JP3432336B2 - Speech synthesizer - Google Patents

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a VICS (Vehicle)
The present invention relates to a speech synthesizer used for Information and Communication System), and particularly to a speech synthesizer that enables smooth voice guidance in text synthesis.

[0002]

2. Description of the Related Art FIG. 9 is a view for explaining display data transferred to a vehicle in VICS. As shown in FIG. 9A, in the VICS, the display data is currently 1
Transferred in 5.5 characters x 2 lines.

[0003]

In VICS, since voice data is not transferred in addition to display data, when voice synthesis is attempted using the display data shown in FIG. 9A, As shown in Fig. 9 (b), fragmentary vocabulary synthesizing is used as in "Tomei Expressway Westbound accident Toyota → Okazaki traffic jam 7km", and smooth voice guidance cannot be provided, which is very difficult to understand. I have a problem.

Further, in the above speech synthesizing apparatus, text synthesis for converting an arbitrary sentence composed of written sentences into speech and synthesizing it, and recording / reproducing system synthesis for coding (and compressing) a speech waveform to obtain a synthesized speech. There is. In text synthesis, arbitrary sentences can be synthesized, but the processing is extremely complicated and requires a large CPU (Central P
rocessing Unit) Power is required and there is a problem in sound quality due to an error in the language processing unit.

On the other hand, the sound quality of the recording / playback synthesis system synthesis is good because the voice waveform data is held as it is in the memory or the like.
There is a problem that the number of synthesized words is very limited because it requires a very large amount of memory. Therefore, in view of the above problems, it is an object of the present invention to provide a speech synthesizer that is easy to understand, has no processing error, and does not require a large amount of memory.

[0006]

In order to solve the above problems, the present invention provides a speech synthesizer having the following configuration. That is, in the voice synthesizer that synthesizes the display information sent from the center into voice, the fixed part and the variable part
Applicable based on the above display information from multiple example sentences composed of
A variable part of the selected combined sentence example is selected.
The display information sent from the center.
A synthetic sentence example determination unit (12) that determines a sentence example, and an acoustic parameter generation unit (6) that combines the acoustic parameters of the fixed part and the acoustic parameters of the variable part to generate an acoustic parameter sequence of a sentence to be finally synthesized. ), And the word dictionary part (7) which holds the reading for the variable part and searches the determined synthetic sentence example for giving the reading to the acoustic parameter generating part (6) and the acoustic parameter generating part (6). An acoustic processing unit (8) that obtains a synthesized voice of an output sentence based on an acoustic parameter sequence and a phoneme voice waveform data, and a phoneme dictionary unit (9) that holds the phoneme voice waveform data.

The composite sentence example determination unit may determine the composite sentence example according to the type of the display information. The composite sentence example determination unit may determine the composite sentence example based on the number of parameters of the display information. The composite sentence example determination unit may determine a composite sentence example based on the type of display information and the number of parameters.

The composite sentence example determination unit may determine the composite sentence example based on the number of place names in the section information of the traffic information included in the display information. The synthesized sentence example determination unit may convert a phrase included in the display information that is not suitable as a spoken word into a phrase prepared for voice synthesis to perform voice synthesis. The composition sentence example determination unit may change the unit of composition for the distance information of the traffic information included in the display information according to the number of distances.

The composite sentence example determination unit may change the information on a minute-by-minute basis for the time information of the traffic information included in the display information into hours and minutes. The composite sentence example determination unit is V
The road traffic information from the ICS may be systematically classified into some patterns and used for voice synthesis. The synthetic sentence example determination unit may systematically classify the road guidance information from the VICS into several patterns and use them for voice synthesis.

The composite sentence example determination unit may add a number or code to each of the classified patterns for management. The synthesis sentence example determination unit may set a synthesis sentence example suitable for speech synthesis for each of the classified patterns. The composite sentence example determination unit may set a plurality of sentence examples to the composite sentence example of the same pattern.

The synthesis sentence example determining unit selects a synthesis sentence example in which the variable portion of the display information is fitted into a fixed portion suitable for voice synthesis based on the variable portion, and a sound of the plurality of fixed portions. A fixed phrase table unit storing parameters is provided. The fixed phrase table unit may hold, in advance, acoustic characteristic parameters including prosodic information and a pause for the fixed portion of the synthetic sentence example.

The fixed part table unit may cause the announcer to utter the synthetic sentence example with respect to the acoustic feature parameter, and hold the spontaneously uttered voice parameter extracted from the utterance data as an acoustic parameter. The word dictionary unit may store readings and accents for the variable portion, retrieve the variable words, and give the readings and accents to the acoustic parameter generation unit.

According to the voice synthesizing apparatus of the present invention, the variable portion of the display information is fitted into the fixed portion suitable for voice synthesis selected based on the variable portion to determine the synthesized sentence example, and the sound of the fixed portion is determined. By combining the parameter and the acoustic parameter of the variable part to generate an acoustic parameter string of a sentence to be finally synthesized, a fragmentary word string for display is converted to synthesize a speech with a natural sentence. Can be realized. Also, by exchanging the variable part, the variation of the composite sentence example increases. Therefore, it is possible to realize the inset combining for combining the VICS road information and the road guidance information with high quality. By converting a phrase included in the display information that is not suitable as a spoken language into a phrase prepared for voice synthesis and performing voice synthesis, it is possible to realize voice synthesis without discomfort as a spoken language. .

The road traffic information from the VICS is systematically classified into several patterns and used for speech synthesis, and the road guidance information is systematically classified into several patterns and used for speech synthesis, thereby It is possible to limit the number, and it is possible to reduce the number of synthetic sentence examples, to accommodate inset synthesis, and to reduce the memory capacity. By adding numbers or codes to the classified patterns and managing them, it is possible to realize simplification of sentence example management and sentence example selection processing. By setting a synthesized sentence example suitable for speech synthesis for each of the classified patterns and setting a plurality of sentence examples for the same sentence of the same pattern, it is possible to perform speech synthesis with an optimal sentence for each information. .

The fixed phrase table section holds acoustic characteristic parameters including pre-reading, prosodic information and poses for the fixed part of the synthetic sentence example, and further announces the synthetic sentence example for the acoustic characteristic parameters. By uttering the voices and holding the natural voiced voice parameters extracted from the voiced voice data as acoustic parameters, natural and natural voice synthesis can be performed.

The word dictionary section holds readings and accents for the variable parts, retrieves the variable words to be synthesized, and gives the readings and accents to the acoustic parameter generating section, thereby providing a more natural and comfortable voice. Synthesis is possible.

[0017]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing an example of a case where the voice synthesis apparatus according to the embodiment of the present invention performs inset synthesis. As shown in the figure, the voice synthesizer is a VICS.
An information receiving unit 2 for receiving traffic information transmitted from 1;
A display unit 3 for displaying traffic information and a composite sentence example determination unit 12 for determining a composite sentence example suitable for the received traffic information are provided. The composite sentence example determination unit 12 will be described in detail below.

FIG. 2 is a diagram showing an example of display information transmitted from the center of VICS 1 in FIG. 1 relating to (natural) traffic congestion. As shown in this figure, 15.5 characters x 2
In the line display data, the traffic jam display information includes route name information, direction information, section information, traffic jam level information, and distance information. FIG. 3 is a diagram for explaining inset combining by the combined sentence example determination unit 12 of FIG. 1 when inputting (natural) congestion display information. If there is display information from the information receiving unit 2,
Based on the type of display information, as shown in this figure (a), the composite sentence example determination unit 12 displays the display information of the (natural) traffic jam of the variable part in the fixed part for inset in this figure (b). Inset as shown. If the section information is only "Okazaki", it is fitted into a fixed part such as "Okazaki at the beginning" or "near Okazaki".

FIG. 4 is a diagram for explaining the number of parameters of the display information example of FIG. As shown in the figure, the number of parameters of the display information example is five, that is, a route name, a section, a direction, a degree of traffic jam, and a distance. Based on the parameter number 5 of the display information example, the composite sentence example determination unit 12 may determine that the display information example is (natural) traffic congestion display information and select the correct sentence example.

FIG. 5 is a diagram showing an example of display information transmitted from the VICS 1 shown in FIG. 1 and related to accident traffic congestion. As shown in this figure, the display information of accident traffic congestion includes route name information, direction information, event information, section information, traffic congestion degree information,
It consists of distance information. FIG. 6 is a diagram for explaining inset combining by the combined sentence example determination unit 12 in FIG. 1 when inputting display information of accident traffic jam. When there is display information from the information receiving unit 2, as shown in FIG.
The composite sentence example determining unit 12 fits the display information of the natural congestion in the variable portion into the fixed portion for fitting as shown in FIG.

FIG. 7 is a diagram for explaining the number of parameters in the display information example of FIG. As shown in this figure, the number of parameters in the display information example is the route name, section, direction, event, congestion degree,
There are five distances. The composite sentence example determination unit 12 may determine that this display information example is the display information of the accident congestion based on the parameter number 6 of the display information example and select the correct sentence example from the sentence example table.

In this way, it becomes possible to convert a fragmentary word string for display and synthesize speech with a natural sentence. Note that the synthesis example may be determined based on both the type of display information example and the number of parameters. This is to facilitate identification of similar display information examples. In the above description, VICS1
Although an example has been described as the display information transmitted from, a lot of information is actually transmitted. Composite sentence example determination unit 12
Systematizes the road traffic information from VICS1 in order to reduce the number of composite sentence examples and enable embedded composition.
It is classified into several patterns as follows.

[0023]

[Table 1]

[0024]

[Table 2]

[0025]

[Table 3]

Further, the road guidance information from the VICS 1 may be systematically classified into several patterns as follows.

[0027]

[Table 4]

As described above, by limiting the pattern, it is possible to reduce the number of composite sentence examples, to accommodate inset composition, and to reduce the memory capacity. The patterns classified as above are numbered A1 to A50,
You may manage by B1-B27 and other codes. In this manner, sentence example management and sentence example selection processing can be simplified.

The synthesis sentence example determination unit 12 sets a sentence example for speech synthesis for each of the patterns classified as described above, as follows.

[0030]

[Table 5]

[0031]

[Table 6]

[0032]

[Table 7]

Next, in Table 5, a plurality of sentence example patterns may be set for the same information, such as sentence example A1 and sentence example A5 or A2, A3, A6, A7. In actual use, this is to let the user select the most favorite of these sentence examples. In this way, it is possible to synthesize a speech with an optimum sentence for each information.

Here, the composite sentence example determining unit 12 is composed of the composite sentence selecting unit 4 and the fixed form table unit 5, and the composite sentence selecting unit 4 displays the display information of the knitted portion to the variable portion in Tables 5 to 7. The type and the number of parameters are determined, the fixed part table part 5 stores the inset part of the fixed part excluding the knitted parts of Tables 5 to 7, and the synthetic text selection part 4 selects the optimum part from the fixed part table part 5. Select an example sentence and insert the display information into it.

The fixed phrase table section 5 holds acoustic characteristic parameters for reading the fixed inset portion in advance and giving prosody information (accent / intonation), pause length and the like in a memory or the like. That is, with respect to the fixed portion, the reading of the sentence example and the acoustic characteristic parameter may be held in advance in order to perform highly natural speech synthesis. This acoustic feature parameter is
Information for giving intonation, accent, etc. at the time of synthesis in a sound processing process described later, such as a pause position, a cut of a breath paragraph, an accent phrase, etc. Regarding the acoustic parameter of the fixed part, a sentence example set for speech synthesis for each of the classified parameters, a plurality of sentence examples set for the same pattern,
For example, an announcer or the like may be uttered and the spontaneously uttered voice parameter extracted from the utterance data may be held as an acoustic parameter.

In this way, natural and natural voice synthesis can be achieved. Next, the composite sentence example determining unit 12 converts words that are not suitable as spoken words in the display information into prepared words and combines them as follows. FIG. 8 is a diagram illustrating an example of voice synthesis in which a part of display information is converted. This figure (a) shows the display information and this figure (b) shows the speech synthesis information. As shown in the figure, in the case of voice synthesis, the display of the directional information may be changed to the “upper and lower lines” which is often used only for “both directions”.
Specifically, change as shown in the table below.

[0037]

[Table 8]

Regarding the information display of the congestion distance, the unit is changed according to the number of distances in the case of voice synthesis, as shown in the table below. This is because the decimal value is difficult to hear with voice.

[0039]

[Table 9]

The simple travel time is changed in the case of voice synthesis as shown in the table below. This is because it is difficult to hear the voice only by the minute unit. For example, XX
Voice synthesis is performed as in time ΔΔ minutes.

[0041]

[Table 10]

Further, information on traffic accidents and accidents from the VICS 1 is organized and displayed, and voice synthesis is also organized as shown in the table below.

[0043]

[Table 11]

[0044]

[Table 12]

In this way, it is possible to realize speech synthesis without any discomfort as a spoken word. Returning to FIG. 1, in the subsequent stage of the synthetic sentence example determining unit 12, the acoustic parameters of the fixed part (entire sentence) output from the synthetic sentence selecting unit 4 and the acoustic parameters of the variable words obtained from the word dictionary described later are combined. , The acoustic parameter generation unit 6 that finally generates an acoustic parameter sequence of a sentence to be synthesized, and holds the reading for the variable portion of the sentence example divided into the fixed portion and the variable portion, and for the input text data, A word dictionary unit 7 for searching and giving a reading to the acoustic parameter generating unit 6 is provided. The word dictionary unit 7 may retain readings and accents (designation of accent patterns and accent kernels, etc.) with respect to the variable portion, retrieve the input text data, and give it to the acoustic parameter generation unit 6.

In this way, by exchanging the variable part, the variation of the composite sentence increases. Further, the speech synthesizer includes an acoustic processing unit 8 that obtains a synthetic speech of an output sentence based on the acoustic parameter string output from the acoustic parameter generation unit 6 and the phoneme speech waveform data, and the phoneme speech waveform data. The phoneme dictionary unit 9 held, the amplifier 10 connected to the sound processing unit 8, and the speaker 11 connected thereto.
And are provided.

In this way, the inset synthesis for synthesizing the road traffic information and the road guidance information of VICS1 with high quality can be realized.

[0048]

As described above, according to the present invention, the variable portion of the display information is fitted to the fixed portion suitable for speech synthesis based on the variable portion to determine the synthesized sentence example, and the acoustic parameter and the variable of the fixed portion are changed. Since the acoustic parameter string of the sentence to be finally synthesized is generated by combining with the acoustic parameter of the part, it is possible to convert a fragmentary word string for display and realize speech synthesis with a natural sentence. become. Also, by exchanging the variable part, the variation of the composite sentence example increases. Therefore, it is possible to realize the inset combining for combining the VICS road information and the road guidance information with high quality. Regarding words and phrases included in the display information that are not suitable as spoken words,
Since the speech synthesis is performed by converting the words and phrases prepared for speech synthesis, it is possible to realize speech synthesis without discomfort as a spoken word. Since the road traffic information from the VICS is systematically classified into several patterns and used for voice synthesis, and the road guidance information is systematically classified into several patterns and used for voice synthesis, the patterns can be limited. You can
It is possible to reduce the number of combined sentence examples, to accommodate embedded composition, and to reduce the memory capacity. Since a number or code is added to each of the classified patterns for management, simplification of sentence example management and sentence example selection processing can be realized. Since a synthesized sentence example suitable for voice synthesis is set for each of the classified patterns and a plurality of sentence examples are set for the synthesized sentence example of the same pattern, it is possible to perform voice synthesis with an optimal sentence for each information. The fixed part table part holds acoustic characteristic parameters including pre-reading, prosody information and poses for a fixed part of the synthetic sentence example, and further causes the announcer to utter the synthetic sentence example for the acoustic characteristic parameter. Since the extracted spontaneously uttered voice parameters are held as acoustic parameters, natural and natural voice synthesis can be performed. The word dictionary unit holds readings and accents for variable parts, searches the determined synthesized sentence examples, and gives the readings and accents to the acoustic parameter generation unit, which enables more natural and comfortable speech synthesis. .

[Brief description of drawings]

FIG. 1 is a diagram showing an example of a voice synthesizing apparatus according to an embodiment of the present invention, which performs inset synthesis.

FIG. 2 is a diagram showing an example of display information transmitted from VICS 1 of FIG. 1 relating to natural congestion.

FIG. 3 is a diagram illustrating inset combining by the combined sentence example determination unit 12 in FIG. 1 when inputting display information of natural traffic.

FIG. 4 is a diagram illustrating the number of parameters of the display information example of FIG.

5 is a diagram showing an example of display information transmitted from the VICS 1 of FIG. 1, which is related to an accident traffic jam.

6 is a diagram showing an example of display information transmitted from the VICS 1 of FIG. 1, which is related to an accident traffic jam.

7 is a diagram illustrating the number of parameters of the display information example of FIG.

FIG. 8 is a diagram illustrating an example of voice synthesis in which a part of display information is converted.

FIG. 9 is a diagram illustrating display data transferred to a vehicle in VICS.

[Explanation of symbols]

1 ... VICS (sender system) 4 ... Synthetic sentence selection section 5: Fixed phrase table section 6 ... Acoustic parameter generation unit 7 ... Word dictionary 8 ... Sound processing unit 9 ... Phoneme dictionary 12 ... Synthetic sentence example determination unit

─────────────────────────────────────────────────── --- Continuation of the front page (72) Inventor Kazuya Sako 1-2-2 Gosho-dori, Hyogo-ku, Kobe-shi, Hyogo Prefecture Fujitsu Ten Limited (72) Inventor Toshitaka Yamato 1 Gosho-dori, Hyogo-ku, Kobe-shi, Hyogo No. 2-28 Fujitsu Ten Limited (72) Inventor Nobuyuki Katae 1015 Kamitadanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Fujitsu Incorporated (72) Shinta Kimura 1015 Kamitadanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Fujitsu Incorporated (56) Reference JP-A-3-213075 (JP, A) JP-A-5-334292 (JP, A) JP-A-63-47798 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 13/04

Claims (17)

(57) [Claims] 1. A voice synthesizing apparatus for synthesizing display information sent from a center into a voice , comprising a plurality of synthetic sentence examples each including a fixed part and a variable part.
Select a suitable synthetic sentence example based on the display information, and select
Sent from the center to the variable part of the synthesized sentence example
A synthesis sentence example determination unit (12) that fits display information to determine a synthesis sentence example, and an acoustic parameter string of the sentence to be finally synthesized by combining the acoustic parameter of the fixed portion and the acoustic parameter of the variable portion. An acoustic parameter generation unit (6), a word dictionary unit (7) that retains readings for the variable portion, searches the determined synthesized sentence example, and gives the readings to the acoustic parameter generation unit (6). An acoustic processing unit (8) that obtains a synthetic speech of an output sentence based on an acoustic parameter string output from the acoustic parameter generation unit (6) and phoneme speech waveform data; and a phoneme that holds the speech waveform data of the phoneme. Dictionary section (9)
A voice synthesizer comprising: 2. The speech synthesis apparatus according to claim 1, wherein the synthesis sentence example determination unit (12) determines a synthesis sentence example according to the type of the display information. 3. The speech synthesis apparatus according to claim 1, wherein the synthesis sentence example determination unit (12) determines a synthesis sentence example according to the number of parameters of the display information. 4. The speech synthesis apparatus according to claim 1, wherein the synthesis sentence example determination unit (12) determines a synthesis sentence example based on the type of the display information and the number of parameters. 5. The speech synthesis according to claim 1, wherein the synthesis sentence example determination unit (12) determines the synthesis sentence example based on the number of place names of the section information of the traffic information included in the display information. apparatus. 6. The synthesis sentence example determining unit (12) converts a phrase included in the display information that is not suitable as a spoken word into a phrase prepared for voice synthesis to perform voice synthesis. The voice synthesizer according to claim 1. 7. The composition sentence example determining unit (12) changes the unit at the time of composition of the distance information of the traffic information included in the display information according to the number of distances.
The speech synthesizer according to claim 1. 8. The composite sentence example determining unit (12) according to claim 1, wherein the time information of the traffic information included in the display information is changed from minute-by-minute information to hour and minute. Speech synthesizer. 9. The composite sentence example determining unit (12) is a VIC.
Characterized in that road traffic information from S (1) is systematically classified into several patterns and used for voice synthesis.
The speech synthesizer according to claim 1. 10. The composite sentence example determination unit (12) is
The speech synthesizer according to claim 1, wherein the road guidance information from the CS (1) is systematically classified and used for speech synthesis. 11. The voice synthesizing apparatus according to claim 9, wherein the synthesis sentence example determining unit (12) manages each of the classified patterns by adding a number or a code. . 12. The voice synthesizing apparatus according to claim 9, wherein the synthesis sentence example determining unit (12) sets a synthesis sentence example suitable for voice synthesis for each of the classified patterns. . 13. The voice synthesizing apparatus according to claim 12, wherein the synthesis sentence example determination unit (12) sets a plurality of sentence examples in the same sentence example of the same pattern. 14. The synthetic sentence example determining unit (12) includes: a synthetic sentence selecting unit (4) for selecting a synthetic sentence example in which a variable portion of the display information is fitted into a fixed portion suitable for voice synthesis based on the variable portion. To store a plurality of embedded parts of the fixed parts and to allow the composite sentence selection unit 4 to select a composite sentence example,
The speech synthesis apparatus according to claim 1, further comprising: a given fixed phrase table unit (5). 15. The fixed-form part table part (5) holds acoustic feature parameters for giving reading and prosody information and a pause in advance for the fixed part of the composite sentence example. Item 14. The speech synthesizer according to item 14. 16. The fixed part table unit (5) causes an announcer to utter the synthetic sentence example for the acoustic feature parameter, and holds a spontaneously uttered voice parameter extracted from the utterance data as an acoustic parameter. The speech synthesizer according to claim 15, wherein 17. The word dictionary section (7) retains readings and accents of the words of the variable portion, retrieves the variable words to be synthesized, and gives the readings and accents to the acoustic parameter generating section (6). The voice synthesizing apparatus according to claim 1, wherein