JP2000322088A - Voice recognition microphone, voice recognition system, and voice recognition method - Google Patents

Abstract

(57) [Summary] To provide an easy-to-use interface for voice recognition in a voice recognition system used for a car navigation system, a small information device, a game, and the like. SOLUTION: A voice recognition microphone 1 having a function of executing voice recognition processing is connected to a system main body 5 by a communication means 4, and a recognition result recognized by a voice recognition section 13 is transferred to the system main body 5 to operate the system. Make it work. The voice recognition microphone 1 includes a voice recognition unit 13 having an acoustic model 135 and a voice recognition processing unit 134, a command dictionary 131, a recognition target dictionary 152, a user registration dictionary 153, and a data communication unit 1.
The necessary words are registered in the user registration dictionary 154 from the recognition target dictionary 152 transferred from the system main body 5, and the command dictionary 151 and the user registration dictionary 15 are usually stored.
4 to perform speech recognition.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a speech recognition system and method, and more particularly to a car navigation system,
This is a voice recognition system used for in-vehicle PCs, car electronics, small information devices such as PDAs and handheld PCs, portable voice translators, game devices, and home electric appliances. In the field of car multimedia represented by PC and car electronics, the present invention relates to an easy-to-use speech recognition system and method in terms of improving recognition response time and recognition rate.

[0002]

2. Description of the Related Art In recent years, small information systems using voice recognition technology have been around for a long time. In addition to car navigation systems, small information devices represented by PDAs, portable translators, and the like. As an example of such a speech recognition system, Japanese Patent Application Laid-Open No. 5-35776 discloses a "translation device with an automatic language selection function" that recognizes, translates, and translates an operator's voice input from a microphone. A technology related to a portable translation device that outputs voice is disclosed.

[0003] An outline of such a speech translation apparatus according to the prior art will be described below with reference to FIG. FIG. 6 is a block diagram showing a configuration of a speech translator according to the related art. The speech translation device 7 including the speech recognition means includes a microphone 71, a speech recognition unit 72, a translation unit 73, a control unit 7
4 and a bus 75 for transferring data through these sections 72, 73 and 74.

[0004] The voice recognition section 72 includes a voice section extraction section 721.
, A voice recognition processing unit 722, and a voice model storage unit 723
And a voice recognition dictionary unit 724.

[0005] The translation unit 73 has a memory card 731 for translated word data, a speech synthesis unit 732, and a display unit 735. Further, a speaker amplifier 733 and a speaker 734 are connected to the voice synthesizer 732.

The microphone 71 converts a user's voice or the like into an electric signal and inputs the electric signal.

[0007] The voice section cut-out section 721 converts the voice input from the microphone 71 and the voice signal containing noise into a digital signal, cuts out the voice section, and sends the voice section signal to the voice recognition processing section 722.

[0008] The voice recognition processing unit 722 receives an operation signal 79 from a keyboard or a switch, and receives an instruction from the control unit 74.
The extracted speech is analyzed using the acoustic model stored in the acoustic model storage unit 723. Further, the voice recognition processing unit 722 compares the analyzed result with the standard voice pattern stored in the voice recognition dictionary unit 724,
Perform voice recognition.

[0009] The acoustic model storage unit 723 stores an acoustic model of a cut-out speech section used for speech recognition.

[0010] The voice recognition dictionary unit 724 is composed of a RAM or the like, and stores a standard voice pattern corresponding to the utterance of the operator. This standard voice pattern is stored in advance by the operator.

On the other hand, the translated word data memory card 731 of the translation unit 73 is composed of a ROM card or the like, and stores translated words corresponding to the words recognized by speech. Stores audio data. Also, from this translated word data memory card 731, a character code corresponding to the translated word is read,
It is displayed on the display unit 735. By exchanging the translation word data memory card 731 with one for another language, it becomes possible to translate the word recognized by speech in a plurality of languages.

The voice synthesizing section 732 includes a voice recognition processing section 72.
2 is read from the translated word data memory card 731, converted into an audio signal, and output via the speaker amplifier 733 and the speaker 734.

The display unit 735 gives instructions to the user of the translation apparatus, displays translated words in characters, and the like.

The control section 74 is composed of a microprocessor or the like, and controls each section of the speech translator 7.

[0015] In the field of such speech recognition and speech synthesis techniques, their development is expected from the demand that the system should provide a more human-like user interface with the improvement of semiconductor technology. In the above-mentioned small information systems using the conventional speech recognition technology, as car navigation systems, portable information devices represented by PDAs, portable translators, and information home appliances with voice interfaces, the future will continue. It is expected to become more and more popular.

Therefore, the problems in practical use in the field using such a speech recognition technology are to improve the recognition rate and to shorten the recognition response time. In the related art, it is necessary to set a limit on the number of words to be recognized in order not to lower the performance of the recognition rate and the recognition response time. Under the constraints, for words and sentences registered in advance, the statistical characteristics of the speaker's voice in the character string and the characteristics of the voice actually spoken by the speaker are compared. The value closest to the probability is regarded as the recognition result. In particular, this method is indispensable for securing a recognition rate above a certain level in a noisy environment.

[0017] In the future, it is expected that the performance of the recognition rate and the recognition response time will be improved by the technical innovation in the speech recognition and the improvement of the software and hardware for realizing the speech recognition without limiting the number of words to be recognized. Can be However, from the practical viewpoint of the speech recognition system, it is preferable that the processing amount is as small as possible in terms of the single performance of the speech recognition at the recognition rate and the recognition response time, and the performance and usability of the whole system incorporating the speech recognition. Further, a problem for making the speech recognition an easy-to-use interface is to make the speech recognition a single user interface using speech.

For this purpose, the voice recognition processing is not performed by the system itself, but is realized by an interface such as a microphone. Thereby, the system main body and the voice recognition interface such as the microphone can be easily connected to the existing system. Further, since the result of the voice recognition processing is transferred to the system main body, the influence of noise from the environment can be reduced as compared with the case where a conventional analog voice signal is transferred to the system main body and then voice recognition is performed. Therefore, the recognition performance of the entire system can be improved. For that purpose, it is necessary to provide an easy-to-use system by separating the system main body and an interface unit such as a microphone for executing voice recognition.

In a speech recognition system in a conventional car navigation system, hundreds of thousands of dictionaries to be subjected to speech recognition such as place names, intersection names, building names, station names, and telephone numbers are prepared, and the dictionaries are hierarchically divided. To perform speech recognition hierarchically. It generates several voices until it reaches the target word, and repeats it until it recognizes the voice. When the target word is recognized, for example, a route search to the destination is performed.

In such a system, the number of words in the dictionary to be recognized is enormous, and hierarchies of the dictionary and speech recognition are executed hierarchically so as not to reduce the performance of the recognition rate and recognition response time. In this case, several times of speech recognition must be performed to recognize one target word. Therefore, an interface based on voice recognition, which should be convenient, is inconvenient, and the usability of the entire system is reduced.

Further, in a speech recognition system in which a dictionary is composed of hundreds of thousands of words and a target word is uttered from the beginning without making the dictionary hierarchical, even if the recognition rate is improved in the future due to the innovation of speech recognition technology, In particular, in reducing the price of the system, sufficient performance cannot be obtained in terms of recognition rate and recognition response time.

Furthermore, even if the dictionary is hierarchized and speech recognition is performed several times, and the basic speech recognition performance is obtained, the recognition rate does not reach 100%. It is the same as a human misunderstanding a phonetically similar word.

For example, a speech recognition system applied to a car navigation system will be described with reference to FIG. FIG.
FIG. 2 is a diagram illustrating a flow of a voice recognition process in a conventional voice recognition system in which a voice recognition process is performed on a car navigation system body side. In the conventional system, a series of processes related to voice recognition are performed by the car navigation system itself, so that not only the car navigation system but also other high-priority processes occur and the CPU load increases. . Also, in a system that does not originally have a voice interface, hardware modification is involved in addition to the load on the CPU.

Here, after the first occurrence "hotel",
The speech recognition processing when the second occurrence “@hotel” is issued will be described. In this system, a plurality of dictionaries are provided for each object. When the first utterance "hotel" is input (S1), the voice recognition process P1 performs a series of voice recognition processes, outputs a recognition result "hotel", and sends it to the application P2 (S2). The recognition result "hotel"
The application P2 starts the dictionary selection process P3 (S3), and selects the hotel word dictionary 531 in the large-scale recognition target dictionary 53 (S4).

The hotel word dictionary 531 is a dictionary composed of hotel name words, and it is assumed that the number of words is 5000 words. As the large-scale recognition dictionary 53, in addition to the hotel word dictionary, a park word dictionary 531 and a ski resort word dictionary 5
32, etc., and the number of words is 100, respectively.
Assume that there are 00 words and 4000 words. The selected hotel word dictionary 531 is used as a recognition target dictionary 534 in the subsequent speech recognition processing P1.

When the second or later utterance "@hotel" is input (S5), the speech recognition facility P1 performs a speech recognition process, and outputs the recognition result "@hotel" to the application P2.
(S6). The application P2 passes this data to processes such as destination setting and route search (S
7).

In such a system, the user must perform the exchange starting from the input of "hotel" as described above every time a destination is set. As mentioned earlier,
Such a method degrades the recognition performance in terms of the recognition response time and the recognition rate from the load on the CPU and the number of words in the dictionary, particularly in reducing the price of the system as a whole.

[0028]

SUMMARY OF THE INVENTION In view of the above, a first object of the present invention is to separate a speech recognition interface unit for executing speech recognition from a system body using a speech recognition result, and to a system body. An object of the present invention is to provide a voice recognition microphone that transfers only a recognition result. A second object of the present invention is to provide a speech recognition microphone that is easy to use even in terms of performance.

That is, for example, among a large number of dictionaries registered in the car navigation system, the number of words in a plurality of dictionaries frequently used by users is 100 words or less in total, such as commands and place names of destinations. Assume that Therefore, as for the words to be recognized by the voice recognition microphone, the user executes the voice recognition process only on the necessary words from the huge dictionary on the system body side, and only on the words registered in the voice recognition microphone side from the system body side. Therefore, the recognition rate is 100
% And the amount of processing is reduced, so that sufficient performance can be achieved even with low-cost hardware.

[0030]

In order to achieve the above object, according to the first aspect of the present invention, words and sentences to be subjected to speech recognition are collected and defined as a dictionary, and these words and sentences are defined based on the result of speech recognition. And a sentence, and output it as data for subsequent information processing, a character string display, an image indicated by a word, or the like, or a speech recognition system that outputs recognition results as speech using speech synthesis. In a voice recognition microphone, a microphone section having a function of a microphone alone, an A / D converter for converting an analog signal from the microphone section into a digital signal, a voice section detection process for detecting a voice section, and Voice analysis processing, and voice models that have voice characteristics in phoneme units, and pre-registered dictionaries and voice models are connected, and the voice models are connected. In the dictionary of Te, against the results voice analysis of the voice input, and configured to include a voice recognition processing section for outputting a probable recognition result, and outputs a speech recognition result from the microphone.

According to a second aspect of the present invention, in the voice recognition microphone according to the first aspect, a voice recognition unit that performs a series of voice recognition processes from inputting a voice to outputting a voice recognition result, and a new voice recognition unit using the voice recognition result. The system includes a data communication unit that transfers a recognition result to the system main unit that executes processing and transfers recognition target data from the system main unit, and a dictionary unit that includes a dictionary to be recognized.

According to a third aspect of the present invention, in the voice recognition microphone according to the second aspect, the data communication unit is configured as an interface for connecting the voice recognition microphone and the system body by wire, wireless, or infrared communication, and the data to be communicated is transmitted. The contents were digital data, and the contents were text data representing speech recognition results and / or digitized speech waveform data and / or S / N ratio data indicating the levels of noise N and speech S which are not speech.

According to a fourth aspect of the present invention, the voice recognition microphone includes an interface for connecting voice input notifying means for notifying the voice recognition microphone that a voice is to be input.

According to a fifth aspect of the present invention, in the above-mentioned speech recognition microphone, a command dictionary registered in advance in the speech recognition microphone and a dictionary to be recognized are read from a large-scale dictionary registered in a storage medium of the system body. It has a target dictionary and a user registration dictionary created by a user registration process.

According to a sixth aspect of the present invention, there is provided a speech recognition microphone having a speech recognition function and outputting a result of recognizing inputted speech, and an information processing means for executing subsequent processing using the speech recognition result. A recognition system was built.

According to a seventh aspect of the present invention, in the voice recognition system, the voice recognition microphone has a voice recognition processing unit, a dictionary unit, and a data communication unit, and the information processing unit uses a recognition target dictionary used for voice recognition. Having a data communication unit, transferring a part of the recognition target dictionary of the information processing means to a dictionary unit of the voice recognition microphone based on a voice recognition result of the voice recognition microphone, and using the transferred recognition target dictionary Recognized.

According to the present invention, in the speech recognition system, a user registration dictionary is provided in a dictionary section of the speech recognition microphone, and words required by a user in the dictionary to be recognized are registered in the user dictionary. In normal speech recognition, speech recognition is performed with the command dictionary and the user registration dictionary as recognition targets.

According to a ninth aspect of the present invention, there is provided a voice recognition method for a voice recognition system comprising a voice recognition microphone having a voice recognition processing unit, a dictionary unit and a data communication unit, and information processing means for performing processing based on the voice recognition result. In the dictionary part of the voice recognition microphone, a command dictionary, a recognition target dictionary transferred from the information processing means, and a user registration dictionary, and a plurality of words finally required by the user from the recognition target dictionary Then, a user registration dictionary is created by collecting the command dictionary and the normal dictionary recognizes the command dictionary and the user registration dictionary as voice recognition targets.

[0039]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments according to the present invention will be described below with reference to FIGS. FIG. 1 is a block diagram showing a functional configuration of a voice recognition microphone according to the present invention.

The voice recognition microphone 1 shown in FIG. 1 outputs a result 3 of voice recognition. Output 3 of this voice recognition microphone 1
Is not limited to the result of speech recognition, but may be information related to speech recognition. For example, it is a digital waveform signal obtained by collecting voices and surrounding sounds which are basic functions of an original microphone and digitizing and outputting a signal conventionally transmitted as an analog signal. In addition, it is S / N ratio data representing the level of noise N which is not voice and the level of voice S in a relative ratio. These digitized information is basic information necessary for an application when a new system is constructed by combining a voice recognition microphone and a system body.

The voice recognition microphone 1 is composed of functional blocks of a microphone 11, a voice recognition unit 13, a dictionary unit 15, and a data communication unit 17.

The microphone 11 captures voice and noise, and is configured by a conventional condenser microphone or the like, and has directivity.

The voice recognition unit 13 detects only voice from the input voice or noise and performs voice analysis. Further, the speech recognition unit 13 connects all the registered dictionaries and acoustic models from a pre-registered dictionary and an acoustic model having speech characteristics in phoneme units, and performs a speech analysis of the actually input speech. The result is compared with the result and a reliable recognition result is output.

The dictionary section 15 stores a dictionary to be subjected to speech recognition. The dictionary stored in the dictionary unit 15 includes a command dictionary registered in advance in the voice recognition microphone 1, a recognition target dictionary transferred from a large-scale recognition dictionary registered in a storage medium of the system main body, and a command dictionary. There is a user registration dictionary configured by extracting and registering only words required by a user from a dictionary or a dictionary to be recognized. Dictionary part 1
5 receives the speech recognition result and “register in user dictionary (register in dictionary)” registered in the command dictionary, and creates a user registration dictionary.

The data communication unit 17 performs a process for transferring the speech recognition result of the speech recognition unit to the system body to which the speech recognition microphone 1 is connected or to the system body using the speech recognition microphone 1. Further, similarly to the speech recognition result, a digitized speech signal and information related to speech are transferred.

The hardware configuration of the voice recognition microphone 1 will be described with reference to FIG. The voice recognition microphone 1 is the microphone 1
1, an amplifier 21, an A / D converter 22, and a CPU 23
, ROM 24, RAM 25, wired interface 26-1, and infrared interface (IR) 26-2.
, A wireless interface 26-3, a voice recognition mode interface 26-4, a system bus 27 for interconnecting them, and a voice input button 29.

The microphone 11 is the same as the microphone 11 shown in FIG.

The amplifier 21 is an amplifier composed of electronic components such as a resistor and a capacitor, and includes a high-pass filter for removing noise and a band-pass filter.

The A / D converter 22 converts a voice or noise analog signal input from the microphone 11 via the amplifier 21 into a digital signal. The A / D converter 22 is connected to the system bus 27.

The CPU 23 is a central processing unit or a CPU core which performs all processes of voice recognition, dictionary registration and data communication in the voice recognition microphone 1 by software.

The ROM 24 stores a program for executing all processes of initialization of the voice recognition microphone system, a series of voice recognition, dictionary registration, and data communication by software. ROM 24
Also, an acoustic model, a dictionary, a grammar, and the like necessary for speech recognition are written in the file.

The RAM 25 is a memory for transferring a program for executing all processes of a series of voice recognition, dictionary registration and data communication by software to a high-speed access memory and executing the program. This is a memory for securing a necessary work area during execution.
Further, it is a memory for preventing dictionaries and the like registered by the user from being erased even when the power is turned off.

The wired interface 26-1 is an interface for exchanging information (data 31) between the voice recognition microphone 1 and the system body connected thereto. The system itself is connected by wire, and the data bit width is
It may be serial or parallel.

The data 31 is data that is bidirectionally transferred between the voice recognition microphone 1 and the system body via a wire. The first is a recognition result output from the voice recognition microphone 1. The recognition result may be character information of text data or coded data. Second,
Dictionary data to be recognized. The dictionary data is transferred from the system body to the voice recognition microphone 1. Third,
This is S / N ratio data indicating the levels of noise N and voice S that are not voice. Fourth is information for notifying that the system is in the voice recognition mode. For example, in a car navigation system on the system body side,
This is the case where the utterance button of the remote control is pressed as the voice recognition mode. Therefore, the voice recognition microphone 1 performs a voice recognition process on the input voice.

The infrared interface 26-2 is an interface for exchanging information (data 32) between the voice recognition microphone 1 and the connected system body.
The system itself is wirelessly interfaced using infrared IR, and the data bit width may be serial or parallel.

The data 32 is data that is bidirectionally transferred between the voice recognition microphone 1 and the system main body via the infrared communication system. The contents are the same as the data 31.

The wireless interface 26-3 is an interface for exchanging information (data 33) between the voice recognition microphone 1 and the connected system main body. The system itself is wirelessly interfaced using a wireless LAN or the like, and the data bit width may be serial or parallel.

The data 33 is data that is bidirectionally transferred between the voice recognition microphone 1 and the system body via radio. The contents are the same as the data 31.

Here, the wired interface 26-1, the infrared interface 26-2, and the wireless interface 26-3 for interfacing with the system main unit can be used even if the voice recognition microphone 1 has only one interface. Good, or all interfaces may be provided.

Voice recognition mode interface 26-4
Is an interface for directly informing the voice recognition microphone 1 that the system is in the voice recognition mode in the information for reporting that the system body is in the voice recognition mode. Voice recognition mode interface 2
6-4, a voice input button 2 for notifying a voice input state;
9 is connected. The meaning of directly telling the voice recognition microphone 1 that the apparatus is in the voice recognition mode is, for example, that the system body receives the recognition results 31 to 33 from the voice recognition microphone 1 in one direction, and the system body This is effective when performing another process or work. When such a method is required, it is conceivable that the method is applied to an audio remote controller.

The voice recognition microphone 1 includes a plurality of LSIs and ICs.
Or may be formed on one semiconductor element such as an ASIC. If it is an ASIC, the CPU 23
It is configured as a PU core.

Referring to FIG. 3, a system configuration in which the voice recognition microphone 1 and the system main body 5 operating using the results of voice recognition are connected wirelessly or by wire, and voice recognition is provided as an interface, and the flow of processing will be described. I do. For example, a car navigation system is an example of the system in the present application.

This system is configured by connecting the voice recognition microphone 1 and the system main body 5 by a communication method such as wireless or wired 4.

For example, in a car navigation system, the voice recognition microphone 1 is installed or built in a sun visor, a seat belt, a steering column, a pillar or a steering wheel in a vehicle. Or voice recognition microphone 1
Is built into the remote control that controls the system.

The car navigation system body 5 is integrated with an audio system and an air conditioning system, and is mounted in the vehicle including a display.

The voice recognition microphone 1 includes a microphone 11, a voice recognition unit 13, a dictionary unit 15, a data communication unit 17, and a voice input button 29.

The speech recognition unit 13 includes an A / D converter 131
Voice section cutout section 132 and voice analysis processing section 133
, A voice recognition processing unit 134, and an acoustic model storage unit 135
And is configured.

The dictionary section 15 includes a command dictionary 151, a recognition target dictionary 152, a dictionary registration processing section 153, and a user registration dictionary 154.

The data communication unit 17 includes the voice recognition processing unit 13
4 is a data communication unit that performs a process of transferring a result recognized by 4 to the system main body 5. Alternatively, it receives dictionary data to be recognized transferred from the system main body 5 and information for notifying that it has entered the voice recognition mode.

The system body 5 is configured as, for example, a car navigation system. In this case, the system body 5 includes a data communication unit 51, a dictionary reading unit 52, a large-scale dictionary 53 to be recognized, and application software 54.
, A voice synthesizing unit 55 and a display unit 56.

The A / D converter 131 converts an analog signal output from the microphone 11 into a digital signal, and outputs the digital signal to the voice section extracting unit 133.

The voice section cutout section 132 cuts out a voice from an input signal containing a voice or noise converted into a digital signal and outputs the voice to the voice analysis processing section 133.

The voice analysis processing unit 133 analyzes the voice and outputs the analysis result to the voice recognition processing unit 134.

The speech recognition processing unit 134 is a speech recognition unit that stochastically performs a series of speech collation processing using the speech analysis result, the acoustic model, and the dictionary, and outputs a reliable recognition result.

The acoustic model storage section 135 stores acoustic models necessary for speech recognition. In a general speech recognition system that is being put into practical use, a so-called “unspecified speaker correspondence” in which anyone can recognize the voice without registering the voice in advance has become mainstream. As an acoustic model used for such speech recognition, for example, a Hidden Markov Model (HMM) can be used.

The command dictionary 151 is a dictionary registered in the speech recognition microphone 1 in advance, and describes words such as a genre for executing speech recognition processing and instructions for speech recognition processing.

The recognition target dictionary 152 is a dictionary to be subjected to voice recognition, and is transferred from the system body 5 as needed.

The dictionary registration processing section 153 allows the user to input a “dictionary” of the command dictionary 151 for words that the user wants to always register in the voice recognition microphone 1 from among the recognition target dictionaries 152 to be recognized. This is a processing unit that performs registration when a “registration to dictionary” is output from the speech recognition processing unit 134 as a result of speech input of “register to”.

The user registration dictionary 154 is a dictionary for registering words to be registered in the voice recognition microphone 1 at all times.

The configuration of the dictionary shown in FIG. 3 will be described with reference to FIG. The command dictionary 151 registered in advance in the voice recognition microphone 1 is configured as a dictionary composed of, for example, 300 words, and the data D shown in FIG.
151 as an example. In this example, command dictionary 1
Reference numeral 51 denotes "hotel", "park", "ski resort", "golf course" indicating a recognition target genre, "register in dictionary", "delete from dictionary", "contents of dictionary" indicating processing instructions. It is composed of such words.

The dictionary for speech recognition 152 is a dictionary to be subjected to speech recognition, and is transferred from the large-scale dictionary for recognition 53 of the system body 5 as necessary. The voice recognition microphone 1 performs voice recognition processing on the input voice only in the range of the words in the command dictionary 151 and the recognition target dictionary 152.

The recognition target dictionary 152 has a word count of about 5000 words, and is configured as a dictionary including words corresponding to commands specifying the genre of the command dictionary. For example, in response to the command “hotel” for specifying a genre, as shown as data D152 in FIG. 4B, “OO Hotel”, “△△△ hotel”, “hotel □□□□” , “@Ryokan”: Only words representing hotel names such as “xxxxxx hotel” and “☆☆☆ hotel” are registered.

The user registration dictionary data D154 shows an example of specific contents of the user registration dictionary 154. For example, the user registration dictionary 154 is composed of words that the user wants to register, and the number of words constituting the user registration dictionary 154 is as follows.
Assume about 100 words.

The user registration dictionary 154 is configured as a dictionary having about 100 words. The user registration dictionary 154 is configured as a dictionary in which words that the user wants to register are extracted from the command dictionary 151 and the recognition target dictionary 152 and registered, and “@” illustrated as data D154 in FIG. Words such as a place and a processing instruction such as "hotel", "Tokyo xxx land", ... "return to home" are registered.

Next, the processing on the system body 5 side will be described. The data communication unit 51 of the system main body 5 transmits the result recognized by the voice recognition microphone 1 to the system main body 5.
Is a data communication unit for receiving the data. Alternatively, the system main body 5 transfers to the voice recognition microphone 1 dictionary data to be recognized and information for notifying that the voice recognition mode has been entered.

The dictionary reading unit 52 is a CD-ROM or DVD
A dictionary classified as a dictionary to be recognized is read out of the large-scale dictionary for recognition 53 stored in a storage medium having a large storage capacity represented by, and transferred from the data communication unit 51 to the voice recognition microphone 1. I do.

In the large-scale dictionary 53 to be recognized, the dictionary to be recognized is classified for each item and stored in a CD-ROM or a DVD. For example, the hotel word dictionary 531 is a dictionary composed of hotel name words, and the number of words is 500.
Assume that there are 0 words. Park word dictionary 532 and ski resort word dictionary 5 in addition to the hotel word dictionary as recognition target dictionaries
Various recognition target genres such as 33 are stored.
The number of words is assumed to be 10,000 words and 4000 words, respectively.

The application software 54 is application software for performing main processing of the car navigation system, and performs processing of the whole system such as GPS processing, navigation processing, and voice interface.

The display section 56 is composed of a liquid crystal display (LCD) or the like, and displays a map, a progress status, information on a voice recognition result, and the like.

The voice synthesizing unit 55 synthesizes and processes the progress status, the callback of the voice recognition result, voice guidance, and the like.

The speech recognition processing in the speech recognition microphone 1 and the system main unit 5 will be described with reference to FIG. The voice recognition system of the present invention performs voice recognition processing on the voice recognition microphone 1 side with respect to the conventional voice recognition system shown in FIG. 7, so that the processing in the system body 5 is not limited to the car navigation system. The system side CPU
Load does not change. Also, even in a system that does not originally have a voice interface, the CP
The load on U does not change, and the hardware modification requires only minor changes. In a system having a communication interface, voice operation can be realized with existing hardware without a voice interface.

In FIG. 5, the upper part from the broken line shows the dictionary registration processing and the voice recognition processing in the voice recognition microphone 1 and the dictionary used for this processing, and the lower part from the broken line shows the use of the voice recognition result in the system main unit 5 and the dictionary. I have. The voice recognition microphone 1 performs a series of voice recognition. In this process, a user registration dictionary creation process performed before dictionary registration and a voice recognition process using the user registration dictionary are performed.

First, when the first occurrence "hotel" is input before registration (S11), the voice recognition process P11 is executed using the command dictionary 151, and the voice recognition result "hotel" is output (S12). . The recognition result “hotel” is transferred to the system main body 5 side, and the application 54 of the system main body side activates the dictionary selection process P52 (S13), and the hotel word in the recognition target large-scale dictionary 53 covering the recognition targets. Select the dictionary 531.

The selected hotel word dictionary 531 is transferred from the system main body 5 to the voice recognition microphone 1, and stored in the recognition target dictionary 152 (S14).

When the second or later utterance "@hotel" is input (S15), speech recognition processing P11 is performed using the hotel dictionary data D531 stored in the recognition target dictionary 152, and the recognition result "@hotel" △ Hotel ”(S1
6). The recognition result “@hotel” is transferred to the system body side, and the application 54 of the system body 5
The processing is passed to processing P57 such as destination setting and route search (S17).

When the third utterance “register in dictionary” is input (S 18), the voice recognition process P 11 executes the command dictionary 1
The speech recognition process is performed using the command 51, and the recognition result "register in dictionary" is output (S19). Upon receiving the recognition result “register in dictionary”, dictionary registration processing 153 registers “@hotel” in user registration dictionary 154 on the voice recognition microphone side (S20). The number of words in the user registration dictionary 154 is 100
Words.

After registration, the user suddenly utters "@hotel" with the destination as the first utterance (S2).
1), voice recognition processing P11 is performed using the user registration dictionary data D154 stored in the user registration dictionary 152, and a recognition result "@hotel" is output (S16).
The recognition result “@hotel” is transferred to the system main body side, and is passed to processing P57 such as destination setting and route search by the application 54 of the system main body 5 (S1).
7).

As described above, according to the present invention, after selecting a necessary word and registering it in the user dictionary 154, the user immediately executes "@hotel" to perform voice recognition. Further, as described above, since the number of words in the user registration dictionary 154 is as small as 100 words, good recognition performance can be achieved in terms of recognition response time and recognition rate, and the speech recognition interface can be improved. Can be expected.

Further, in this speech recognition system, the load on the system itself in the speech recognition processing is greatly reduced, so that an existing interface can be received simply by adding an interface capable of receiving a recognition result from a speech recognition microphone. Can be applied.

[0100]

According to the present invention, in a car navigation system, a small information system, and a voice recognition system used in a game, a threshold for voice section detection is set in accordance with a noise level in an environment actually used. Can be realized, and a good speech recognition system can be realized in which speech section detection by automatic threshold setting and recognition performance do not deteriorate even in a real environment.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an outline of functions of a voice recognition microphone according to the present invention.

FIG. 2 is a block diagram showing a hardware configuration of a voice recognition microphone according to the present invention.

FIG. 3 is a block diagram illustrating an outline of a speech recognition system according to the present invention.

FIG. 4 is a view for explaining a dictionary configuration in the speech recognition system according to the present invention.

FIG. 5 is a view for explaining the operation of the speech recognition system according to the present invention.

FIG. 6 is a block diagram illustrating a configuration of a portable translation device using a conventional speech recognition system.

FIG. 7 is a diagram for explaining the operation of a conventional speech recognition system.

[Explanation of symbols]

 1 Voice Recognition Microphone 11 Voice Recognition Microphone 13 Voice Recognition Unit 131 A / D Converter 132 Voice Section Extraction Unit 133 Voice Analysis Processing Unit 134 Voice Recognition Processing Unit 135 Acoustic Model Storage Unit 15 Dictionary Unit 151 Command Dictionary 152 Recognition Dictionary 153 Dictionary registration processing unit 154 User registration dictionary 17 Data communication unit 21 Amplifier 22 A / D converter 23 CPU 24 ROM 25 RAM 26 Interface 29 Voice input button 3 Recognition result 4 Communication means 5 System body 51 Data communication unit 52 Dictionary reading unit 53 Large-scale dictionary for recognition 531 Hotel word dictionary 532 Park word dictionary 533 Ski area word dictionary 54 Application software 55 Speech synthesis unit 56 Display unit

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G10L 3/00 571A (72) Inventor Hiroaki Kokubo 1-280 Higashi-Koigabo, Kokubunji-shi, Tokyo Hitachi, Ltd. (72) Inventor Nobuo Hataoka 1-280 Higashi Koikekubo, Kokubunji-shi, Tokyo F-term in Central Research Laboratory, Hitachi, Ltd.F-term (reference)

Claims (9)

[Claims] 1. Words and sentences to be subjected to speech recognition are collected and defined as a dictionary, and those words and sentences are extracted based on the speech recognition result. A voice recognition microphone having a voice recognition function in a voice recognition system that outputs a recognized image or the like, or outputs a recognition result as voice using voice synthesis. A / D to convert analog signal to digital signal
A converter, a voice section detection process for detecting a voice section, a voice analysis process for performing voice analysis on the captured voice, a voice model having voice characteristics in phoneme units, a dictionary and a voice model registered in advance. In all of the dictionaries linked to the acoustic model, the dictionary is provided with a voice recognition processing unit that compares the input voice with the voice analysis result and outputs a reliable recognition result, and outputs the voice recognition result from the microphone. A voice recognition microphone characterized by the following. 2. A speech recognition unit for performing a series of speech recognition processes from inputting speech to outputting a speech recognition result, and transferring the recognition result to a system main unit for executing a new process using the speech recognition result. The voice recognition microphone according to claim 1, further comprising a data communication unit that transfers recognition target data from the main body, and a dictionary unit that includes a dictionary to be recognized. 3. The data communication unit is an interface for connecting a voice recognition microphone and the system body by wire, wireless, or infrared communication, and data to be communicated is digital data, and the content of the data is text representing a voice recognition result. 3. An S / N ratio data indicating data and / or digitized audio waveform data and / or non-audio noise N and audio S levels.
Voice recognition microphone described in. 4. An interface for connecting a voice input notifying means for notifying a voice recognition microphone that a voice is to be input from now on.
A voice recognition microphone according to claim 3. 5. A command dictionary registered in advance in a voice recognition microphone, a recognition target dictionary in which a dictionary to be recognized is read from a large-scale dictionary registered in a storage medium of a system main body, and a user registration process. The voice recognition microphone according to any one of claims 1 to 4, further comprising a user registration dictionary created. 6. A speech recognition system comprising a speech recognition microphone having a speech recognition function and outputting a result of recognizing input speech, and an information processing means for executing subsequent processing using the speech recognition result. 7. The speech recognition microphone according to claim 1, wherein the speech recognition microphone has a speech recognition processing unit, a dictionary unit, and a data communication unit, and the information processing unit has a recognition target dictionary and a data communication unit used for speech recognition. 7. A part of the recognition target dictionary of the information processing means is transferred to a dictionary part of a voice recognition microphone based on the voice recognition result in step (a), and voice recognition is performed using the transferred recognition target dictionary. A speech recognition system as described. 8. A user registration dictionary is provided in a dictionary section of the voice recognition microphone, and words required by a user in the dictionary to be recognized are registered in the user dictionary. In a normal voice recognition, a command dictionary and a user registration are registered. The speech recognition system according to claim 6, wherein speech recognition is performed with a dictionary as a recognition target. 9. A speech recognition method for a speech recognition system, comprising: a speech recognition microphone having a speech recognition processing unit, a dictionary unit, and a data communication unit; and information processing means for performing processing based on the speech recognition result. The dictionary section includes a command dictionary, a recognition target dictionary transferred from the information processing means, and a user registration dictionary. A plurality of words finally required by the user are collected from the recognition target dictionary for user registration. A speech recognition method for a speech recognition system, comprising: creating a dictionary; and performing speech recognition using a command dictionary and a user registration dictionary as recognition targets in normal speech recognition.