JP3700533B2 - Speech recognition apparatus and processing system - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a voice recognition device that is effective when, for example, a destination setting or the like in a navigation system is input by voice, and a processing system including the voice recognition device.
[0002]
[Prior art and problems to be solved by the invention]
Conventionally, a speech recognition apparatus that compares input speech with a plurality of comparison target pattern candidates stored in advance and uses a recognition result having a high degree of coincidence has already been put into practical use. This is used for a user to input a place name by voice. In particular, when the driver himself / herself uses a navigation system for a vehicle, since it is not accompanied by button operation or screen gaze if it is a voice input, it is effective because it is highly safe even when the vehicle is running.
[0003]
In order to satisfy such a function, it is desired that a sufficiently detailed point can be easily specified. Specifically, it is preferable to be able to input not the prefecture or city level but the level of the town name under the city and the level of large or small letters in the town and village. Furthermore, if the user wants to set “Showacho, Kariya City, Aichi Prefecture”, for example, he / she must divide it into levels called “Aichi Prefecture”, “Kariya City”, “Showamachi”, etc. Then, since it is troublesome, it is preferable to enable continuous input (batch input).
[0004]
However, it is necessary to increase the number of recognizable words and increase the amount of data in the dictionary in order to make it possible to specify more detailed points on the premise that collective input is possible in this way. For example, when it comes to address recognition, it is common to have dictionaries only up to the level of large letters. It is possible to specify more detailed points by preparing a dictionary up to small letters, but as the dictionary data increases, the memory such as RAM for storing the dictionary data becomes very large, It will also lead to cost increase.
[0005]
The present invention suppresses the disadvantages that may occur when pursuing the advantage of increasing the number of recognizable objects by increasing the number of words prepared in the dictionary in such a speech recognition technology, and enjoys the above advantages more realistically. It aims at proposing the technology which makes it easy to do.
[0006]
[Means for Solving the Problems and Effects of the Invention]
First, it is described in claim 1 Prefecture name Template data, By prefecture The template data, the first storage unit, and the second storage unit will be described.
Prefecture name Template data has multiple final recognition targets Place name Are connected hierarchically Street address If Name of prefectures This is data for comparison with matching data obtained based on input speech. Also, By prefecture Template data is By each prefecture Be prepared, Name of prefectures In addition to City name or even lower level place name A word group including up to is stored. For example In the case of Japan Prefecture name Template data But Have 47 prefecture names as template data By the prefecture divided into 47 Template data But Will be prepared.
[0007]
Prefecture name Or By prefecture The template data may be dictionary data (upper layer dictionary or lower layer dictionary) as shown in claim 2, or may be voice data as shown in claim 3. For example, a person may pronounce a word or a group of words in advance, input it, and store it as voice data. In other words, data that can be recognized by matching with some data based on voice data uttered by the user is not limited to recognition using a dictionary.
[0008]
Further, the first storage means has a relatively low high-speed accessibility during the speech recognition process, and the second storage means has a relatively high high-speed accessibility during the speech recognition process. Examples of the first storage means include a DVD and a CD-ROM, and examples of the second storage means include a RAM. That is, in executing the actual speech recognition process, the dictionary is usually read into the second storage means such as a RAM from the viewpoint of shortening the processing time and improving the response.
[0009]
According to the speech recognition apparatus of claim 1, at least By prefecture The template data is stored in the first storage means, and when recognizing the voice input through the voice input means that can be input continuously, first, matching data obtained based on the input voice and Prefecture name By comparing with template data But It is determined whether or not the prefecture name is included. And it was supposed to be included in the preliminary judgment Name of prefectures Corresponding to By prefecture Read the template data into the second storage means, By prefecture The final recognition result is obtained using the template data. In other words, for example, if the preliminary judgment shows that the prefecture name “Aichi Prefecture” is included, To use Prepared By prefecture Only the template data can be read into the second storage means for recognition.
[0010]
In this way, it is not necessary to read all template data into the second storage means. That is , Prepared by prefecture By prefecture It is only necessary to read at least one template data (a plurality of template data in some cases), and it is sufficient that the second storage unit has a storage capacity corresponding to the template data. In other words, the present invention Address where multiple place names are connected hierarchically Against Prefectures To the unit By prefecture In other words, template data is prepared, that is, “divide” the template data, By prefecture It was determined whether template data should be used, and after narrowing down to “really necessary” template data, it was read into the second storage means. Therefore, even when pursuing the advantage of increasing the number of recognizable objects by increasing the number of words prepared in the template data, it is not necessary to store all the recognizable vocabulary in the second storage means. Therefore, even if the second storage means has a relatively small capacity, it is possible to realize appropriate speech recognition corresponding to batch input.
[0011]
In addition, Address where multiple place names are connected hierarchically Because it is possible to flexibly apply the upper hierarchy and lower classification for , Contract It may be as shown in claim 4. That means Street address Is more than 3 levels With the place name If configured, Prefecture name Template data, By prefecture Template data and By prefecture Considering template data as a higher hierarchy By city prepared for each municipality, so that the place name level lower than the municipality is also distinguished. Template data The To prepare . This By doing so, it is only necessary to read the necessary template data, and the second storage means can cope with a smaller capacity.
[0012]
By the way, finally By prefecture Template data Or by city in claim 4 To recognize using template data, By prefecture Template data Or by city in claim 4 Preliminary determination for selecting template data is performed. This preliminary judgment is Prefecture name Template data Or by prefecture in claim 4 Template day T The template data may be configured as follows. That is, as shown in claim 5 , Te A wild card model that can be matched with a speech input in which other words or word groups are attached after a plurality of types of words or word groups constituting template data.
[0013]
For example, Prefecture name Template data of In this case, match any sound after the prefecture name. If only the prefecture name is stored as dictionary data, since only a part of the actual recognition target (a group of words including municipalities and capital letters below the prefecture) has the prefecture name, the matching degree as a whole decreases. . On the other hand, in the case of the wild card model, since the matching itself can be performed as the entire recognition target, such a problem does not occur. Examples of the wild card model include a garbage model and a syllable connection model described later.
[0014]
On the other hand, instead of using such a wild card model, as shown in claim 6, Prefecture name Template data Is the name of the municipality or a lower-level place after the prefecture name Is attached The template data by prefecture in claim 4 is the name of the place at a lower level than the municipality after the municipality name. And redundant template data may be used.
[0015]
The use of a wild card has the advantages as described above. However, since this wild card matches anything gently, it also increases the possibility of erroneous recognition. So, for example Prefecture name Redundant state with template data including not only the prefecture name but also the city name In Prepare template data. At the time of recognition, matching is performed up to the municipality name, but as a result, it is determined which prefecture name is matched. Matching with longer voices, and since it does not match anything gently like a wild card model, an improvement in recognition rate can be expected. However, the amount of template data to be prepared increases relatively.
[0016]
By the way, such a speech recognition device and a processing device that executes a predetermined process based on a result recognized by the speech recognition device, a predetermined that needs to be specified when the processing device performs processing. It is possible to construct a processing system that allows the user to input the above command by voice. At this time, as shown in claim 7, command template data, which is template data for recognizing a command, is read from the first storage means into the second storage means, or similar to the second storage means. Further, it is stored in advance in a third read-only storage means (for example, a ROM) having relatively high high-speed accessibility. And the next (1)-(4) Recognize according to the procedure.
[0017]
(1) Prefecture name template data or by prefecture in claim 4 Preliminary determination is performed by performing recognition using template data.
(2) Recognize using command template data and respond to preliminary judgment results in parallel with this recognition Template data by prefecture or city by claim 4 The template data is read into the second storage means.
[0018]
(3) That read Template data by prefecture or city by claim 4 Recognition is performed using template data.
(4) the above (2) Recognition result and (3) Of these recognition results, the one with higher accuracy is set as the final recognition result.
For example, if the processing device is a navigation device, a place name (address) having a hierarchical configuration may be input by voice for setting the destination, etc., and of course, commands for using various functions of the navigation device May be instructed. If this navigation system is assumed, by executing the recognition process described above, not only the place name (address) but also the command is input, the command template data It is not necessary to separately perform recognition processing using. That is, the response is improved and the usability for the user is improved.
[0019]
The processing system described in claim 7 can naturally be applied to systems other than the navigation system, but the following contrivances can also be made especially when limited to the navigation system. That is, as shown in claim 8, on the premise of having the function of detecting the present location, (1)-(4) Recognize according to the procedure.
[0020]
(1) Prefecture name template data or by prefecture in claim 4 Preliminary determination is performed by performing recognition using template data.
(2) Corresponds to the current location detected by the current location detection means Template data by prefecture or city by claim 4 Read the template data into the second storage means, Template data by prefecture or city by claim 4 Recognize using template data and handle the result of preliminary judgment in parallel with this recognition Template data by prefecture or city by claim 4 The template data is read into the second storage means.
[0021]
(3) That read Template data by prefecture or city by claim 4 Recognition is performed using template data.
(4) the above (2) Recognition result and (3) Of these recognition results, the one with higher accuracy is set as the final recognition result.
The situation to be solved by this method is as follows. In other words, if a vehicle equipped with a navigation system is traveling in Aichi Prefecture, for example, and the same Aichi Prefecture is set as “Showa Town, Kariya City, Aichi Prefecture”, the destination is “Showa Town, Kariya City, Aichi Prefecture” It is more natural not to input voice, but to input “Akari prefecture” and “Kariya city Showa-cho”. In this method, since the recognition is performed using the lower hierarchy dictionary corresponding to the current location in the second recognition, even speech input in which the prefecture name is omitted can be handled.
[0022]
On the other hand, if it is desired to relatively improve the recognition speed on the premise of having the function of detecting the current location in the same manner, it may be as shown in claim 9. In this case, it corresponds to the current location detected by the current location detection means prior to the recognition process. Template data by prefecture or city by claim 4 Template data is read in advance in the second storage means. And the next (1), (2) Recognize according to the procedure.
[0023]
(1) Prefecture name template data or by prefecture in claim 4 Template data and pre-loaded Template data by prefecture or city by claim 4 Perform recognition using template data. And the recognition result was read in advance Template data by prefecture or city by claim 4 If it is obtained using template data, the recognition process is terminated with the final recognition result.
[0024]
(2) On the other hand, the above (1) The recognition result of Prefecture name template data or by prefecture in claim 4 If template data is used, it corresponds to the recognition result. Template data by prefecture or city by claim 4 Read the template data into the second storage means, Template data by prefecture or city by claim 4 The recognition result obtained using the template data is used as the final recognition result.
[0025]
In this way, when recognizing place names in a given area including the current place that is considered to be frequently used, Template data by prefecture or city by claim 4 Since the template data is read in advance, the recognition process can be performed relatively quickly.
[0026]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments to which the present invention is applied will be described below with reference to the drawings. Needless to say, the embodiments of the present invention are not limited to the following examples, and can take various forms as long as they belong to the technical scope of the present invention.
[0027]
FIG. 1 is a block diagram showing a schematic configuration of a navigation system 2 having a voice recognition function. The navigation system 2 is a so-called car navigation system used by being mounted on a vehicle. The navigation system 2 is connected to a position detector 4, a data input device 6, an operation switch group 8, a control circuit 10 connected thereto, and a control circuit 10. And an external memory 12, a display device 14, a remote control sensor 15, and a voice recognition device 30. The control circuit 10 is configured as a normal computer, and includes a well-known CPU, ROM, RAM, I / O, and a bus line for connecting these configurations.
[0028]
The position detector 4 has a known geomagnetic sensor 16, a gyroscope 18, a distance sensor 20, and a GPS receiver 22 for detecting the position of the vehicle based on radio waves from a satellite. Since these sensors 16, 18, 20, and 22 have errors of different properties, they are configured to be used while being interpolated by a plurality of sensors. Depending on the accuracy, a part of the above may be used, and further, a steering rotation sensor, a wheel sensor of each rolling wheel, or the like may be used.
[0029]
In addition to so-called map matching data for improving the accuracy of position detection, various data for navigation including map data and landmark data, the data input device 6 uses dictionary data used when the speech recognition device 30 performs recognition processing. Is a device for inputting. As a storage medium, it is considered that a DVD is generally used because of its data amount, but another medium such as a CD-ROM may be used. When a DVD is used as the data storage medium, the data input device 6 is a DVD player.
[0030]
The display device 14 is a color display device. On the screen of the display device 14, the vehicle current position mark input from the position detector 4, the map data input from the map data input device 6, and further displayed on the map. Additional data such as guidance routes to be set and landmarks for setting points can be displayed in an overlapping manner. In addition, when a menu screen that displays a plurality of options, or when an option is selected, a command input screen that displays a plurality of options can be displayed.
[0031]
In addition, when the navigation system 2 inputs the position of the destination from the remote control sensor 15 or the operation switch group 8 via a remote control terminal (hereinafter referred to as a remote controller) 15a, the navigation system 2 can move from the current position to the destination. A so-called route guidance function is also provided which automatically selects an optimum route to form and display a guidance route. As a method for automatically setting an optimal route, a method such as the Dijkstra method is known. For example, a touch switch or a mechanical switch integrated with the display device 14 is used as the operation switch group 8 and is used for inputting various commands.
[0032]
The voice recognition device 30 can input various commands in the same manner when the user inputs voices, while the operation switch group 8 or the remote controller 15a is used for inputting various commands by manual operation. It is an apparatus for doing so.
[0033]
The voice recognition device 30 includes a voice recognition unit 31, a dialogue control unit 32, a voice synthesis unit 33, a voice extraction unit 34, a microphone 35, a switch 36, a speaker 37, and a control unit 38. Yes.
The voice recognition unit 31 performs input voice recognition processing on the voice data input from the voice extraction unit 34 according to an instruction from the dialogue control unit 32, and returns the recognition result to the dialogue control unit 32. That is, the speech data acquired from the speech extraction unit 34 is collated using the stored dictionary data, and the upper comparison target pattern having a higher degree of coincidence than a plurality of comparison target pattern candidates is displayed in the dialog control unit 32. Output to. The recognition of the word series in the input speech is performed by sequentially analyzing the speech data input from the speech extraction unit 34 to extract an acoustic feature amount (for example, cepstrum), and at the time of the acoustic feature amount obtained by this acoustic analysis. Get series data. Then, the time series data is divided into several sections by a known HMM (Hidden Markov Model), DP matching method or neural network, and it is determined which word is stored as dictionary data. .
[0034]
Based on the recognition result in the voice recognition unit 31 and the instruction from the control unit 38, the dialogue control unit 32 instructs the control circuit 10 to execute a response voice output instruction to the voice synthesis unit 33 or the processing of the navigation system itself. For example, a destination or command necessary for the navigation process is notified, and processing for instructing to set the destination or execute the command is performed. As a result of such processing, if the voice recognition device 30 is used, a destination can be instructed to the navigation system by voice input without manually operating the operation switch group 8 or the remote controller 15a. .
[0035]
The voice synthesizer 33 synthesizes a voice based on a response voice output instruction from the dialogue control unit 32 using a voice waveform stored in the waveform database. This synthesized voice is output from the speaker 37.
The voice extraction unit 34 converts the surrounding voice captured by the microphone 35 into digital data and outputs the digital data to the voice recognition unit 31. Specifically, in order to analyze the feature amount of the input voice, for example, a frame signal of a section of about several tens of milliseconds is cut out at a constant interval, and whether the input signal is a voice section including the voice is included. Determine if there is no noise interval. Since the signal input from the microphone 35 includes not only the speech to be recognized but also noise, the speech section and the noise section are determined. Many methods have been proposed as this determination method. For example, the short-time power of the input signal is extracted at regular intervals, and depending on whether or not the short-time power equal to or greater than a predetermined threshold continues for a certain period. A method of determining whether a speech section or a noise section is often used. Then, when it is determined that it is a voice section, the input signal is output to the voice recognition unit 31.
[0036]
In the present embodiment, the user inputs voice through the microphone 35 while pressing the switch 36. Specifically, the control unit 38 monitors the timing at which the switch 36 is pressed, the timing at which the switch 36 is returned, and the time during which the pressed state continues, and if the switch 36 is pressed, The voice recognition unit 31 is instructed to execute processing. On the other hand, when the switch 36 is not pressed, the processing is not executed. Accordingly, voice data input via the microphone 35 while the switch 36 is being pressed is output to the voice recognition unit 31.
[0037]
By having such a configuration, in the in-vehicle navigation system 2 of the present embodiment, various processes such as route setting, route guidance, facility search, and facility display can be executed by the user inputting a command.
Here, the voice recognition unit 31 and the dialogue control unit 32 will be further described. As shown in FIG. 2, the voice recognition unit 31 includes a collation unit 311, a dictionary unit 312, and an extraction result storage unit 313, and the dialogue control unit 32 includes a processing unit 321, an input unit 322, a dictionary control unit 323, and the like. have.
[0038]
In the speech recognition unit 31, the extraction result storage unit 313 stores the extraction result output from the speech extraction unit 34, and the collation unit 31a stores the extracted result in the dictionary unit 312. Collation is performed using existing dictionary data. The dictionary data in the dictionary unit 312 is not fixed, but is set and updated as appropriate. This point will be described later. Then, the higher-level recognition result compared with the dictionary data by the collating unit 311 and having a high degree of coincidence is output to the processing unit 321 of the dialogue control unit 32, and the processing unit 321 of the dialogue control unit 32 performs the control circuit. The recognition result is output to 10.
[0039]
On the other hand, the processing unit 321 can issue a request (dictionary read request) to read out the dictionary data from the DVD and output it to the voice recognition device 30 side to the control circuit 10. The dictionary data thus input is input via the input unit 322 of the dialog control unit 32. Then, the dictionary control unit 323 sets (writes) / updates the dictionary data in the dictionary unit 312 of the voice recognition unit 31.
[0040]
Here, the dictionary data will be described. As the dictionary data, not only the data of the vocabulary itself but also the vocabulary is prepared by dividing a plurality of words in a hierarchical manner as follows. Here, a place name dictionary will be described as an example of dictionary data that is divided and prepared in this way.
[0041]
First, the upper hierarchy dictionary is dictionary data of prefecture names. That is, it has keywords corresponding to the names of 47 prefectures (Aichi Prefecture, Aomori Prefecture,..., Wakayama Prefecture) as dictionary data. The lower hierarchy dictionary is a prefecture-specific dictionary prepared by dividing into prefectures. That is, 47 prefecture-specific dictionaries are prepared, such as Aichi prefecture-specific dictionaries, Aomori prefecture-specific dictionaries, ... Wakayama prefecture-specific dictionaries. This lower-level dictionary is dictionary data in which lower-level keywords are added to higher-level keywords. For example, if it is a prefecture-specific dictionary in Aichi prefecture, Aichi prefecture XX city xx town, ..., Kariya city, Aichi prefecture Showa Town, Aichi Prefecture, △△ City, ▽▽ Town, and so on. The same applies to prefecture-specific dictionaries of other prefectures.
[0042]
If necessary, a lower-level dictionary may be prepared. In other words, a city-specific dictionary may be prepared for each municipality existing throughout the country. For example, a city-specific dictionary in Kariya City, Aichi Prefecture, a city-specific dictionary in Obu City, Aichi Prefecture, and so on. In Japan, it is said that there are about 4000 municipalities, so about 4000 city-specific dictionaries will be prepared. If this idea is advanced, it is of course possible to prepare a lower-level dictionary. For example, there are 16 wards in Nagoya City, but a 16-divided dictionary may be prepared for the distinction. Of course, it is also possible to prepare a dictionary that is divided into large-size levels below the municipality as well as the ward.
[0043]
Basically, all dictionaries including the dictionary thus divided are recorded on a recording medium such as a DVD set in the data input device 6. The reason for “basically” is that there may be dictionary data resident in the dictionary unit 312 of the voice recognition unit 31. However, the lower-level dictionary described above is stored in a data storage medium such as a DVD as a rule, and is read into the dictionary unit 312 when necessary.
[0044]
Next, the operation of the navigation system 2 of the present embodiment will be described. Since the portion related to the speech recognition device 30 is characteristic, the general operation as a navigation system will be briefly described, and then the operation related to the speech recognition device 30 will be described in detail.
[0045]
After the navigation system 2 is turned on, the driver can operate the remote controller 15a from the menu displayed on the display device 14 (the operation switch group 8 can also be operated in the same manner). When the route information display process is selected for display, or by inputting a desired menu through the voice recognition device 30 through the microphone 35, the remote controller 15a is connected to the control circuit 10 from the dialogue control unit 32. When the same instruction is given for selection via the above, the following processing is performed.
[0046]
That is, when the driver inputs the destination by operating voice or a remote controller based on the map on the display device 14, the current location of the vehicle is obtained based on the satellite data obtained from the GPS receiver 22, and the destination and current location are determined. In between, the cost is calculated by the Dijkstra method, and the shortest route from the current location to the destination is obtained as a guidance route. Then, the guidance route is displayed on the road map on the display device 14 to guide the driver of the appropriate route. Such calculation processing and guidance processing for obtaining a guidance route are generally well-known processing, and thus description thereof is omitted.
[0047]
Next, the operation in the voice recognition device 30 will be described. Here, some operation examples are given.
[Operation example 1]
FIG. 4 is a flowchart illustrating processing in the voice recognition unit 31 and the dialogue control unit 32 in the case of the first operation example.
[0048]
In the first step S10, a dictionary of the highest hierarchy is set. Specifically, it is the above-mentioned prefecture name dictionary, which is read from the DVD by the data input device 6 and set in the dictionary unit 312 of the voice recognition unit 31 through the control circuit 10 and the dialogue control unit 32. As described above, the prefecture name dictionary may be resident in the dictionary unit 312.
[0049]
If preparation for voice recognition is thus completed, voice recognition processing is subsequently performed (S20). As described above, the voice data input through the microphone 35 while the switch 36 is being pressed is extracted by the voice extraction unit 34 and output to the voice recognition unit 31. Recognition processing is executed.
[0050]
After the voice recognition process is performed, it is determined whether or not the lowest-level dictionary is used for the recognition (S30). If it is a recognition using the prefecture name dictionary set in S10, it is not the lowest hierarchy dictionary (S30: NO), so the lower hierarchy dictionary selected from the result of the recognition process in S20 is set (S40). . For example, when “Aichi Prefecture” is selected in recognition using a prefecture name dictionary, a prefecture-specific dictionary for Aichi Prefecture is set. In this setting, as illustrated in FIG. 3, the dialogue control unit 32 requests the control circuit 10 to read the prefecture-specific dictionary. In response to the request, the control unit 10 reads the corresponding prefecture-specific dictionary from the DVD by the data input device 6 and sends it to the dialogue control unit 32. Then, as described above, the prefecture-specific dictionary is set in the dictionary unit 312 of the speech recognition unit 31 by the dictionary control unit 323 (see FIG. 2) in the dialogue control unit 32.
[0051]
Thereafter, the process returns to S20, and the speech recognition process is performed again using the extraction result stored in the extraction result storage unit 313. If the prefecture-specific dictionary is the lowest-level dictionary (S30: YES), the recognition result obtained using the dictionary is output to the control circuit 10 (S50).
As described above, when a city-specific dictionary, a distinction dictionary, a large dictionary, etc. are prepared as lower-level dictionaries of the prefecture-specific dictionaries, the loop processing of S20 to S40 is repeated, and the lowest-level dictionaries are repeated. The result recognized with the dictionary set may be output.
[0052]
In this way, when recognizing a place name input by voice, it is not necessary to read the entire dictionary relating to the place name into the dictionary unit 312, and the prefecture name dictionary and the prefecture-specific dictionary corresponding to the selected prefecture are not included. Just read it. The dictionary is “divided” for such a hierarchically structured word group, and it is determined which sub-layer dictionary (dictionary for each prefecture) should be used in the preliminary determination. The data was narrowed down to the dictionary and then read into the dictionary unit 312. Therefore, even when pursuing the advantage of increasing the number of recognizable objects by increasing the vocabulary prepared in the dictionary, it is not necessary to store all the recognizable vocabularies in the dictionary unit 312. Therefore, even if the dictionary unit 312 has a relatively small capacity, appropriate speech recognition corresponding to batch input can be realized.
[0053]
[Operation example 2]
FIG. 5 is a flowchart showing processing in the speech recognition unit 31 and the dialogue control unit 32 in the case of the operation example 2. Here, it is assumed that the prefecture name dictionary and the command dictionary are stored in the dictionary unit 312 before the actual recognition process is started.
[0054]
In the first step S110, a prefecture name dictionary is set. Since it is stored in advance, it is set here as a dictionary used for speech recognition. That is, although the command dictionary is also stored in the dictionary unit 312, it is not set. In the subsequent S120, the first speech recognition process is performed using the prefecture name dictionary, and a request is made to read the prefecture-specific dictionary selected from the first recognition result (S130).
[0055]
The dialog controller 32 requests the control circuit 10 to read the dictionary as described in the first operation example. Upon receiving this request, the control unit 10 reads the corresponding prefecture-specific dictionary from the DVD by the data input device 6 and sends it to the dialog control unit 32. The dialog control unit 32 reads the prefecture-specific dictionary ( In S190, the prefecture-specific dictionary is set in the dictionary unit 312 of the voice recognition unit 31 (S160).
[0056]
However, since a certain amount of time is required from when the request is sent to the control unit 10 until the prefecture-specific dictionary is sent, here, the second recognition process is performed using that time. That is, this time, the command dictionary is set as a dictionary used for speech recognition (S140), and the second speech recognition process is performed using the command dictionary (S150). When the second recognition process is completed, the prefecture dictionary read from the DVD is set as a dictionary used for voice recognition (S160), and the third voice recognition process is performed using the prefecture dictionary (S160). S170).
[0057]
The probabilities of the second recognition result and the third recognition result obtained in this way are compared, and a higher candidate (recognition result) is output (S180).
As a vocabulary that users input by voice when using a navigation system, there is a place name (address) for setting the destination, etc. Of course, there are cases where commands for using various navigation functions are instructed. . Therefore, according to this operation example, since the second speech recognition process is performed using the command dictionary, not only the place name (address) but also a command can be promptly dealt with. . This recognition process is executed while the prefecture-specific dictionary selected in the preliminary determination using the prefecture name dictionary is being read, so that time loss can be reduced. That is, the response is improved as a whole, and the usability for the user is improved.
[0058]
[Operation example 3]
FIG. 6 is a flowchart showing processing in the voice recognition unit 31 and the dialogue control unit 32 in the case of the operation example 3. Here, it is assumed that the prefecture name dictionary and the current prefecture-specific dictionary are stored in the dictionary unit 312 before the actual recognition process is started. That is, since the current location can be detected by the position detector 4, for example, when a vehicle equipped with the navigation system is traveling in Aichi Prefecture, the prefecture-specific dictionary of Aichi Prefecture is read in advance from the DVD and stored in the dictionary unit 312. Let me.
[0059]
In the first step S210, a prefecture name dictionary is set. Since it is stored in advance, it is set here as a dictionary used for speech recognition. That is, the dictionary unit 312 stores a prefecture-specific dictionary corresponding to the current location, but it is not set. In the subsequent S220, the first speech recognition process is performed using the prefecture name dictionary, and a request is made to read the prefecture-specific dictionary selected from the first recognition result (S230).
[0060]
As a result of the dictionary reading request in S230, the corresponding prefecture-specific dictionary is read from the DVD (S290), and the prefecture-specific dictionary is set in the dictionary unit 312 of the speech recognition unit 31 (S260). The processing contents of S130, S160, and S190 in Example 2 are the same. In the second operation example, the recognition process using the command dictionary is performed using this interval. In the third operation example, the prefecture-specific dictionary corresponding to the current location read in advance is set as a dictionary used for voice recognition. (S240) A second speech recognition process is performed using the prefecture-specific dictionary (S250). When the second recognition process is completed, the prefecture dictionary read from the DVD is set as a dictionary used for voice recognition (S260), and the third voice recognition process is performed using the prefecture dictionary (S260). S270).
[0061]
The probabilities of the second recognition result and the third recognition result obtained in this way are compared, and a higher candidate (recognition result) is output (S180).
For example, if a vehicle equipped with a navigation system is traveling in Aichi Prefecture and you want to set “Showa Town, Kariya City, Aichi Prefecture” as the destination in the same Aichi Prefecture, enter the voice as “Showa Town, Kariya City, Aichi Prefecture” Instead of doing it, it is natural to omit “Aichi Prefecture” and input “Kariya City Showacho” as a voice. In this method, since the recognition is performed using the lower hierarchy dictionary corresponding to the current location in the second recognition, even speech input in which the prefecture name is omitted can be handled.
[0062]
[Operation example 4]
FIG. 7 is a flowchart showing processing in the voice recognition unit 31 and the dialogue control unit 32 in the case of the fourth operation example. As in the case of the operation example 3, it is assumed that the prefecture name dictionary and the prefecture-specific dictionary of the current location are stored in the dictionary unit 312 before starting the actual recognition process.
[0063]
In the first step 310, a prefecture name dictionary and a prefecture-specific dictionary corresponding to the current location are set as a dictionary used for speech recognition. In subsequent S320, the first speech recognition process is performed using the prefecture name dictionary and the prefecture-specific dictionary corresponding to the current location. When the first recognition result is obtained using the prefecture-specific dictionary corresponding to the current location (S330: YES), the first recognition result is output (S340).
[0064]
On the other hand, if the first recognition result is obtained using the prefecture name dictionary instead of the prefecture dictionary corresponding to the current location (S330: NO), the prefecture dictionary selected from the recognition result A request for reading is made (S350), and the corresponding prefecture-specific dictionary is read from the DVD (S355). In this case, unlike the above-described operation examples 2 and 3, no voice recognition process is executed separately from the dictionary reading request to the actual reading.
[0065]
Then, the prefecture-specific dictionary read in S355 is set as a dictionary used for speech recognition (S360), the second speech recognition process is performed using the prefecture-specific dictionary (S370), and the recognition result is output (S380). ).
In this way, when recognizing a place name in a prefecture that includes the current location that is considered to be frequently used, it can be recognized by the first speech recognition process using a prefecture-specific dictionary for recognizing it, The recognition process can be performed relatively quickly.
[0066]
Although four examples of operations in the speech recognition apparatus 30 have been described and the effects of the respective operation examples have been described, the following effects can also be obtained by devising the configuration of the upper layer dictionary.
[Dictionary configuration example 1]
Here, a prefecture name dictionary is considered as an example of an upper hierarchy dictionary. As described above, the prefecture name dictionary has keywords corresponding to the names of prefectures (Aichi Prefecture, Aomori Prefecture, ..., Wakayama Prefecture) as dictionary data. , Wakayama Prefecture *, and a wild card model in which the * part can be matched to any voice input. For example, “Kariya City” in the voice input “Akari Prefecture Kariya City” matches *. If there is only a keyword for the prefecture name as dictionary data, since only a part of the actual recognition target (a group of words including municipalities and capital letters below the prefecture) has the prefecture name, the matching level as a whole Decreases. On the other hand, in the case of the wild card model, since the matching itself can be performed as the entire recognition target, such a problem does not occur.
[0067]
Here, a supplementary explanation of the wild card model will be given.
First, an HMM (Hidden Markov Model) method generally used in speech recognition will be briefly described. Assuming that speech is generated from a Markov model expressed by states and transitions, this method creates the generation model in advance, matches it with the speech, and matches the best. The result is a recognition result. The expression shown in FIG. 8 is common as an example of this model. The output probability distribution corresponds to each state, and the time series of the feature values (represented in two dimensions for simplicity in FIG. 8) as a result of analyzing the speech are in the order (a1) corresponding to FIG. → a2 → a3), the probability is matched from the probability distribution of FIG. 8B. Finally, the recognition result is the one with the best probability product (score called likelihood) until the end of the speech. In this method, HMMs of recognition target vocabulary are prepared and compared, but it is practically impossible to recognize large vocabulary, so phonemes and syllables (this means the word part) The unit is called a sub-word), HMMs are created, and a word model is created by connecting them.
[0068]
Next, a garbage model that is an example of a wild card model will be described. FIG. 9A shows an example of a probability distribution corresponding to each state of the HMMs of / a /, / i /, and / u /. Here, the feature space is two-dimensional. A speech model called a garbage model does not represent a certain feature of a specific syllable, but has a distribution with a large variance so as to cover a large amount of speech. In this way, because the garbage model matches “broad and shallow” to various voice patterns, it outputs a certain score (= probability) for a wide range of voices, but tends to output smaller values than the correct distribution There is. For example, for the voice pattern indicated by “x” in FIG. 9A, the score of / a /, / i / becomes very small and the score of / u / becomes large. On the other hand, the score in the case of the garbage model is larger than the score of / a /, / i /, but smaller than the score of / u /.
[0069]
Therefore, if you match the template of “Aichiken G” (G is a garbage model) and “Aichi Kenkari Yasushi” with the voice of “Aichi Kenkari Ya”, the score will be “Aichi Ken G” <“Aichi Kenkari Yasushi” "Is likely (but not guaranteed). However, if the “Aichiken G (garbage model)” and “Aichiken Kashigarashi” templates are matched with the “Aichikenkarishi” voice, the score will be “Aichiken G” <“Aichi Kenkaishi”. It does not necessarily become, and it reverses with considerable probability.
[0070]
Next, a syllable concatenation model, which is another example of a wild card model, will be described.
The syllable HMM is a unit constituting a word, but if this is arbitrarily connectable, the utterance of any word can be recognized. That is, the voice connection model as shown in FIG. Here, it is assumed that Japanese is recognized.
[0071]
If this is used as a wild card, such as “Aichiken SCM” (SCM is a syllable concatenation model), it is possible to match even utterances such as “Showacho, Kariya City, Aichi Prefecture”. In this case, since the model of “Aichiken SCM” includes the expression of the model “Aichikenkari and Shishowacho”, a score higher than the latter can be obtained as a score.
[0072]
[Dictionary configuration example 2]
The use of the wild card model described above has the advantages as described above. However, since this wild card is gently matched with anything, it also increases the possibility of erroneous recognition. Therefore, for example, when configuring a prefecture name dictionary, the dictionary is prepared in a redundant state in which not only the prefecture name but also the municipality name is added. At the time of recognition, matching is performed up to the municipality name, but as a result, it is determined which prefecture name is matched. Matching with longer voices, and since it does not match anything gently like a wild card model, an improvement in recognition rate can be expected.
[0073]
As described above, the present invention is not limited to such embodiments, and can be implemented in various forms without departing from the spirit of the present invention.
For example, in the above-described embodiment, the voice recognition device 30 is described as an example applied to the navigation system 2 mounted on a vehicle. However, the embodiment may be realized not only as a vehicle-mounted device but also as a portable navigation device, for example. .
[0074]
Also, the present invention can be applied to the case where various data settings and instructions are given by voice input to a device that executes other processing than navigation.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a schematic configuration of a navigation system as an embodiment.
FIG. 2 is a block diagram showing a configuration of a voice recognition unit and a dialogue control unit in the voice recognition device.
FIG. 3 is an explanatory diagram of a dictionary reading request and dictionary reading corresponding thereto;
FIG. 4 is a flowchart showing processing according to an operation example 1 in the speech recognition apparatus.
FIG. 5 is a flowchart showing processing according to an operation example 2 in the speech recognition apparatus.
FIG. 6 is a flowchart showing processing according to an operation example 3 in the speech recognition apparatus.
FIG. 7 is a flowchart illustrating processing according to an operation example 4 in the speech recognition apparatus.
FIG. 8 is an explanatory diagram of an HMM (Hidden Markov Model).
FIG. 9 is an explanatory diagram of a garbage model and a syllable concatenation model as examples of a wild card model.
[Explanation of symbols]
2 ... Navigation system 4 ... Position detector
6 ... Data input device 8 ... Operation switch group
10 ... Control circuit 12 ... External memory
14 ... Display device 15 ... Remote control sensor
15a ... remote control 16 ... geomagnetic sensor
18 ... Gyroscope 20 ... Distance sensor
22 ... GPS receiver 30 ... Voice recognition device
31 ... Voice recognition unit 32 ... Dialog control unit
33 ... Speech synthesis unit 34 ... Speech input unit
35 ... Microphone 36 ... Switch
37 ... Speaker 38 ... Control unit
311: Verification unit 312: Dictionary unit
313 ... Extraction result storage unit 321 ... Processing unit
322: Input unit 323: Dictionary control unit

Claims (9)

Voice input means that can input voice in a row,
Prefecture name template data corresponding to the prefecture name when the final recognition target is an address obtained by hierarchically connecting multiple place names ,
Prefectural template data that is prepared for each prefecture, and corresponds to a group of words that includes the name of the city or town, or even a lower-level place name in addition to the name of the prefecture ,
A first storage means having a relatively low high-speed accessibility during voice recognition processing;
A second storage means having a relatively high speed accessibility for voice recognition processing,
At least the prefecture-specific template data is stored in the first storage means,
When recognizing speech input continuously through the speech input means, first, by comparing the matching data obtained based on the input speech with the prefecture name template data, which prefecture Preliminarily determine whether the name is included,
Prefectural template data corresponding to the name of one prefecture that is included in the preliminary determination is read into the second storage means, and the final recognition result is obtained using the prefectural template data. A voice recognition device characterized by the above. The speech recognition apparatus according to claim 1,
At least one of the prefecture name or prefecture-specific template data is dictionary data. The speech recognition apparatus according to claim 1,
A speech recognition apparatus, wherein at least one of the prefecture name or the template data for each prefecture is speech data. The speech recognition apparatus according to any one of claims 1 to 3,
When the address is composed of three or more levels of place names , the prefecture name template data and prefecture-specific template data are provided, and the prefecture-specific template data is regarded as the upper hierarchy and the place name level lower than the municipality is also included. It was to distinguish speech recognition apparatus comprising: a city-specific template data prepared in municipal units. The speech recognition device according to any one of claims 1 to 4,
Prefecture template data in the prefecture name template data and claim 4, also matched against the speech input other word or a word group after multiple kinds of words or word groups is included to configure the template data A wild card model The speech recognition device according to any one of claims 1 to 4,
The prefecture name template data, der ones than city name, or even municipal and comes lower level places after the prefecture name is, prefectural template data in Claim 4, the lower level than municipalities after the city name A voice recognition device characterized by the fact that the place name is attached . A voice recognition device according to any one of claims 1 to 6, and a processing device that executes a predetermined process based on a result recognized by the voice recognition device,
The voice input means is a processing system that is also used by a user to input a predetermined command that needs to be specified when the processing device performs processing.
The voice recognition device
Command template data, which is template data for recognizing the command, is read into the second storage means before performing the actual speech recognition process, or has high-speed accessibility as with the second storage means. Pre-stored in a relatively high read-only third storage means,
Perform the preliminary determination by performing recognition using the prefecture name template data or the prefecture-specific template data in claim 4 ,
Thereafter, recognition using the command template data is performed, and in parallel with the recognition, prefecture-specific template data corresponding to the result of the preliminary determination or city-specific template data in claim 4 is stored in the second storage means. read, have line recognition by using a city-specific template data in the prefecture template data or claim 4 read that,
The recognition result obtained by using the command template data and the recognition result obtained by using the prefecture-specific template data or the city-specific template data in claim 4 is the final recognition result. A processing system characterized by A voice recognition device according to any one of claims 1 to 6, and a navigation device that executes predetermined processing based on a result recognized by the voice recognition device,
Said voice input means is a processing system we are used to the user an indication of a predetermined place name-related data that needs to be specified on which at least the navigation device to the navigation processing is input by voice ,
The navigation device includes a current location detecting means for detecting a current location,
The voice recognition device
Perform the preliminary determination by performing recognition using the prefecture name template data or the prefecture-specific template data in claim 4 ,
5. The prefecture-specific template data corresponding to the current location detected by the current location detection means or the city-specific template data in claim 4 is read into the second storage means, and the prefecture-specific template data or city-specific template in claim 4 is read. In addition to performing recognition using data, in parallel with this recognition, the prefecture-specific template data corresponding to the result of the preliminary determination or the city-specific template data in claim 4 is read into the second storage means, and the read prefecture There line recognition using city-specific template data in another template data or claim 4,
The recognition result obtained by using the template data by prefecture corresponding to the present location or the template data by city in claim 4 and the template data by prefecture corresponding to the result of the preliminary determination or the template data by city in claim 4 A processing system characterized in that a recognition result having a higher degree of accuracy is used as a final recognition result. A voice recognition device according to any one of claims 1 to 6, and a navigation device that executes predetermined processing based on a result recognized by the voice recognition device,
Said voice input means is a processing system we are used to the user an indication of a predetermined place name-related data that needs to be specified on which at least the navigation device to the navigation processing is input by voice ,
The navigation device includes a current location detecting means for detecting a current location,
The voice recognition device
Pre-loading the prefecture-specific template data corresponding to the current location detected by the current location detection means or the city-specific template data in claim 4 into the second storage means,
Recognition is performed using the prefecture name template data or the prefecture-specific template data in claim 4 and the pre-read prefecture-specific template data or city-specific template data in claim 4 , and the recognition result is read in advance. If is obtained by using a city-specific template data in Oita Prefecture another template data or claim 4, which was the final recognition result, while the recognition result is the prefecture name template data also When using the template data by prefecture in claim 4, the template data by prefecture corresponding to the recognition result or the template data by city in claim 4 is read into the second storage means and read. it was separated by lower-level template de in the prefecture template data or claim 4 Processing system, characterized in that a final recognition result recognition results obtained using data.

Images