, it outputs a message to the user according to the acquired information. For example, when the difference information that the voice section is short is stored, a message such as "Speak slowly" is output. Examples of this are the length of the vocalization section, the amount of vocalization power, the beginning of the word,
There are features such as the strength of the ending utterance and the pitch of the voice.

These pieces of information are obtained from the difference information,
By outputting the comment message to the user from the output means 8, if the user consciously speaks, the difference from the standard voice can be reduced as much as possible, and the voice recognition rate can be improved. .. The difference information indicates the difference from the standard value, but it has the effect of reducing the probability of error as much as possible by using it as additional information for utterance that is closer to the standard value than the value corrected by the difference value. ..

(Embodiment 4) A fourth embodiment of the present invention will be described below with reference to the drawings. FIG. 2 shows the configuration of the information processing apparatus in the fourth embodiment.
As shown in FIG. 2, the voice recognition unit 2, the storage medium 3, the recognition unit 5, the information storage unit 6, the information control unit 7, and the difference information storage unit 10 are the same as those of the first embodiment. Reference numeral 8 is an output means. 4 is an input means,
The input unit 4 has an input gain changing unit (not shown).

With respect to the information processing apparatus 1 configured as described above, the mutual operation of the constituent elements will be described.
The information control unit 7 determines whether or not the storage medium 3 is available, and if it is determined as available, the difference information in the storage medium 3 is acquired from the difference information storage unit 10.

Next, the output means 8 outputs a message or the like for prompting the user to input information necessary for starting the service. The input voice such as the service name and the personal identification number input from the input means 4 is matched based on the voice information stored in the information storage means 6 and the difference information acquired by the information control means 7 to obtain a voice. Recognition is done.
Various information processing is performed by performing voice recognition, converting the relevant information from the voice signal into information such as appropriate numeral and character information.

When the information control means 7 acquires the difference information, it issues a request for changing the input gain to the input gain changing means according to the acquired information. For example, when the difference information that the audio power amount is small is stored,
Increase audio input gain to improve input sensitivity. Examples of this are the length of the vocalization section, the amount of vocalization power, the beginning of the word,
There are strong and weak utterances of endings. By obtaining these pieces of information from the difference information and adjusting the voice input gain in order to make the input means 4 suitable for the user's voice input,
The difference from the standard voice can be reduced as much as possible, and the voice recognition rate can be improved. The difference information indicates the difference from the standard value. However, by using the difference information as additional information for utterance that is closer to the standard value than the value corrected by the difference value, the probability of error is the same as in the third embodiment. Has the effect of lowering even a little.

Of course, as described in the third embodiment,
If the output unit 8 has a unit for outputting information using the difference information and has both the output unit 8 and the input gain changing unit, the recognition rate is further increased.

(Embodiment 5) A fifth embodiment of the present invention will be described below with reference to the drawings. FIG. 3 shows the configuration of the information processing apparatus in the fifth embodiment.
As shown in FIG. 3, as the components, the voice recognition unit 2, the storage medium 3, the input unit 4, the recognition unit 5, the information storage unit 6, the information control unit 7, and the difference information storage unit 10 are the same as those in the first embodiment. Is. Reference numeral 8 is an output means, and 9 is a difference information updating means.

Regarding the information processing apparatus 1 configured as described above, the mutual operation of the constituent elements will be described.
The information control unit 7 determines whether or not the storage medium 3 is available, and if it is determined as available, the difference information in the storage medium 3 is acquired from the difference information storage unit 10.

Next, the output means 8 outputs a message or the like for prompting the user to input information necessary for starting the service. The input voice such as the service name and the personal identification number input from the input means 4 is matched based on the voice information stored in the information storage means 6 and the difference information acquired by the information control means 7 to obtain a voice. Recognition is done.
Various information processing is performed by performing voice recognition, converting the relevant information from the voice signal into information such as appropriate numeral and character information.

When the difference information is acquired, the information control means 7 outputs the corresponding information to the output means 8 according to the acquired information, as in the third embodiment. Further, as a result of performing this output processing and performing voice recognition, the difference information with the standard voice information is written into the difference information storage unit 10 by the difference information updating unit 9.

With this configuration, there are the following advantages. When the storage medium 3 is newly issued, the difference information of the user may not be stored in the difference information storage means 10 depending on the use. In addition, difference information corresponding to a new vocabulary may be newly required. In such a case, initially, the voice recognition is performed using the standard voice information stored in the information storage means 6. Then, the difference information updating means 9 writes the difference information from the standard value at that time into the difference information storage means 10.
From now on, based on the written difference information, the information of both the information storage means 6 and the difference information storage means 10 is used,
Voice recognition can be performed. Therefore, even if the voice recognition is erroneous because the individual characteristics cannot be grasped at the first time,
From now on, you can perform voice recognition that incorporates individual characteristics.

There are also cases where the user does not use the storage medium 3 for a long time. In some cases, the address environment may change or the utterance may change due to illness. Even in such a case, the information processing apparatus 1 that more surely captures the utterance of the user by first performing voice recognition using the previous difference information and writing the difference information subjected to the voice recognition.
Can be configured.

Depending on the use, even the same vocabulary information may be uttered early or slowly. Eg 0
In the case of targeting the voice and numerical information from (Zero) to 9 (Kyu), in the case of menu-type number selection,
Although one number is uttered slowly, when uttering plural-digit number information like a telephone number, each number tends to be short and continued to be uttered. Therefore, if a plurality of pieces of difference information of the same vocabulary are provided or a function of rewriting the difference information each time is provided according to the purpose, the voice recognition rate is further increased and the convenience of the user is improved.

Further, as in the case of the fourth embodiment, a configuration having an input gain changing means is adopted, and as the difference information, for example, the information that the utterance level tends to be small as a whole is used, and the input gain level of the input means 4 is used. In addition to the above, the output unit 8 may be configured to output a message or the like indicating that a loud voice is desired.

The storage medium 3 may be configured to store the voice recognition result information. The following processing is also possible depending on the use for which security is required. If incorrect input is repeated many times, it can be assumed that it is being used illegally or that an inapplicable service has been selected.Therefore, stop the service or further set the pattern matching threshold. It is necessary to take measures such as tightening.

When the voice recognition error occurs frequently and the utterance which is far from the standard voice information is repeated, the past result information in which the result of the voice recognition is stored is referred to and the information stored in the storage medium 3 is used. It has a function of stopping the service using the processing device 1.

Further, with respect to applications in which security is not required so much, the error correction tendency of voice recognition may be stored as the above result information and corrected. For example, if the utterance of 7 (shichi) is close to 1 (1), and if it is often mistaken for 1 (1), this error tendency information is taken in and 7 (shichi) is automatically selected from the tendency information. Can be configured to.

Therefore, in this embodiment, the voice recognition is performed using the voice information input by voice and the standard voice information in the information storing means 6, and the means for writing the difference information from the standard is used to make the voice of the user. It is possible to provide the information processing apparatus 1 that can cope with changes in the above and that has improved convenience.

[0080]

As is apparent from the above description, the present invention shows the recognition means for recognizing the input voice, the information storage means for storing the standard voice information to be referred at the time of recognition, and the characteristics of the user's individual utterance. By including a difference information storage unit that stores difference information, and a control unit that controls the recognition unit to recognize it from the standard voice information and the difference information when performing voice recognition Since it is possible to perform voice recognition using features, an information processing device having a high voice recognition rate that makes use of the features of the user's utterance is realized. Further, by using the removable storage device for the difference information, it is possible to provide an information processing device which is excellent in terms of confidentiality.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an information processing apparatus according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of an information processing device according to third and fourth embodiments of the present invention.

FIG. 3 is a block diagram showing a configuration of an information processing device according to fifth and sixth embodiments of the present invention.

FIG. 4 is a block diagram showing a configuration of a conventional information processing device.

[Explanation of symbols]

 1 Information Processing Device 2 Voice Recognition Unit 3 Storage Medium 4 Input Means 5 Recognition Means 6 Information Storage Means 7 Information Control Means 10 Difference Information Storage Means

Claims (7)

[Claims] 1. An information storage in which input means for inputting voice, recognition means for recognizing the content of the voice input by the input means, and voice information referred to by the recognition means at the time of voice recognition are stored. Means, and difference information storage means for storing difference information between the standard voice information stored in the information storage means and the voice information peculiar to each individual, and recognizing the content of the voice input by the input means. In this case, the recognizing means has a control means for controlling the voice recognition by referring to the standard voice information stored in the information storage means and the difference information stored in the difference information storage means. Information processing device. 2. The information processing apparatus according to claim 1, wherein the difference information storage means is a removable storage device. 3. An input gain changing means for changing the input level of the input means, wherein the control means controls the input level to a predetermined level according to the difference information stored in the difference information storage means. The information processing apparatus according to claim 1, further comprising: 4. A difference information updating means for changing the difference information stored in the difference information storing means, wherein the difference information updating means is stored in the voice input inputted to the input means and the information storing means. The information processing apparatus according to claim 1, wherein the difference information is updated from standard audio information. 5. An input gain changing means for changing the input level of the input means is provided, and the control means controls the input level to a predetermined level according to the difference information stored in the difference information storage means. 5. The information processing apparatus according to claim 4, wherein the difference information updating means then updates the difference information from the voice input input to the input means and the standard voice information stored in the information storage means. 6. A difference information updating means for changing the difference information stored in the difference information storing means, wherein the difference information updating means is stored in the voice input inputted to the input means and the information storing means. The information processing apparatus according to claim 1, wherein error tendency information peculiar to each individual is added to standard voice information in addition to the difference information. 7. An output means for presenting a message to the user, wherein the control means presents the message to the output means in accordance with the difference information stored in the difference information storage means. The information processing device described.