JP3513988B2 - Navigation system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [0001] [0001] The present invention relates to:Navigation
StemAbout. Especially for adding accents to synthesized sounds
Of the words that compose the synthesized sound
The amount of accent information can be reduced
It was to soNavigation systemAbout. [0002] 2. Description of the Related Art Conventional speech synthesizers include words of fixed sentences.
Output synthetic speech (synthesized speech)
You. In other words, it is suitable for car navigation systems, for example.
In the speech synthesizer used, for example, "here is near XX
is. ”And other frequently used sentences
Was replaced with the place name.
By outputting a synthesized sound corresponding to the replacement sentence, the driver
In addition, the current location is notified. [0003] The synthesized sound output as described above
Must be accented with
You. Therefore, for example, a so-called waveform editing type speech synthesizer is used.
In the dictionary, the replacement dictionary (the above-mentioned place name
Words that can be replaced with words that make up a fixed sentence)
Is replaced in a fixed sentence
The speech waveform of the word is stored.
The dictionary built into the speech synthesizer that performs synthesis contains replacement words.
The accent type is stored. [0004] The Japanese accent is the pitch of the sound.
It depends on the placement, but the accent type here is
This height arrangement is shown. In particular, Japanese
How high should the accent of the idiom be, that is, where
Whether there is a transition from a high level to a low level
Round. Here, change from this high level to the low level
The last high-level beat of the
And the accent nucleus is N mora (beat) from the beginning of the word
, The accent type of the word is N-type.
Here, the mora (beat) is equivalent to the kana character unit.
(But slightly different from syllables). From a high level
If there is no change to low level, type 0 Accen
It becomes a shape. [0005] In this specification, the type 0
Xent type is called flat type,
A unit of prosody having a nucleus is called a prosodic word. The sentence
For accent, the prosodic word unit
The position is specified. [0006] SUMMARY OF THE INVENTION Conventional fixed phrase words are
In a speech synthesizer that outputs synthesized speech by replacing
As mentioned above, some form (in the above case, the audio waveform
Or accent type), and
Information (hereinafter referred to as accent information as appropriate)
Traditionally, such accent information is
It was registered in a word dictionary. However, for car navigation, etc.
Applicable speech synthesizers require proper nouns, especially
Names are often replaced words, and the number is huge.
(The names of places and facilities throughout Japan reach more than 100,000 words.
ing). Because of this, a huge number of replacement words
When accent information is registered in the word dictionary, the word
There was a problem that the volume of the book became large. In addition,
As a result, there has been a problem that the system is increased in size and cost is increased. Therefore, the accent type of the replacement word must be
There is a method of making the sound flat, but in this case, the synthesized sound is
There was a problem that became natural. The present invention has been made in view of such a situation.
And the amount of accent information to be registered in the dictionary.
Reduced and able to generate natural synthesized sounds
It is to be. [0010] [Means for Solving the Problems]Navigation of the present invention
The system provides at least map information,
Map information storage means for storing
Measurement means, and the contents and measurement of the map information storage means
Output for outputting predetermined information according to the measurement result of the means
A navigation system comprising:
Means for outputting predetermined information as synthesized sound.
When the predetermined information includes a predetermined word,
When the predetermined word is less than or equal to 1
Including multiple prosodic words, which are prosodic units with proto-core
Suppose that the prosodic word located at the end of the word is the consequent
When the rest is the antecedent, the accent on the antecedent
Antecedent accent information that stores accent information
Storage means for storing the consequent,
Word accents are stored in the antecedent accent information storage means and
Determining means with reference to the storage means and the consequent storage means;
For a given word with an accent determined by the column
Generating means for generating a corresponding synthesized sound;
Determines whether a given word has a consequent,
Is determined by referring to the specified word has a consequent.
If not, set the accent type of the given word to flat
If a given word has a consequent,
Event information from the antecedent accent information storage means,
Based on the search results, determine the accent of a given word.
Is characterized. [0011]In the navigation system of the present invention
Means the antecedent accent information storage means
Accent information about
Remembers the consequent. The determination means determines the
Xent, the antecedent accent information storage means and the consequent
Is determined by referring to the storage means, and the generation means is determined by the determination means.
A match corresponding to a given word having the determined accent
It is adapted to generate articulated sounds. In this case smell
The determining means determines whether or not the predetermined word has a consequent.
Judgment by referring to the consequent storage means, a predetermined word
If has no consequent, flatten the accent type of the given word.
If it is determined to be a plate type and the given word has a consequent,
Is the accent information storage means
Search, and based on the search results,
The cents have been made to determine. [0012] [0013] [0014] [0015] [0016] [0017] [0018] [0019] [0020] [0021] [0022] [0023] Embodiments of the present invention will be described below.
However, before that, each means of the invention described in the claims
In order to clarify the correspondence between
In the parentheses after the means, the corresponding examples (but one example)
In addition, the features of the present invention are described as follows.
You. [0024]The navigation system of the present invention
Stores map information that is information about at least a map
Map information storage means (for example, the CD-ROM 3 shown in FIG. 1)
Etc.) and measuring means for measuring the current location (for example, FIG. 1)
Measuring unit 4 shown in FIG.
Outputs predetermined information according to the measurement results of the measuring means
Output means (eg, the output unit 5 shown in FIG. 1)
A navigation system comprising:
Speech synthesis means for outputting predetermined information as synthesized sound (for example,
1 etc.).
When the predetermined information includes a predetermined word, the speech synthesis means
A given word is a simple prosodic with less than one accent kernel
If there is more than one prosodic word at the end,
The prosodic word to be placed is the consequent, and the rest is the antecedent
Sometimes remembers accent information about the antecedent's accent
Antecedent accent information storage means (for example, in FIG. 5
And the consequent element dictionary 34B).
Consequence storage means (for example, consequent element dictionary 34A shown in FIG. 5)
Etc.) and the accent of the given word, the antecedent accent
Decision made with reference to the information storage means and the consequent storage means
Determining means (for example, the word accent search unit 33 shown in FIG. 5)
And fixed phrase reading prosody information generating unit 35, etc.)
For a given word with an accent determined by the column
Generating means for generating a corresponding synthetic sound (for example, as shown in FIG.
Ruled speech synthesizer 16 etc.), and
Whether or not the word has a consequent is referred to the consequent storage means.
If a given word has no consequent,
If the word's accent type is determined to be flat,
If the word has a consequent, the accent information of the antecedent
From the antecedent accent information storage means, and the search
Based on the results, determine the accent for a given word.
And features. [0025] [0026] [0027] [0028] [0029] Of course, in this description, each means is described above.
It is not meant to limit to things. FIG. 1 shows a navigation to which the present invention is applied.
It is a block diagram showing composition of one example of a system. This
Navigation systems are installed in, for example, automobiles
Besides the operation of buttons, etc.
Have been made to work. In addition, this navigator
In addition to being installed in automobiles,
It is designed so that it can be used in a band. The input unit 1 includes, for example, a button and a joystick.
And other microphones for inputting audio
And other voice input means. And the input section
From 1, a signal corresponding to the operation of the operation means is sent to the control unit 2.
Has been made to be sent. Further, the input unit 1
It has a built-in voice recognition device,
The voice input to the input means is
Is also transmitted to the control unit 2.
ing. The control unit 2 includes:
Control the operation of the entire system.
It has been made. For example, the CD-ROM 3
Digitized map information (map image data) on a number scale
Not only data, but also text such as place names and facility names on the map.
Required for other navigation)
Necessary information is recorded (stored), and from there, control
The necessary information is read out based on the command from the unit 2.
It has been done. The measuring unit 4 receives radio waves from GPS satellites
Or use a measuring device such as a gyroscope
For example, the present location is measured.
The output unit 5 outputs predetermined information supplied from the control unit 2.
It has been made to be. That is, the output unit 5 outputs the predetermined information.
If the report is an image or other displayable (for example, a map
For example, to display it when
The display unit 6 composed of a monitor, a CRT, etc.
If it is necessary to generate the sound,
Voice synthesizer 7 that performs voice synthesis based on
The synthesized sound generated by the device 7 and other information that can be output as voice
It has a speaker 8 for outputting. The navigation system configured as described above
In the stem, for example, the input unit 1 displays the current location.
Is operated, the operation signal corresponding to the operation
Is output from the input unit 1 to the control unit 2. Then control
In the section 2, the current position measured by the measuring section 4 is received.
And a map of a given scale, including the current location,
It is read from the CD-ROM 3. Further, the control unit 2
Indicates the map read from the CD-ROM 3
It is supplied to the display unit 6. Thereby, in the display unit 6, for example,
For example, if the map centered on the current location is a mark representing itself (for example,
(E.g., an arrow pointing in the direction of the car)
Is done. Thereafter, the automatic operation performed by the user (driver)
For example, a car is approaching a certain area, intersection, facility, etc.
Then, in the control unit 2, the place name (intersection name,
Is read from the CD-ROM 3,
Along with the information necessary for speech synthesis, the speech synthesis
Output to the device 7. In the voice synthesizer 7, the control unit 2
When receiving the information, based on the information, for example, a regular sound
Voice synthesis is performed, and a synthesized sound is generated. That is, the place name is
For example, in the case of "Kitashinagawa", the speech synthesizer 7
Is a synthetic sound for notifying the user of the current location, for example.
"Near Kitashinagawa" is generated. This synthesized sound
The peaker 8 is supplied and output, whereby the user
Recognizes the current location without moving the line of sight to the display unit 6.
can do. As described above, the present location is output as a synthetic sound.
The point of emphasis (area, facility, intersection, etc.) is CD-RO
It can be stored in M3,
The user can also set it in advance. In addition, the user
However, when the input unit 1 is operated in a predetermined manner, the current position is
It is also possible to output. The CD-ROM 3 is compatible with the system.
It is made detachable, so that future maps
Regardless of the change, the map information after the change is stored
CD-ROM instead of CD-ROM3.
To display information stored in the CD-ROM 3.
That decay effectively renders the system unusable.
It has been made so that it can be prevented. FIG. 2 shows a configuration example of the speech synthesizer 7 in FIG.
Is shown. The speech synthesizer 7 converts the words of the fixed phrase into
Replace with the replacement word (input word) (predetermined word)
It is configured to generate a synthesized sound as described above. Immediately
That is, the control unit 2 sends a speech
Identify the fixed phrases to be used, for example, as information necessary for
Information (typical sentence pattern input) for CD-RO
Replacement words (such as place names recorded in M3)
(Word input). Specifically
Is, for example, as described above,
You. Is generated, it is called "around XX."
A fixed phrase and a replacement word "Kitashinagawa" that can be replaced with XX
Is supplied. In addition, the specific information is the fixed phrase reading prosody information
The replacement word is input to the selection unit 11 by the word accent search unit 13.
, Respectively. The standard sentence reading prosody information selecting unit 11 includes a control unit
2 receives the specific information, and is specified by the specific information.
Information about a fixed sentence to be stored in the fixed sentence pattern storage unit 12.
And supplies it to the fixed phrase reading prosody information generating unit 15
It has been done. The fixed phrase pattern storage unit 12 stores
Sentences frequently used in navigation systems
Fixed phrase (for example, as described above, "Around XX."
"Please turn right (left) at the XX intersection."
Or, "And, in the vicinity of ○○ (destination etc.) at ○ km
You. , "Next is the XX intersection."
Information such as phonological information (reading) and prosodic information
(For example, the start position of a phrase in a fixed phrase,
Accent nucleus position of formed word (morpheme), inserted into fixed phrase
(The position of the pause to be entered, etc.)
Information and prosodic information,
"). [0040] The word accent search unit 13 is
Received the replacement word from the
For example, an accent type as accent information
A search is made from the word accent dictionary 14 and a replacement word (replacement word
Along with the standard sentence reading prosody information generation unit 15
It is made to supply to. Word accent dictionary 1
4 is a fixed pattern stored in the fixed sentence pattern storage unit 12.
When replacing a sentence with a word that composes a sentence,
Accent type of replacement word whose xent type becomes something other than flat type
Is paired with the heading and phonological information (reading) of the replacement word.
I remember it. Therefore, the word access is
The cent search unit 13 actually includes a heading of the replacement word,
Or heading and phonological information is provided
The word accent search unit 13 receives the word
Corresponds to the heading of the replacement word, or heading and phonological information
The attached accent type is converted to word accent dictionary 1.
4 is searched. The word accent dictionary 14 is as described above.
Was replaced with words that make up fixed phrases
Sometimes a replacement word whose accent type becomes something other than a flat type
Only the accent type is memorized and it becomes a flat type.
The accent type of the paraphrase is not stored. Therefore, simply
Such a replacement word is input to the word accent search unit 13.
If it is, the accent type of the replacement word is searched.
Cannot be used (get accent type as search result)
Can not), but in such a case, the word accent
The search unit 13 sends the standard sentence reading prosody information generation unit 15
Failed to search accent type of replacement word
And the phonetic information of the replacement word are output.
Have been. The standard sentence reading prosody information generation unit 15
Receive a fixed sentence pattern from the reading prosody information selecting unit 11
And the replacement word from the word accent search unit 13
Phonological information and accent type
Based on the information, there is only one of the words that
Word (the above-mentioned XX part) is replaced with a replacement word
Generate a replacement sentence, and obtain phonemic information and rhyme for the replacement sentence.
Is generated. That is, the fixed phrase reading prosody information generating unit 15
The phoneme information of the XX part of the fixed phrase is replaced with the phoneme information of the replacement word.
To generate phonemic information of the replacement sentence.
You. Also, the fixed phrase reading prosody information generation unit 15
The starting position of the sentence phrase and the words that make up the replacement sentence
Accent nucleus position, pose position to be inserted in replacement sentence
Generate (determine) prosody information such as the position. Here, words are connected to form a sentence.
And if the word is spoken in the sentence,
The accent nucleus position of the word
May change (such a change in accent nucleus position).
Ization is called accent fusion). Therefore, fixed phrase reading
In the prosody information generating unit 15, such accent fusion
From the fixed phrase reading prosody information selecting unit 11
Prosody information and word accents included in fixed sentence patterns
The accent type of the replacement word from the search unit 13 is
We should not combine them into the prosodic information of the replacement sentence.
Based on that information, so to speak,
Generates (determines) prosodic information of a sentence
(However, the prosody information generated in this way is
And the prosody information included in the
May be the same as the combination
). The fixed sentence reading prosody information generation unit 15
From the word accent search unit 13, the accent of the replacement word
If you receive a message that the type could not be searched,
Determines (recognizes) the accent type of the replacement word of the word as a flat type
It has been done. The fixed sentence reading prosody information generating unit 15
After generating the phonological or prosodic information of the sentence,
Is converted to phonetic symbols or prosodic symbols represented by predetermined symbols
To be supplied to the rule speech synthesizer 16.
You. The rule speech synthesizer 16 includes a prosody processor 17,
Prosody control rule / model storage unit 18, phoneme unit selection unit 19,
It is composed of a phoneme segment storage unit 20 and a speech synthesis unit 21,
Phoneme information generated by the standard sentence reading prosody information generation unit 15,
Sentence with prosodic information such as accent and accent nucleus
Is generated by the rule-based speech synthesis process.
It has been made. That is, the prosody processing unit 17 outputs the
With reference to the model storage unit 18, generation of fixed phrase reading prosody information
From the phonetic symbols and prosodic symbols output from the
Constructs a replacement sentence, such as a pitch pattern and a replacement sentence
Control information such as the duration of each phoneme
Calculate parameters to drive the model that generates the report
It has been made to be. Further, the prosody processing unit 17
Prosody control rules / model notation
The model stored in the storage unit 18 is driven.
The pitch pattern described above, the duration of each phoneme, and
Generates control information (specific numerical values) such as
The data is supplied to the segment selection unit 19. In addition,
The prosody processing unit 17 converts the control information (specific numerical values)
In addition, phonemic symbols from the fixed phrase reading prosody information generating unit 15 are also:
It is supplied to the phoneme segment selection unit 19. The prosody control rule / model storage unit 18
Prosody control rules for controlling the prosody of the sound, and
Various models for generating control information are stored. The phoneme segment selecting section 19 sends the
Phoneme segment data corresponding to the phonetic symbol string of
20 and the control information from the prosody processing unit 17 as well.
The connection is made based on the information. That is, phonemes
The segment selector 19 converts the phoneme segment data into a pitch pattern,
Connect and place based on rhyme duration, power, etc.
To generate an audio waveform corresponding to the replacement sentence.
I have. The phoneme segment storage unit 20 stores, for example, CV, CVC /
Speech unit data required for rule-based speech synthesis in units such as VCV
I remember. The voice synthesizer 21 has a built-in D / A converter
Then, the speech waveform output from the phoneme segment selection unit 19 is converted to a D / A
The data is converted and supplied to the speaker 8 for output.
ing. Next, the operation will be described. Control unit
2 provides the specific information and the replacement word,
The fixed information or replacement word is the standard sentence reading prosody information, respectively.
Input to the selection unit 11 or the word accent search unit 13
You. In the standard sentence reading prosody information selecting section 11, the control section 2
When the specific information is received, the fixed phrase corresponding to the specific information
The pattern is read from the fixed phrase pattern storage unit 12.
Then, it is supplied to the standard sentence reading prosody information generating unit 15. Ma
In the word accent search unit 13, the replacement is performed from the control unit 2.
Upon receiving a word, follow the flowchart in FIG.
Processing is performed. That is, first, in step S1
The replacement word (the heading of the replacement word,
Is heading and phoneme information). And step
In S2, the accent type of the replacement word is
Is searched from the Xent dictionary 14 and proceeds to step S3.
Determine whether the accent type of the replacement word was found.
Is determined. In step S3, replacement from the control unit 2
If it is determined that a word accent type has been found,
Proceeding to step S4, the accent type is read.
Then, the process proceeds to step S5, and the accent type is
Along with the substitute words (phonemic information of the replacement words),
This is output to the rule information generation unit 15 and the process is terminated. On the other hand, in step S3, the control unit 2
Judgment that the accent type of these replacement words was not found
If so, the process proceeds to step S6, and a message to that effect is sent.
Page (hereinafter referred to as an accent-less message, as appropriate)
), But with a replacement word from the control unit 2
The prosody information is output to the prosody information generating unit 15, and the process is terminated. Returning to FIG. 2, the standard sentence reading prosody information generation unit 1
Reference numeral 5 denotes a standard sentence pattern from the standard sentence reading prosody information selecting unit 11.
And the word accent search unit 13
Receive the phonological information and accent type of the replacement word
And, based on the information, a fixed phrase (a fixed phrase reading prosody)
Corresponding to the standard sentence pattern supplied from the information selection unit 11
Replace the XX part of the words that make up the fixed form sentence
Generates a replacement sentence replaced with a word, and the replacement sentence
Phonological information and prosody information are generated. Furthermore, fixed phrases
The reading prosody information generation unit 15 outputs the phonetic information or the prosody information.
Information into phonetic or prosodic symbols, respectively.
It is supplied to the speech synthesis unit 16. Note that the standard sentence reading prosody information generating unit 15
Is no accent type from the word accent search unit 13.
If a message is received, replace it as described above.
The accent type of the word is determined to be flat, and the replacement sentence
Phonetic and prosodic symbols are generated. In the rule speech synthesizer 16, first, the prosody processing
In the unit 17, the prosodic control rule / model storage unit 18 is referred to.
And output from the fixed phrase reading prosody information generation unit 15
From phonetic symbols and prosodic symbols, control information as described above
The parameters for driving the model that generates the
And the model is driven using the parameters.
It is. As a result, specific numerical values of the control information are generated.
You. The specific numerical value of this control information is the phonetic transcription of the replacement statement.
Along with the number, it is supplied to the phoneme segment selection unit 19. Phoneme selection
The selection unit 19 corresponds to the phonetic symbol string from the prosody processing unit 17
The read speech unit data is read out from the speech unit storage unit 20.
These are also used as control information from the prosody processing unit 17.
Replaced by being connected based on specific numbers
An audio waveform corresponding to the sentence is generated. In addition, the fixed reading rhyme
Of the prosody information of the replacement sentence generated by the
One accent type is the pitch pattern of the audio waveform.
Etc., which has a natural accent
A speech waveform of the synthesized sound is generated. Further, the phoneme segment selecting section 19
Of the resulting speech waveform,
The pose of one of the prosody information determined by the generation unit 15
A pause is inserted corresponding to the position, and the pause is
Paid. The voice synthesizer 21 has a built-in D / A converter.
The sound waveform from the speech element selection unit 19 is changed to D /
The signal is A-converted and supplied to the speaker 8. This allows
The synthesized sound corresponding to the replacement sentence is output from the peaker 8.
It is. Next, from the speaker 8, for example, the synthesized sound "Kita
It is near Shinagawa 1-chome. Is output as an example,
The operation of the navigation system will be further described.
You. In this case, the control unit 2 sends a signal to the speech synthesizer 7.
As the specific information, for example, the fixed phrase "Around XX."
And a replacement word (exactly
Is the heading of the replacement word, or
For example, "Kitashinagawa 1-chome" (Kitashi
(In this case, the characters in ""
The characters in parentheses that correspond to the headwords of the substitute words,
(Corresponding to the phoneme information). In the standard sentence reading prosody information selecting section 11,
Receive the specific information corresponding to the sentence "Near XX."
From the fixed phrase pattern storage unit 12, "It is near XX"
Of the fixed sentence pattern, that is, its reading (phonological information) and
Prosody information is retrieved and read. Hereafter, the reading (phonological information) is appropriately changed as follows.
The position of the accent nucleus, which is one of the prosodic information in Hiragana
Is represented by a 'mark, and the boundary between prosodic words is represented by a | mark. In addition, like this
The notation is phonological information and prosodic information (here, prosodic information
Both the accent type (the position of the accent nucleus) in the report
Therefore, it is called phonological information. In the following,
The turn storage unit 12 and the word accent dictionary 14 include
It is assumed that such phonological information is stored. Therefore, the fixed phrase prosody reading information selecting unit 11
Is the phonemic prosody information (fixed sentence pattern)
"Is searched. On the other hand, the word accent search unit 13
Paraphrase "Kitashinagawa 1-chome" (Kitashinagawa Ichome)
Received from the word accent dictionary 14, "Kitashinagawa
1-chome phonetic prosody information "Kitashina" |
Is searched. Here, the fixed phrase "Around XX" is constructed.
Word to be replaced with the replacement word "Kitashinagawa 1-chome"
Replacement sentence "It is near Kita Shinagawa 1-chome."
Put a 韻 mark on the high part and a _ mark on the low part, and
When expressed together with the information, the result is as shown in FIG. What
In the figure, | marks indicate the boundaries between prosodic words as described above.
Represents the world. As shown in FIG. 4A, the replacement statement
In the "Kitashinagawa 1-chome area", the replacement word "Kitashinagawa
The accent type of Kawa 1-chome is not a flat type. Follow
And, the word accent dictionary 14 contains "Kitashinagawa 1-chome"
Phonetic prosody information "Kitashina" is registered.
As a result, the word accent search unit 13
Is the phonetic prosody information of "Kitashinagawa 1-chome"
| Icchome "is retrieved and read. In the word accent dictionary 14,
As mentioned above, the accent type of the replacement word
It is stored in association with the word heading and phonological information.
If the registered font in the word accent dictionary 14
The mat is, for example, as follows. That is, for example, the replacement word “Kitashinagawa 1-chome”
Is Headline: Kitashinagawa 1-chome Phoneme information: Kita Shinagawa Accent type: 4/6 + 0/5 Is registered in the word accent dictionary 14
You. Note that a / b in the accent type is a certain prosodic word
Is composed of b mora and its accent type is a type
+ Indicates a boundary between prosodic words.
Therefore, in the above example, the replacement word “Kitashinagawa 1-chome” is
It consists of two prosodic words, and the first prosodic word is
Accent type 4 with 6 mora, second prosodic word
Must be of type 0 and have 5 mora
And The standard sentence reading prosody information selection unit 11 searches
The phonetic prosody information "Fuki'n."
You. )) And the word accent search unit 13
Phonetic prosody information "Kitashina" |
Both are supplied to the fixed phrase reading prosody information generating unit 15,
There, these are combined and the replacement statement
It is near the river 1-chome. Is generated. That is, fixed phrase reading
In the prosody information generation unit 15, the replacement sentence "Kitashinagawa 1-chome
It's close. Reading and prosodic information about
You. Here, usually, in the preceding stage of “Nearby.”
"Kitashinagawa 1-chome" is added, and as a result, "Kitashinagawa 1
It is near the street. ”Means“ 1 chome ”
The so-called connecting part of "Nearby."
A merge occurs and they are united. As a result, "1 chome" and
And “Nearby.” Are replaced by one prosodic word.
Read about the sentence "It is near Kita Shinagawa 1-chome."
The prosody information, that is, the phoneme prosody information, is already shown in FIG.
As you did, "Kitashina" is
It is. " The replacement statement generated as described above
"Kitashinagawa is near 1-chome."
"Nanakawa" is a ruled voice
Is supplied to the component 16 and the above-described processing is performed.
Thus, the speaker 8 outputs the phonetic prosody information “Kitashi
"I'm |
The sound "Near Kita Shinagawa 1-chome" is output. That is,
A natural synthesized sound is output. Next, in the above case, the replacement word "Kita
The replacement word “Kitashinagawa” is replaced by the control unit instead of “Shinagawa 1-chome”.
2 to the word accent search unit 13
Will be described. In this case, the fixed sentence prosody reading information selecting unit 11
Then, as described above, phonological prosody information (fixed sentence pattern
N) “○○ Fuki” is searched, and the fixed phrase reading prosody
The information is supplied to the information generation unit 15. On the other hand, the word accent search unit 13
The replacement word “Kitashinagawa” from the word accent dictionary 14
A search for the phonetic prosody information of T. Shinagawa is performed. Here, the fixed phrase "Around XX" is constructed.
Word replaced with the replacement word "Kitashinagawa"
The paraphrase "Near Kitashinagawa."
When expressed in the same way as in the case, the result is as shown in FIG. Immediately
Usually, "Kitashinagawa" is added before "Nearby."
And, by this, "it is near Kitashinagawa."
The connection between "Kitashinagawa" and "Nearby."
Fusion, which results in the replacement word "Kitashinagawa"
Cent says "Kitashinagawa" from "Kitashinagawa"
And flatten. Therefore, as shown in FIG.
In the replacement sentence "Near Kitashinagawa."
The accent type of the word "Kitashinagawa" is a flat type. Therefore,
The word accent dictionary 14 contains the phonetic prosodic information of "Kitashinagawa".
The report "Kitashina'agawa" has not been registered, and as a result,
In the word accent search unit 13, the phonetic prosody of "Kitashinagawa"
Information cannot be retrieved. Therefore, simply
As described above, the word accent search unit 13
Along with the phonetic information "Kitashinagawa" of the word "Kitashinagawa",
Xententless message generates fixed phrase reading prosody information
Output to the unit 15. In the standard sentence reading prosody information generation unit 15, the standard sentence reading prosody information
Phonetic prosody information “blow” from the sentence reading prosody information selection unit 11
I'm ”(“ ○ ふ ”) and the word access
Paired with phonetic information “Kitashinagawa” from the client search unit 13
Combined, the replacement sentence "Near Kitashinagawa."
Reading and prosody information are generated. Here, as described above, the standard sentence reading prosody
The information generation unit 15 receives a word from the word accent search unit 13.
A message without accent type with the paraphrase "Kitashinagawa"
When the message is received, the accent type of the replacement word is
It is designed to be treated as a type. Therefore, in this case,
In the fixed sentence reading prosody information generation unit 15, the replacement sentence "Kitashin
It is near the river. And the prosodic information about
As shown in FIG. 4B, as phonological prosody information,
"Kitashinawafuki" is generated. The replacement statement generated as described above
"Near Kitashinagawa." Phonetic prosody information "Kitashinagawa
Is the speed through the rule speech synthesizer 16.
8 is supplied to the phonetic prosodic information “Kitashina”.
Synthetic sound according to "Gawabuki"
You. Is output. That is, a natural synthesized sound is output
You. [0077] A fixed form used in the navigation system
Replace words that make up the sentence with, for example, place names or other replacement words.
In the replacement sentence, the replacement word is often
Causes accent fusion in the fixed phrase,
The flattening of word accents has been reported by the present inventor.
Has been confirmed. Therefore, the word accent dictionary 14 contains
You have registered the accent type of many replacement words like
That is, in the replacement sentence,
Only the accent type of the replacement word whose type is other than the flat type
Reduce the capacity (dictionary size) because it only needs to be recorded
Can be Further, it is registered in the word accent dictionary 14.
Unreplaced words have a flat accent type.
Although it is treated as a plate type, as described above, the word Accen
Replacement words that are not registered in the dictionary 14
Whether the accent type becomes a flat type
However, even if a synthetic sound is generated in this way,
Is not unnatural. In the above description, the replacement word is “Kitashinagawa”.
I explained the case of "Kitashinagawa 1chome" as an example, but this example only
As can be seen from the word accent dictionary 14.
Words that need to be replaced (replacement words) are composed of multiple prosodic words.
Prosodic words other than the last prosodic word are flat
Outside. The word accent dictionary 14 contains
Words except for replacement words that flatten accents like
Alternatively, certain accents that do not flatten the accent
It is also possible to register words except for replacement words
But such replacement words have flattened accents
The number is very small compared to the replacement word
From the viewpoint of reducing the capacity of the accent dictionary 14,
Words except for replacement words whose accents are flattened.
It is desirable to record it. Next, FIG. 5 shows another example of the speech synthesizer 7 shown in FIG.
Is shown. In the figure, the case in FIG.
Are assigned the same reference numerals as those corresponding to
Hereinafter, the description thereof will be appropriately omitted. That is, this sound
The voice synthesizer 7 replaces the word accent search unit 13 with a simple
The word accent search unit 33 is added to the word accent dictionary 14.
Instead, the consequent element dictionary 34A, the antecedent element dictionary 34B, and
And the specific antecedent dictionary 34C is a fixed sentence reading prosody information generation unit 1
5 instead of the standard sentence reading prosody information generation unit 35,
Other than being provided, the configuration is the same as that in FIG.
Have been. The word accent search unit 33 determines whether the control unit 2
From the consequent element dictionary 34A,
34B or the specific antecedent dictionary 34C as necessary.
Access to the replacement word's accent
Search and replace, for example, accent type
Along with the word (phonemic information of the replacement word), the standard sentence reading prosody
The information is supplied to the information generating unit 35. The consequent element dictionary 34A stores words (substitution words).
Heading of the consequent (except for the rare consequent described below)
(And phonetic information if necessary)
In the antecedent element dictionary 34B, the word (substitution word) antecedent (
(Excluding specific antecedents described later)
For example, memorize accent type as Xent information
I have. Here, the consequent of a word means that the word
The last rhyme if it consists of a number of prosodic words
The antecedent means all the remaining prosodic words.
To taste. The specific antecedent element dictionary 34C is an antecedent element dictionary.
Like 34B, accent type of word (replacement word) antecedent
I remember. However, the specific antecedent element dictionary 34C includes:
Access to only specific antecedents (hereinafter referred to as specific antecedents as appropriate)
I remember the cent type. Here, the specific antecedent is a word
In the accent search unit 33, the
Are rarely entered (eg,
In navigation systems, for example,
Is less specific), and the antecedent of the place name is a specific antecedent). The standard sentence reading prosody information generation unit 35
Consequence element dictionary 34A, antecedent by Xent search unit 33
Referring to the element dictionary 34B or the specific antecedent dictionary 34C
Generates a replacement sentence based on the search results of
Cents (including replacement word accents) and other prosody
The information is generated (determined). With the speech synthesizer 7 configured as described above,
Is supplied with specific information and a replacement word from the control unit 2.
And the specific information or replacement word
Prosody information selection unit 11 or word accent search unit 13
Is entered. The standard sentence reading prosody information selection unit 11
When the specific information is received from the control unit 2, as described above,
Phonemic prosody information of the fixed sentence corresponding to the specific information
Turn) is read out from the fixed phrase pattern storage unit 12.
This is supplied to the standard sentence reading prosody information generating unit 35. Ma
In the word accent search unit 33, the replacement is performed from the control unit 2.
When a word is received, follow the flowchart in FIG.
Processing is performed. That is, first, in step S11,
The replacement word (the heading of the replacement word,
Is heading and phoneme information). And step
In S12, the search for the consequent of the replacement word is performed
This is performed with reference to the elementary dictionary 34A. Note that step
In S12, the search for the consequent of the replacement word is
The last part (possibly the whole replacement word) and the consequent
Match with the consequent registered in the elementary dictionary 34A
It is done by doing. Here, the consequent element dictionary 34A contains, for example,
"1-chome" (Ichome) and "2-chome" (Nicho)
), "3-chome" (Sanchome), "4-chome" (Yon
In addition to "~ chome" such as chome, "~ bancho" (~ ban
Town), "North exit" (Kitaguchi), "Honmachi" (honmachi),
Place names (such as "below", "one", etc.)
Are the consequents of, for example, facility names, road names, intersection names, etc.)
Possible words (except for unusual consequents) are registered
I have. When the search of the consequent element dictionary 34A is completed,
In step S13, after the replacement word from the control unit 2
It is determined whether a case has been found. In step S13
Then, if the consequent of the replacement word from the control unit 2 is found
If it is determined, the process proceeds to step S14, and the replacement word
Is searched while referring to the antecedent element dictionary 34B.
Is In step S14, before the replacement word,
Search is performed by excluding the consequent of the replacement word and the antecedent element
Matches the antecedent registered in the dictionary 34B.
Is performed by Here, the antecedent element dictionary 34B contains, for example,
"Kitashinagawa" (Kitashinagawa) and "Higashishinagawa" (Higashishi
(Nagawa), "Nishishinagawa" (Nishishinagawa), "Minamishinagawa"
(Such as Minami Shinagawa)
Words that can be the antecedent of a building name, road name, intersection name, etc.)
(However, except for the specific antecedent) is registered. When the search of the antecedent element dictionary 34B is completed,
In step S15, before the replacement word from the control unit 2
It is determined whether a case has been found. In step S15
Then, if the antecedent of the replacement word from the control unit 2 is found
If it is determined, the process proceeds to step S16, and the
The Xent type is read from the antecedent element dictionary 34B.
Then, the process proceeds to step S17, and the accent type is
Fixed phrase reading along with replacement words (phonemic information of replacement words)
The output is output to the prosody information generation unit 35, and the process ends. The standard sentence reading prosody information generation unit 35 outputs the word
From the xent search unit 33 with the phoneme information of the replacement word.
When the accent type of the antecedent is received,
Information and accent type, and fixed sentence reading prosody information selection
Based on the phonetic prosodic information of the fixed sentence from the selector 11,
Phonological information of paraphrase, accent and other prosodic information
Is generated (determined). That is, for example, “Kitashinagawa 1-chome”
(Kitashinagawa Icchome) is entered as a replacement word
If it was, "Kitashinagawa" (Kitashinaga)
“Wa” is the antecedent and “1chome” (Icchome) is the antecedent.
Are searched in step S14 or S12, respectively.
It is. Therefore, in this case, the word accent search unit
33, the fixed phrase reading prosody information generating unit 35
Phonetic prosody information "Kitashina" for "Kitashinagawa 1chome"
Is output. The consequent element dictionary 34A
Has no registered consequent accent type.
As shown in Fig. 4 (A), the consequents such as "1-chome"
Accents are usually flattened in replacement sentences
So, the accent type of the consequent of the replacement word is, for example,
Is treated as On the other hand, from the standard sentence reading prosody information selecting section 11
For example, the standard sentence reading prosody information generation unit 35
Phonological prosody information of sentence "○○ Fuki'n."
Then, in the fixed sentence reading prosody information generating unit 35,
As in the case of the standard sentence reading prosody information generation unit 15,
Phonological prosody information of replacement sentence "Kitashina" |
Umefukin 'is generated. " Therefore, also in this case, the case shown in FIG.
Similarly, a natural synthesized sound is output. Further
In this case, the replacement word is divided into the consequent and the antecedent,
The consequent element dictionary 34A or the antecedent element dictionary 34B respectively
So that the replacement word is
Compared to the case of registering in the dictionary, the capacity of the entire dictionary is
Can be smaller. That is, for example, "1 Kitashinagawa
"Eye", "Kitashinagawa 2-chome", "Kitashinagawa 3-chome", "Kitashinagawa
4-chome ”,“ Higashishinagawa 1chome ”,“ Higashishinagawa 2-chome ”,“ East
8 words of "Shinagawa 3-chome" and "Higashi-Shinagawa 4-chome" are replaced words.
The replacement word in the dictionary
You need to remember all of these,
And the antecedent are divided and registered, the consequent element dictionary 3
4A includes “1 chome”, “2 chome”, “3 chome” and
And “4” are stored in the antecedent element dictionary 34B.
Remembers "Kitashinagawa" and "Higashishinagawa"
It only needs to be done. In this case, the search for the consequent is replaced with the search for the predecessor.
Speed up the process because it is done before the search
Can be. That is, for example, a place name is set as a target of a replacement word.
If the number of consequents is much smaller than the number of consequents
Become. Then, in the flowchart shown in FIG.
When it is determined that the consequent was not found in step S13
As described later, the antecedent search process in step S14
Is performed, and the processing after step S18 is performed.
Therefore, the processing can be speeded up. On the other hand, in step S13, the control unit 2
Determined that the consequent of the replacement word from was not found
Or the consequent of the replacement word from the control unit 2 is found.
In step S15, the antecedent was found.
If it is determined that there has not been, the process proceeds to step S18,
The search for the antecedent (specific antecedent) of the replacement word from the control unit 2
This is performed while referring to the specific antecedent element dictionary 34C. What
In step S18, the antecedent of the replacement word (before identification)
Search) is the part of the replacement word excluding the consequent or the replacement
The first part of the word (possibly the whole replacement word)
Match with the antecedent registered in the fixed antecedent element dictionary 34C
(Heading or phonemic information matching)
Done by Here, for example, in one area of Hyogo prefecture, the
"Bamachi Ikuno" and "Dojomachi Day"
Bottom part ”, Dojocho Shioda
Unusual place names such as
Place name) exists. In such a place name, "Dojomachi"
Is an antecedent, and “Ikuno” (go
), "Kusakabe" (Kusakabe), "Shiota" (Shiota)
Which is the consequent. In this case, the consequent “Ikuno” (Ikuno
), Kusakabe (Kusakabe), Shioda (Shiota)
Is frequently used, such as "1 chome" mentioned above.
Unlike consequents, they are mostly used as consequents of place names
Can not be (uncommon), so to speak, a rare consequent
I can say. Therefore, in addition to the general consequents,
All unusual consequents were registered in the consequent element dictionary 34A.
Then, the capacity of the consequent element dictionary 34A becomes large,
It takes time to search for the consequent in step S12
It will be. On the other hand, the antecedent is "Dojo-machi"
The compound place name is one of the Hyogo prefectures mentioned above.
Except for now, it does not exist. [0103] Therefore, "Dojocho" (Dojocho)
Replacement word (place name) with antecedent (or its
The first part is "Dojo-cho"
Replacement word), its antecedent or consequent,
Registered in the element dictionary 34B or the consequent element dictionary 34A, respectively.
Without recording, "Dojo-cho" is registered in the specific antecedent element dictionary 34C.
And further, as shown in FIG.
A, Search processing for the antecedent element dictionary 34B (step
Separate from the specific antecedent element dictionary
Search processing for 34C (processing in step S18)
To do. In this case, the consequent of the replacement word from the control unit 2
Is not found, or when the control unit 2
If the consequent was found, but the antecedent was not found
Only when the search for a specific antecedent is performed.
Can be speeded up. Furthermore, the consequent element dictionary 3
Prevents increase in 4A capacity and delays in subsequent search processing
You can also. The specific antecedent element in step S18
(Specific antecedent element dictionary 34C is referred to
The search for the antecedent (specific antecedent) of the replacement word is terminated.
Then, the process proceeds to step S19, where a specific antecedent is found.
Yes or No (before the specifics registered in the specific antecedent element dictionary 34C)
Is included in the replacement word from the control unit 2?
No) is determined. In step S19, before the identification
If it is determined that a case has been found, the process proceeds to step S20.
Mi, the accent type is from the specific antecedent element dictionary 34C.
Is read. Then, the process proceeds to step S21,
Xent type is replaced with the replacement word (phonetic information of the replacement word)
In addition, the standard sentence reading prosody information generation unit 35 outputs
finish. Thus, for example, as a replacement word,
"Dojomachi Ikuno" is input and the specific antecedent element dictionary 34C
, The phonetic prosody information of the specific antecedent "Dojomachi"
If "cho" is registered, the word accent search unit 3
3 to the standard sentence reading prosody information generating unit 35,
Phonological prosody information of "Dojomachi Ikuno"
"Go" is output. Note that the replacement word “Michi
The part after the specific antecedent of "Bamachi Ikuno", that is, the consequent "Ikuno"
The accent type is the flat type, as described above.
It is. Hereinafter, the fixed phrase reading prosody information generation unit 35 and
And in the rule speech synthesizer 16 as in the case described above.
The processing is performed, so that a natural
A synthesized sound is output. On the other hand, in step S19, the control unit 2
Predecessor (specific antecedent) of replacement word from was not found
Is determined, that is, the replacement word is
34A, antecedent element dictionary 34B, or specific antecedent element dictionary
Consequent, antecedent, or
If none of the specific antecedents is included, proceed to step S22.
Only, a message to that effect, ie no accent type
The message is a fixed form with the replacement word from the control unit 2.
This is output to the sentence reading prosody information generating unit 35, and the processing is terminated.
You. In this case, the fixed phrase reading prosody information generation unit 35
In, the accent type of the replacement word is
Phonological prosodic information of the sentence is generated. And, below,
The synthesized sound is output from the speaker 8 in the same manner as
It is. In the above case, the specific antecedent
34C, and search for it
However, it is possible to prevent such processing
It is. That is, in step S13, the control unit 2
If it is determined that the consequent of the replacement word is not found
Or the consequent of the replacement word from control unit 2 is found.
However, in step S15, the antecedent was found.
If it is determined that there is no data, the process of step S18 is performed.
Alternatively, the processing of step S22 may be performed.
Noh. In this case, the processing speed can be increased.
Since there is no need to provide the specific antecedent element dictionary 34C,
It is possible to reduce the cost and size of the system.
You. In this case, however, the place name (place
Paraphrases) are to be all flat accents
Therefore, if the accent type is not flat,
As a result, an unnatural synthesized sound is generated. However
However, the number of place names including specific antecedents is small,
Since the probability that the place name is included in the replacement sentence is low,
In this way, the search processing for the specific antecedent dictionary 34C is performed.
Even if it is not, unnatural synthesized sounds will not be output frequently.
No. Further, in the above case, the antecedent element dictionary 34B
(The same applies to the specific antecedent element dictionary 34C)
Regardless of the accent type in the paraphrase,
The cent type was registered, but the antecedent element dictionary 3
4B shows the case in the word accent dictionary 14 of FIG.
In the same way as in
Be sure to register only the antecedents other than the flat type.
Both are possible. In this case, the antecedent element dictionary 34B (specific
The capacity of the antecedent element dictionary 34C) can be reduced.
You. For example, for example,
There is a place name consisting of one prosodic word, such as
And such place names do not exist in reality).
The replacement word is input to the word accent selection unit 33.
In this case, the processing according to the flowchart of FIG.
In this case, this replacement word is a specific antecedent "dojocho"
And the consequent "i" is erroneously divided. So
Here, in order to prevent such an incorrect division,
For example, if the number is less than 2 moras, the word accent
In the selection unit 33, the replacement word is set to two prosody words,
And the consequent. Further, the notation of the place name on the map has a consequent
What consists only of numbers, for example, "Kitashinagawa 5"
Shinagawa). For example, the fixed phrase "around XX
is. "Is replaced with" Kitashinagawa 5 "
Ta (combined with "Kitashinagawa 5") replacement sentence "Kitashinagawa
It is around 5. ), When you actually say
"5" which is the consequent of "Kitashinagawa 5" is "Nearby."
Composing accents to form one prosodic word
And the accent is flattened. But
In this way, the place name whose consequent consists only of numbers (replacement
")", "5" and "Nearby" are accented
Rather than fusing them, split them into separate prosodic words
It is better to make the character part accent type that the number has
The synthesized sound becomes easy to hear. That is, for example,
About "Near Kita Shinagawa 5", as shown in Fig. 7
Prosody information makes the synthesized sound easier to hear.
You. Therefore, the consequent element dictionary 34A contains the consequent
For the number, the heading (or heading
And phonetic information) as well as their accent types
And further, in step S13 of FIG.
When a case is found, a process as shown in FIG. 8 is performed.
To do. That is, in step S31, the consequent is a number.
It is determined whether the consequent is not a number or not.
Step S14 and subsequent steps are performed. The consequent is a number.
If the search for the antecedent is performed, the process proceeds to step S32.
U. Then, the process proceeds to step S33, where the antecedent and the
Both accent types (however, if the antecedent was not found
If the antecedent, no message with accent type)
From the word accent search unit 33 to the standard sentence reading prosody information
The data is output to the generation unit 35. Then, the fixed phrase reading prosody information generating unit 35
From the word accent search unit 33,
If both accent types are received, the consequent is accented
Instead of fusing, treat them as independent prosodic words,
And use both accent types of the consequent
Phonological prosody information of a paraphrase is generated. More than
By doing so, the fixed phrase reading prosody information generation unit 35
Is, for example, the replacement sentence "Near Kitashinagawa 5".
Is "Kitashina" |
Phonological prosody information will be generated. In addition to numbers, no accent fusion
Consequences that are easier to hear (e.g.
("O'tsu"), Hei ("Hey"), etc.
The accent information is stored in the consequent element dictionary 34A.
In addition, the fixed phrase reading prosody information generation unit 35 includes the consequent element
Has consequent that accent information is stored in dictionary 34A
Have only the above-mentioned consequents for the replacement word
Prosody information is generated in the same way as
It is possible to make it. In this case,
Xcent is more logical than flattening
It is possible to provide a synthesized sound that is easy to solve. In the above case, the consequent element dictionary 34
A to register the consequent accent type in A
However, the consequent element dictionary 34A has
Can be registered. This
, The replacement word is the consequent element dictionary 34A,
In the contract 34B or the specific antecedent element dictionary 34C, respectively.
Any of the recorded consequents, antecedents, or specific antecedents
When there is an exact match, look for the match.
After attaching, without further processing,
The accent type is input from the word accent search unit 33 as a fixed phrase.
It is possible to make the reading prosody information generation unit 35 output the information.
It is possible. As described above, the present invention is applied to a navigation system mounted on an automobile.
Explained the case of applying to the gating system
However, the present invention is also mounted on, for example, an aircraft or a ship.
Navigation system, etc.
You. In addition, the present invention, in addition to the navigation system,
Voice guidance device that performs voice response using synthesized sound
Applicable to any other devices. In this embodiment, the current location is
For example, the fixed phrase "Around XX."
As described above, the output of the synthesized sound has been described.
It is also applicable when using fixed phrases. Further, in this embodiment, the fixed pattern storage is performed.
Part 12 and word accent dictionary 14 (or consequently
Elementary dictionary 34A, antecedent element dictionary 34B, and specific antecedent required
Elementary dictionary 34C) in the speech synthesizer 7.
However, the stored information is stored, for example, on a CD-ROM3 or the like.
It is possible to register. In this embodiment, the speech synthesizing device 7
To generate synthesized speech by regular speech synthesis processing.
However, the speech synthesizer 7 generates synthesized sounds by other methods.
It is possible to make it. Further, in the present embodiment, the phoneme information
, The word accent dictionary 14 (or the consequent element
Dictionary 34A, antecedent element dictionary 34B, and specific antecedent element
The same applies to the dictionary 34C).
Is provided by the control unit 2 as the replacement word.
(If this is the case, this is
What you can do by recording kana readings
Can be registered in the word accent dictionary 14)
You do not need to do this (for example, Hiragana reading
Only the accent and accent types are sufficient). In this case, the word
To further reduce the capacity of the accent dictionary 14
Can be. [0124] 【The invention's effect】According to the navigation system of the present invention
If the antecedent accent information storage means is
Accent information is stored in the consequent storage means
Has stored the consequent. And the given word is the consequent
Is a consequent storage means in which the consequent is stored
Is determined by referring to the
If not, set the accent type of the given word to flat
Is done. If a given word has a consequent,
Information about the previous accent
Antecedent accent information memory with cent information stored
Column, and based on the search result, a predetermined word
The accent is determined. Therefore, a natural accent
Can be quickly obtained. In addition,
In the case of, the antecedent accent information storage means and the consequent memory
The capacity of the stage can be reduced. [0125] [0126] [0127] [0128]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a configuration of an embodiment of a navigation system to which the present invention is applied. FIG. 2 is a block diagram showing a configuration example of a speech synthesizer 7 of FIG. FIG. 3 is a flowchart for explaining the operation of a word accent selection unit 13 in FIG. 2; FIG. 4 is a diagram showing pitches of sounds, phoneme information, and boundaries between prosodic words. FIG. 5 is a block diagram showing another example of the configuration of the speech synthesizer 7 in FIG. 1; 6 is a flowchart for explaining the operation of the word accent selection unit 33 of FIG. FIG. 7 is a diagram showing pitches of sounds, phoneme information, and boundaries between prosodic words. FIG. 8 is a flowchart for explaining a case where a consequent consisting only of numbers is treated as one prosodic word. [Description of Signs] 1 Input unit 2 Control unit 3 CD-ROM 4 Measurement unit 5 Output unit 6 Display unit 7 Speech synthesis device 11 Standard sentence reading prosody information selecting unit 12 Standard sentence pattern storage unit 13 Word accent search unit 14 Word accent Dictionary 15 Fixed sentence reading prosody information generation unit 16 Rule speech synthesis unit 33 Word accent search unit 34A Consequent element dictionary 34B Predicted element dictionary 34C Specific antecedent element dictionary 35 Fixed sentence reading prosody information generation unit

──────────────────────────────────────────────────続 き Continued on front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G10L 13/08

Claims (1)

(57) [Claims] [Claim 1] A map which is at least information on a map
A map information storage means for storing information, and measuring means for measuring the current position, the stored content and the measurement unit of the map information storage means
Output means for outputting predetermined information according to the measurement result;
A navigation system comprising said output means, a sound to output the predetermined information in the synthesized speech
It is configured to include a voice synthesis unit, the predetermined information, when containing a predetermined word, the speech synthesizing means, the predetermined words, prosody having 1 following accent nucleus
If there are multiple prosodic words that are units of
The prosodic word located at
When the case is made, the accent information on the accent of the preceding antecedent is stored.
An antecedent accent information storage means, a consequent storage means for storing the consequent, and an accent of the predetermined word,
Decision determined with reference to the information storage means and the consequent storage means
Wherein having means, accent determined by the determining means
Generating means for generating a synthetic sound corresponding to a predetermined word.
And the determining means determines whether the predetermined word has the consequent.
Is determined by referring to the consequent storage means.
And if the predetermined word does not have the consequent,
The accent type of a given word is determined to be a flat type, and
If the word has the consequent, the accent information of the antecedent
Information is retrieved from the antecedent accent information storage means.
Based on the search result of
A navigation system characterized by determining.