JPH10312193A - Voice input device - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To provide a voice input device capable of inputting a large vocabulary voice with a small number of input times. SOLUTION: A microphone 500 and an A / D converter 501 are provided.
Takes in a voice, converts it into a digital voice signal, and outputs it to the signal processing device 5. The external storage device 503 stores a hierarchical word dictionary. The monitor 60 with a touch panel displays the recognition result and can change the partial dictionary by touch input. The signal processing device 5 first inputs a predetermined partial dictionary at the lowest hierarchy, calculates the degree of coincidence between the voice signal and a word in the dictionary, and determines the word with the highest degree of coincidence on the monitor including its upper-layer structure. indicate. If the word is not the word to be recognized, the hierarchy to be changed is touch-input. The touch signal is detected, and the corresponding dictionary is read from the external storage device 503. After that, the hierarchy is determined by voice, and the lowest dictionary under the hierarchy is read. The target word is newly uttered, and voice recognition is performed by collation with the dictionary.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice input device capable of performing large vocabulary recognition with a small number of inputs.

[0002]

2. Description of the Related Art A voice input device for recognizing voice and reading information is capable of inputting commands and data without any hand, so that it is difficult to perform manual operations such as a word processor for persons with disabilities and in-vehicle navigation. It is being adopted. As shown in FIG. 13, a typical voice input device is a microphone 500 that captures voice and converts it into an electric signal and outputs a voice signal.
An A / D converter 501 for A / D converting an analog voice signal into digital information, an external storage device 503 storing a word dictionary used for word recognition from the voice signal, a CPU and a memory And a signal processing device 502 that reads a word dictionary into a memory and performs speech recognition by determining the degree of coincidence between a speech signal and a word in the dictionary. The switch 505 is operated by the user of the apparatus and gives the signal processing apparatus 502 the timing of capturing the audio signal. The monitor 504 displays the recognized word.

[0003] In general, in the case of speech input, the recognition rate decreases as the number of words to be recognized increases, and a hierarchical dictionary is used for large vocabulary recognition. At the time of input, voice input is repeated a plurality of times in accordance with the dictionary classification, the hierarchy is advanced, recognition is performed in a state where the number of words to be recognized is narrowed down, and the false recognition rate is reduced.

[0004] The details of speech recognition in the above-described conventional apparatus will be described with reference to the flowchart of FIG. Here, a case in which “Sakuragicho”, which is the JR station name, is input by voice is shown. Such a station name input is used for setting a destination of an automobile navigation device. The external storage device 503 stores, for example, a hierarchical word dictionary as shown in FIG.

First, in step 521, in an initial state of voice input, the signal processing device 502 sets a partial dictionary at the highest level of the hierarchical word dictionary as a first recognition target. Here, the partial dictionary a in FIG. 2 is set.
Next, the device user operates the switch 505 to operate the microphone 5.
At 00, say "facility". This is because the user has recognized that "Sakuragicho" is included in the category of "facilities". The operation of the switch 505 is detected in step 522. Thereby, the voice recognition processing is started.

In step 523, the signal processing device 502
CPU reads the partial dictionary of the set hierarchy from the external storage device 503 into the memory. Initially a partial dictionary a
Is read. The signal processing device 502 includes the switch 50
The average power of the voice until the operation of the voice signal 5 is operated. When the instantaneous power of the voice signal becomes larger than the average power of the voice by a predetermined value or more after the switch 505 is operated, the user speaks. It is determined that it has started, and step 52
At 4, the voice capture is started.

In step 525, the CPU calculates the degree of coincidence between the input speech and the words in the partial dictionary read into the memory. Initially, the degree of matching with each of "address" and "facility" is calculated.

When the instantaneous power of the audio signal falls below a predetermined value, it is determined that the utterance of the user has ended, and in step 526, the capture of the audio ends. When the calculation of the matching degree between the input voice and the word in the partial dictionary is completed, in step 527, the CPU selects the word having the highest matching degree. Here, “facility” is selected because the degree of coincidence is higher.

In step 528, the selected word having a high degree of matching is displayed on the monitor 504 as a result of recognition. In step 529, it is determined whether or not the selected word is a partial dictionary name of a lower hierarchy. If the selected word is a dictionary name, the process proceeds to step 523; otherwise, the process proceeds to step 530.
Proceed to. Here, since the word “facility” is the dictionary name of the partial dictionary e in the lower hierarchy as shown in the word dictionary of FIG. 2, the process returns to step 523.

In step 523, the CPU newly reads the partial dictionary e from the external storage device 503 into the memory. As described above, the hierarchy advances each time the processing is repeated. When there are no partial dictionaries below the selected word, in step 530,
The recognized word is output as an object. Here, "Sakuragicho" is output.

[0011]

As described above, in the conventional voice input device, in order to transmit one word, the partial dictionary name up to the lowest layer to which the word belongs must be input for each layer. Since the hierarchy becomes more complicated as the number of words increases, there is a problem that the recognition load from a large vocabulary imposes a greater input burden on the user. Further, there is a problem that although the user wants to convey the object most, the input of the object gives a sense of incongruity at the end and the usability is not good. The present invention has been made in view of the above-described conventional problems, and has as its object to realize an accurate voice input with a small number of input times.

[0012]

Means for Solving the Problems Speech input means for converting uttered speech into information and inputting the word, a word dictionary comprising a plurality of words and having a hierarchical structure, and a partial dictionary for determining a partial dictionary to be used Determining means, calculating means for calculating the degree of coincidence between the input voice from the voice input means and the determined word in the partial dictionary, and selecting and outputting the word having the highest degree of matching from the calculated words Selection means, wherein the partial dictionary determination means first determines a predetermined lowermost partial dictionary, and changes the hierarchy when receiving a hierarchy change instruction to determine the partial dictionary.

The partial dictionary determining means, when receiving an instruction to change to a higher hierarchical level, determines a partial dictionary in the upper hierarchical level, changes to a hierarchical level indicated by the word selected by the word selecting means by input, and changes the hierarchical level to a level indicated by the word selected by the word selecting means. A predetermined bottom sub-dictionary can be determined. It is preferable that the partial dictionary first determined by the partial dictionary determining means is the lowest partial dictionary used last time. It is also possible to use a lower-level partial dictionary that is frequently used instead of the lowest-level partial dictionary used last time.

[0014] It is desirable that display means is connected to the voice input device to display the word selected by the word selection means and the higher hierarchical structure. The display unit may have a touch panel and output a hierarchy change command corresponding to the hierarchy touched by the user to the partial dictionary determination unit. Further, it is desirable that a changeable hierarchy is displayed on the display means when a touch input is made on the touch panel.

The partial dictionary determining means determines that a change instruction to a higher hierarchical level is received when an input is made following the object, and determines all partial dictionaries higher than the set lowest partial dictionary. Then, the hierarchy can be changed to the hierarchy indicated by the word selected by the word selecting means.

The sub-dictionary deciding means, when an input is made following the object, decides the lowest sub-dictionary set and all sub-dictionaries higher than the set sub-dictionary, and selects the word selected by the word selecting means. Is a word in the upper sub-dictionary,
Assuming that a change instruction to a higher hierarchy is received, the hierarchy can be changed to the hierarchy indicated by the word.

[0017] In addition, when the hierarchy is changed, the word selecting means uses the voice stored in the storage means to store the first input voice. The degree of coincidence with a word in the partial dictionary can be calculated and output.

[0018]

When a partial dictionary is set using a hierarchical word dictionary, a predetermined lowermost partial dictionary is first set, so that large vocabulary recognition can be performed and the first word is used as an object. You can also. If it is not recognized by one input, the hierarchy is changed by the hierarchy change command, so for example, change to the upper hierarchy common to the current partial dictionary and the partial dictionary with the object, and determine the lowest partial dictionary This eliminates the need for input from the uppermost partial dictionary, thereby achieving efficient input. In addition, since the input is performed from the beginning with the object and the change in the hierarchy of the dictionary cannot be recognized, the voice input matching the sense can be performed.

The partial dictionary determining means, when receiving an instruction to change to a higher hierarchical level, determines a partial dictionary in the upper hierarchical level, and changes to a hierarchical level indicated by the word selected by the word selecting means by input.
If the predetermined lowermost sub-dictionary under the hierarchy is determined, even if the hierarchy is changed, the recognition of the partial dictionary between the object partial dictionary and the upper hierarchy is not necessary, and the hierarchy is changed to the higher hierarchy. Even if this is done, the object is immediately input, so that an effect can be obtained that can be input without frustrating the user of the device.
Further, if the partial dictionary determined first by the partial dictionary determining means is the lowest partial dictionary used last time, the dictionary configuration can be understood and the hierarchy can be changed smoothly. Further, the same effect can be obtained by using a lower-level partial dictionary frequently used in place of the lowest-level partial dictionary used last time.

When the display means is connected to the voice input device, and the word selected by the word selection means and the higher hierarchical structure are displayed on the display means, when the hierarchical change is required, the displayed hierarchical structure is relied on. Since the hierarchy can be changed, the input load can be reduced. When the display means has a touch panel and outputs a hierarchy change command corresponding to the hierarchy touched by the user to the partial dictionary determination unit,
The input becomes interactive and the input load is further reduced. In addition, when there is a touch input, when a hierarchy that can be changed by voice input is displayed on the display unit, the structure of the dictionary can be known from the display screen, and even if there is no prior knowledge of the dictionary configuration, etc. The effect that the hierarchy can be changed is obtained.

The partial dictionary determining means determines that when an input is made following the object, an instruction to change to a higher hierarchical level is received, and determines all partial dictionaries higher than the set lowermost partial dictionary. If the hierarchy is changed to the hierarchy indicated by the word selected by the word selection means, the hierarchy can be changed only by the words in the upper hierarchy, and the input load can be further reduced.

The sub-dictionary deciding means, when an input is made following the object, decides the lowest sub-dictionary set and all the sub-dictionaries higher than the set sub-dictionary, and selects the word selected by the word selecting means. Is a word in the upper sub-dictionary,
When a change instruction to a higher hierarchy is received and a hierarchy change is performed to the hierarchy indicated by the word, the hierarchy change due to erroneous word recognition is prevented while the partial dictionary is accurate.

When the hierarchy is changed, the word selection means uses the speech stored in the storage means to determine the partial dictionary of the lowest layer when the storage means for storing the first input speech is stored. By calculating and outputting the degree of matching with the words in the word, even if the hierarchy is changed and the partial dictionary at the bottom layer changes, there is no need to input the target speech again, and the number of times of voice input is further reduced. It is.

[0024]

Next, embodiments of the present invention will be described with reference to examples. FIG. 1 is a block diagram showing the configuration of the first embodiment of the present invention. The audio signal captured by the microphone 500 is converted into digital information by the A / D converter 501 and input to the signal processing device 5. The switch 505 is operated just before the device user speaks for voice input, and gives the signal processing device 5 a detection timing of the speech voice. The external storage device 503 stores a word dictionary. The monitor 60 is a monitor with a touch panel, which displays the processing results of the signal processing device 5 and allows the partial dictionary to be changed using the touch panel.

As the word dictionary, a hierarchical word dictionary as shown in FIG. 2 is used. In the figure, the connecting lines between the partial dictionaries indicate the connection relationship between the upper and lower layers, and what protrudes into the frame is the display of words representing the lower hierarchy. "Street address"
The partial dictionary a, which is composed of two words, “a” and “facility”, is located at the top layer of the word dictionary, and each word is a lower-level dictionary in which the “address” is a part in which the name of the prefecture is written in accordance with the government-designated city. A dictionary b is provided, and for each prefecture, a partial dictionary c in which city and ward names are described and a partial dictionary d in which ward and village names are described are sequentially provided.

Below the word "facility" in the partial dictionary a, a partial dictionary e of words describing the type names of "facility" such as "station", "department store" and "hotel" follows. "station"
A sub-dictionary f in which the names of prefectures are described is provided below. From partial dictionary f, partial dictionary g describing transportation company names
And a partial dictionary h of station names connected to the partial dictionary g is provided for each prefecture.

The word "department store" is also composed of a lower partial dictionary i in which the names of prefectures are written, a partial dictionary j in which the names of cities are written,
It has a partial dictionary k describing the name of "department store", and is linked to a partial dictionary of cities and wards and a partial dictionary of wards and villages by prefecture. The partial dictionary d, the partial dictionary h, and the partial dictionary k are the lowermost partial dictionaries, in which object words are registered, and words to be recognized are collated and recognized here.

The signal processor 5 calculates the average power of the audio signal digitized from the microphone 500 until the switch 505 is operated in the audio input unit 53. When the instantaneous power becomes larger than the average power value, it is determined that the utterance has started, and the capture of the audio signal is started. The voice signal is subjected to word recognition in the word selection unit 52. At this time, the word dictionary used first is the partial dictionary in the lowest hierarchy read from the external storage device 503 by the partial dictionary determination unit 51.

The word output unit 54 temporarily outputs the result of the recognition and the name of the upper layer of the word to the monitor 60, but suspends the output as a device. During this time, the partial dictionary can be changed by touch-inputting the name of the upper hierarchy displayed on the monitor 60. If there is no touch on the touch panel within the predetermined time, the word output unit 54 outputs the recognition result as an output of the device.

When there is a change in the partial dictionary, the partial dictionary determination unit 51 reads the partial dictionary having the touched name from the external storage device 503, and the device user operates the switch 505 to determine the lower-level dictionary. To speak. The partial dictionary determining unit 51 determines a predetermined lowest-order partial dictionary related thereto and reads it from the external storage device 503. Thereafter, voice input is performed again, and word recognition is performed in the word selection unit 52. The lowest partial dictionary input first or after the hierarchy change is the partial dictionary used last time or the most frequently used partial dictionary.

Next, the operation flow of the apparatus will be described with reference to the flowchart of FIG. First, in step 101, in the initial state of voice input, the partial dictionary determination unit 51 sets the lowest partial dictionary as the first recognition target in the hierarchy. The most frequently used partial dictionary or the lowest partial dictionary used last time is used as the initial setting,
Here, for example, it is assumed that the partial dictionary h in FIG. 2 has been set.

The device user operates the switch 505 to
Speak into the microphone 500. In step 102, the voice input unit 53 detects whether or not the switch 505 has been operated. If the switch 505 has been operated, the average power value of the voice after the microphone 500 has been operated is calculated, and the process proceeds to step 103. In step 103, the determined partial dictionary is read from the external storage device 503 into the partial dictionary determination unit 51. Here, the partial dictionary h is read.

In step 104, the utterance voice of the user is started. That is, when the instantaneous power becomes larger than a predetermined value as compared with the average power of the continuously calculated voice signal of the input voice signal, it is determined that the user's utterance has started, and the capturing of the utterance voice starts. Here, for example, it is assumed that the user has spoken “Sakuragicho” as the object. At step 105, the word selection unit 52 calculates the degree of coincidence between the word in the partial dictionary h read by the partial dictionary determination unit 51 and the uttered voice. In this process, a speech section portion divided by a predetermined time is compared with each word while taking a speech signal, and the acquisition of the speech signal is proceeding at the same time.

At step 106, when the instantaneous power of the audio signal falls below the average power, the user determines that the utterance has been completed, ends the voice capture, and proceeds to step 107. In step 107, the word having the highest matching degree calculated by the word selection unit 52 is output to the word output unit 54 as a recognition result.

In step 108, the word output unit 54 temporarily outputs the result of the recognition and the name of the upper layer of the word to the monitor 60 for display, but suspends the output as the device.
FIG. 4 shows a display screen of the monitor 60. h is a dictionary of the selected word, and a, e, f, and g are high-order dictionaries of the word. The object displayed from the box of h is the object, and the words arranged in the box above it are words in the upper dictionary related to the object, and the hierarchical order is displayed from right to left. I have. When changing the hierarchy, the user can touch the place where the word is displayed to change the hierarchy. And step 1
In step 09, if there is no touch input to the touch panel within a predetermined time, in step 110, the recognition result of the word "Sakuragicho" is output as an output of the apparatus. In this case, there is no need to change the hierarchy, and the number of utterances ends once.

If the user has uttered "SOGO" in step 104, another word will be displayed because the initial partial dictionary is h. In this case, since it is necessary to change the upper layer, the user touches e in FIG.

When the touch input is detected in step 109, the process proceeds to step 111. In step 111, the monitor 6 determines which layer the touched layer can be changed to.
Display at 0. Here, “station”, “department store”, and “hotel” are displayed. FIG. 5 shows the display state. In step 112, the partial dictionary determination unit 51 newly reads the partial dictionary e from the external storage device 503.

In step 113, the user selects from the upper hierarchy displayed by touching and inputs a voice. "Sogo"
In the case of (1), the user inputs the "department store" related thereto by voice and is recognized by the word selection section 52. In this step, the same processing as steps 102 to 107 may be performed. In step 114, the partial dictionary determination unit 51 determines the lowermost partial dictionary k connected to the recognized word “department”.

In step 115, an input instruction mark is displayed again on the monitor 60 in order to notify the user of the voice input retry again, and the process returns to step 102. Thereafter, the user utters “SOGO” and the processing from step 102 to step 108 is performed again.
In, the word “SOGO” is output.

Microphone 500, A / D converter 501,
The voice input unit 53 constitutes voice input means. The partial dictionary determination section 51 constitutes a dictionary determination unit. The word selecting section 52 constitutes a word selecting means. The monitor 60 constitutes a display unit.

The present embodiment is configured as described above. When the first speech is input, the degree of coincidence between the speech signal and a word in a predetermined lowermost partial dictionary is calculated. Can be used as the object, and the input burden on the user is reduced. Further, when the partial dictionary needs to be changed, even if there are a plurality of layers below the upper layer, the lowest part of the partial dictionary is set and used. Then, since the partial dictionary used first is the one frequently used or the one used last time, the user can easily understand the output result and smoothly change the hierarchy.

Further, since the monitor displays the upper hierarchical structure of the word when displaying the word, it is possible to easily determine which layer the user wants to change. When there is a touch input, by displaying what hierarchy can be changed by voice input, it is possible to operate to change to which hierarchy without knowing the dictionary configuration.

FIG. 6 is a block diagram showing the configuration of the second embodiment. In this embodiment, a signal processing device 6 provided with an audio storage unit 66 is used instead of the signal processing device 5 in the first embodiment shown in FIG. The voice signal storage unit 66 outputs the voice signal of the voice input unit 63 to the word selection unit 62, and stores the first voice signal in the voice storage unit 66.
When there is a change in the partial dictionary, the degree of coincidence with the partial dictionary is determined using the voice signal stored in the voice storage unit 66, and the object is recognized. Others are the same as the first embodiment.

Next, the operation flow of the apparatus will be described with reference to the flowchart of FIG. Step 201 to Step 2
Steps 15 to 15 in the flowchart of FIG.
The same processing as in step 14 is performed. That is, first, the partial dictionary at the lowest level of the hierarchy is set as the first recognition target, and is read from the external storage device 503. A word in the set partial dictionary and the degree of coincidence are calculated while taking in the user's uttered voice signal. Then, when the capture of the audio signal is completed, in step 207, the audio storage unit 66 stores the audio signal. The word having the highest matching degree is selected as a result of the recognition, and is displayed on the monitor 60 including the upper hierarchy.

If there is no touch input on the touch panel of the monitor 60 within a predetermined time, the selected word is output as an output of the device. If there is a touch input, the partial dictionary is changed to the touched hierarchy. Do. The partial dictionary whose hierarchy has been changed is newly read from the external storage device 503. Thereafter, an utterance for changing the hierarchy is made, and when a word indicating the hierarchy is specified by voice input, the lowest partial dictionary under the hierarchy is read from the external storage device 503.

Thereafter, in the first embodiment, the user is notified again of the re-input of the voice. In the present embodiment, however, in step 216, the voice signal stored from the voice storage unit 66 is used to store the voice in the lowermost partial dictionary. Calculate the degree of matching with the word. The word having the highest matching degree from each word is determined in step 208.
Selected in Then, in step 210, monitor 6
If it is determined that there is no touch input to 0, step 21
The word recognized in 1 is output as an output of the device.

According to this embodiment, the same effect as that of the first embodiment can be obtained. Further, since it is not necessary to first speak and input the object again after the partial dictionary is changed, the number of times of input is reduced. Will be.

Next, a third embodiment will be described. FIG. 8 is a block diagram showing the configuration of the third embodiment. In this embodiment, a signal processing device 7 is used instead of the signal processing device 6 in the second embodiment shown in FIG. The voice signal storage unit 66 stores the voice signal of the voice input unit 73 into the word selection unit 62.
The first audio signal is output to the audio storage unit 6
6 memorizes. If there is a partial dictionary change, the voice storage unit 6
6. Recognition is performed by using the audio signal stored in 6.

In the first or second embodiment, the hierarchy is changed by the touch panel of the monitor 60. In this embodiment, the hierarchy is changed by voice. That is, the first and second embodiments differ from the first and second embodiments in that the first speech input corresponds to the object and the subsequent speech inputs correspond to the hierarchy change. A monitor 70 for display only is used instead of the monitor 60. Others are the same as the first embodiment.

Next, the operation flow of the apparatus will be described with reference to the flowchart of FIG. Step 201 to Step 2
Up to 16, processes similar to those in the flowchart of FIG. 7 in the second embodiment are performed except for steps 310 and 313. That is, first, the lowest partial dictionary in the hierarchy is set as the first recognition target, and is read from the external storage device 503. A word in the set partial dictionary and the degree of coincidence are calculated while taking in the user's uttered voice signal. Then, when the acquisition of the audio signal is completed, the audio storage unit 66 stores the audio signal. Then, in step 208, the word having the highest degree of coincidence is selected as a result of the recognition, and in step 209, when the word including the upper layer is displayed on the monitor 70, the process proceeds to step 310.

In step 310, the voice input section 73 detects whether or not the switch 505 has been operated. If the switch 505 has been operated, the flow proceeds to step 313 because there is a voice input. In step 313, all the partial dictionaries higher than the currently used partial dictionary are read from the external storage device 503. After that, as in the first and second embodiments, voice input of the upper hierarchy is performed, and when a word indicating the hierarchy is specified, the lowest partial dictionary under the hierarchy is read from the external storage device 503.

In step 216, the voice storage unit 66
Then, the degree of coincidence with the word in the partial dictionary is calculated by using the speech signal stored from. From each word, the word with the highest degree of matching is selected in step 208. If it is detected in step 310 that there is no operation on the switch 505, the word recognized in step 211 is output as an output of the device. In this way, for example, when the partial dictionary h is initially set and the word to be input is “Sogo” that is not in the dictionary, if the user speaks “Department store”, “Sogo” as shown in FIG. Since the hatched partial dictionaries including the upper dictionary are automatically set, there is no troublesome operation even if there is a hierarchy change.

That is, when the apparatus outputs words of a different hierarchy from the hierarchy specified by the user in response to the first utterance, the user can utter the name of the hierarchy desired by the user, so that the utterance, reply, A natural conversation with detailed explanations can be realized, and an effect that voice input can be performed with the most natural feeling can be obtained. Since the partial dictionaries of each hierarchy are set, there is a concern that the recognition rate may decrease due to speech.However, words in the upper partial dictionary are generally words for classification, and the number of words is in the lowest partial dictionary. Since the number of words is extremely small as compared with the number of words, the recognition rate is sufficient for use.

Next, a modification of the third embodiment will be described. In each of the above embodiments, even when the partial dictionary at the lowest level is set correctly, words may be erroneously recognized due to the influence of noise or the like. In this case, even if the partial dictionaries of all the upper layers are reset, the original partial dictionary is reached, which results in an unnecessary input count. This modification is intended to cope with this, and even if there is an erroneous recognition, words in the same partial dictionary can be processed without changing the hierarchy.

FIG. 11 is a flowchart showing the flow of the operation of the apparatus. This flowchart is configured by replacing Steps 412, 413, and 414 with Steps 313 and 214 in the third embodiment of FIG. That is, first, the lowest partial dictionary in the hierarchy is set as the first recognition target, and is read from the external storage device 503. While taking the user's utterance voice signal, the word and the degree of coincidence in the set partial dictionary are calculated. Then, when the acquisition of the audio signal is completed, the audio storage unit 66 stores the audio signal. The word having the highest matching degree is selected as a result of the recognition, and is displayed on the monitor 70 including the upper layer in step 209.

In step 310, the voice input unit 73 detects whether or not the switch 505 has been operated. If the switch 505 has been operated, there is a voice input.
Proceed to 12. In step 412, the currently used partial dictionary and all the partial dictionaries thereof are read. In step 413, the speech uttered by the user is input, and the degree of coincidence with the word in the read partial dictionary is calculated.

At step 414, it is determined whether or not the word having the highest degree of coincidence exists in the partial dictionary of the upper hierarchy. When the word is in the partial dictionary of the upper hierarchy, it is for changing the hierarchy, and the process proceeds to step 215 to perform the same processing as in the third embodiment. If the word is in the lowermost sub-dictionary, it is an already recognized object, so that step 209 is executed.
The process proceeds to step 310, where it is determined whether a new operation has been performed on the switch. In step 211, a word is output.

As described above, if the partial dictionary h is initially set and the word to be input is "Sakuragicho", the voice signal of "Sakuragicho" is distorted due to noise or the like, and is incorrectly identified as "Kannai". May be displayed. Also "Sakuragicho"
Erroneously speaking "Kannai". Even if such a partial dictionary is correct, FIG.
Since the partial dictionaries of all upper layers including the partial dictionary h are automatically set as indicated by hatching in FIG. 2, if "Sakuragicho" is re-uttered, voice input can be performed without changing the layer. As a result, the hierarchy is changed only when necessary, and the voice input burden is further reduced.
In addition, as in the third embodiment, there is also obtained an effect that a speech input can be performed in a natural dialogue style such as utterance, reply, and detailed explanation.

[0059]

As described above, according to the present invention,
Since the first word can be used as the object, the user can first input the object, and if it is not accurate, change the hierarchy and re-enter the object to recognize it, and can input speech with a natural feeling. Also, when changing the dictionary, for example, go back from the current partial dictionary, change to a higher partial dictionary with the same content as the object, and take steps to determine the lowermost partial dictionary, so that until the object is recognized Fewer inputs and less typing

When the instruction to change to the upper hierarchy is received, the partial dictionary determination means determines the partial dictionary in the upper hierarchy, changes the hierarchy to the hierarchy indicated by the word selected by the word selection means by voice input, and changes the hierarchy to the lower hierarchy. In the case where the predetermined lower-level partial dictionary is determined, there is no need to recognize the partial dictionary up to the partial dictionary containing the object even if the hierarchy is changed, so that an effect of not causing a large input burden can be obtained. . Further, if the partial dictionary determined first by the partial dictionary determining means is the lowest partial dictionary used last time, the dictionary configuration can be understood and the hierarchy can be changed smoothly. Further, the same effect can be obtained by using a lower-level partial dictionary frequently used in place of the lowest-level partial dictionary used last time.

When the display means is connected to the voice input device and the word selected by the word selection means and the upper hierarchical structure are displayed on the display means, when the hierarchical change is required, the displayed hierarchical structure is relied on. Since the hierarchy can be changed, the input burden can be further reduced. When the display means has a touch panel and outputs a hierarchy change command corresponding to the hierarchy touched by the user to the partial dictionary determination unit, the input becomes interactive and the input load is further reduced. Further, when there is a touch input, when a hierarchy that can be changed by voice input is displayed on the display unit,
The structure of the dictionary can be known from the display screen, and the effect of changing the hierarchy without any prior knowledge of the dictionary configuration can be obtained.

The partial dictionary determining means determines that a change instruction to a higher hierarchical level has been received when a voice is input following the object, and determines all partial dictionaries higher than the set lowermost partial dictionary. When the determination is made and the hierarchy is changed to the hierarchy indicated by the word selected by the word selecting means, an operation for the hierarchy change is unnecessary and the input load is further reduced.

The sub-dictionary deciding means decides the lowermost sub-dictionary set and all the sub-dictionaries higher than the set sub-dictionary when the speech is input following the object, and the word selecting means selects the sub-dictionary. If the word is a word in the upper partial dictionary, it is assumed that a change instruction to the upper hierarchy has been received, and if the hierarchy is changed to the hierarchy indicated by the word, the hierarchy due to incorrect recognition of the word while the partial dictionary is accurate is changed. The change is prevented, and an effect of not causing useless input is obtained.

When the hierarchy is changed, the word selection means uses the speech stored in the storage means to determine a partial dictionary of the lowest layer when the hierarchy is changed. By calculating the degree of coincidence with the words in the word and outputting it, even if the hierarchy is changed and the partial dictionary at the bottom layer changes, there is no need to input the target word again, and the number of times of voice input is further reduced. It is. Further, when the apparatus outputs words of a different hierarchy from the hierarchy desired by the user in response to the first utterance, the user can utter the hierarchy name desired by the user in the same manner as in the second embodiment. , Utterance, response, and detailed explanation can be realized, and the effect of voice input with the most natural feeling can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a first embodiment.

FIG. 2 is a diagram showing a configuration of a hierarchical word dictionary.

FIG. 3 is a flowchart of the first embodiment.

FIG. 4 is a block diagram showing a monitor display screen.

FIG. 5 is a diagram showing a monitor display screen when a partial dictionary is changed.

FIG. 6 is a block diagram showing a configuration of a second embodiment.

FIG. 7 is a flowchart of the second embodiment.

FIG. 8 is a block diagram showing a configuration of a third embodiment.

FIG. 9 is a flowchart of the third embodiment.

FIG. 10 is a diagram showing an upper partial dictionary temporarily set when the lowermost partial dictionary is changed.

FIG. 11 is a flowchart showing a modification of the third embodiment.

FIG. 12 is a diagram illustrating a partial dictionary temporarily set when the lowest partial dictionary is changed.

FIG. 13 is a block diagram showing a configuration of a conventional example.

FIG. 14 is a flowchart of a conventional example.

[Explanation of symbols]

 5, 6, 7 Signal processing device 51, 71 Partial dictionary determination unit 52, 62 Word selection unit 53, 63, 73 Audio input unit 54, 64 Word output unit 60, 70, 504 Monitor 66 Audio storage unit 500 Microphone 501 A / D converter 502 Signal processing device 503 External storage device 505 Switch a, b, c, d, e, f partial dictionary g, h, i, j, k partial dictionary

Claims (10)

[Claims] 1. A speech input means for converting an uttered speech into information and inputting the information, a word dictionary having a hierarchical structure including a partial dictionary including a plurality of words, a partial dictionary determining means for determining a partial dictionary to be used, and Calculating means for calculating the degree of matching between the voice information from the voice input means and the determined word in the partial dictionary; and word selecting means for selecting and outputting the word having the highest matching degree from the calculated words. Wherein the partial dictionary determining means first determines a predetermined lowermost partial dictionary, and when a hierarchical change command is received, changes the hierarchy and newly determines a partial dictionary. apparatus. 2. When a change instruction to an upper hierarchy is received, the partial dictionary determination means determines a partial dictionary of the specified upper hierarchy, and changes to a hierarchy indicated by the word selected by the word selection means by input. 2. The voice input device according to claim 1, wherein a predetermined lowermost partial dictionary under the hierarchy is determined. 3. The voice input device according to claim 1, wherein the partial dictionary first determined by the partial dictionary determining means is the lowest partial dictionary used last time. 4. The speech input device according to claim 1, wherein the partial dictionary determined first by said partial dictionary determining means is a lower-level partial dictionary frequently used. 5. The voice input device according to claim 1, wherein a display unit is connected to the voice input device, and the word selected by the word selection unit and a higher hierarchical structure are displayed. apparatus. 6. The voice input device according to claim 5, wherein the display unit has a touch panel, and a hierarchy change command corresponding to a hierarchy touched by a user is output to the partial dictionary determination unit. . 7. The voice input device according to claim 6, wherein said display means displays a changeable hierarchy when a touch input is made. 8. The partial dictionary determining means determines that, when an input is made following an object, an instruction to change to a higher hierarchical level is received, and all partial dictionaries higher than the set lowest partial dictionary are set. 3. The voice input device according to claim 1, wherein the voice input device determines the hierarchy and changes the hierarchy to the hierarchy indicated by the word selected by the word selection unit. 4. 9. The partial dictionary determining means, when an input is made following an object, determines a set lowest partial dictionary and all partial dictionaries higher than the set partial dictionary, and the word selecting means selects the partial dictionary. 3. The speech input device according to claim 1, wherein if the word is a word in the upper partial dictionary, a change instruction to a higher hierarchy is received, and the hierarchy is changed to the hierarchy indicated by the word. 10. The voice input device further has a storage unit for storing a voice signal input first, and when the hierarchy is changed, the word selection unit uses the voice signal stored in the storage unit. 10. The voice input device according to claim 1, wherein the degree of coincidence with the determined word in the lowermost partial dictionary is calculated and output.