JPH02195397A - Speech synthesizing device - Google Patents

Abstract

PURPOSE:To vary the voicing speed of a speech uniformly by providing a voicing speed standard value and setting the standard value with a command described in a text or an external switch. CONSTITUTION:This speech synthesizing device is equipped with a speed control means which controls the speed relatively to the preset standard speed. Namely, the number of one set of samples per voice parameter corresponding to the voicing speed standard value is determined and sent out to a synthesizer 8 to control the voicing speed. Then the text which contains command information specifying the voicing speed is inputted and analyzed by using a rule file 4, and then a voice pattern is generated by using a specific rule according to the analysis result. Then the voice pattern is outputted at the voicing speed relative to the preset standard speed. Consequently, the voicing speed of the synthesized voice is varied uniformly.

Description

Detailed Description of the Invention [Field of Industrial Application] The present invention relates to a speech synthesis device that analyzes input text and generates speech according to predetermined rules, and particularly relates to a speech synthesis device that analyzes input text and generates speech according to predetermined rules. The present invention relates to a speech synthesis device that changes the speaking speed of the voice.

[Prior Art] Conventionally, in a speech synthesis device that analyzes input text and generates speech according to predetermined rules, attempts have been made to control the speech rate of the generated speech using commands written in the input text. .

For example, (S3) shown in FIG. 2(a) is a speech rate of "3".
This is a command that specifies , and generates the sound ``Namamugi'' at this speed. Similarly, (S5) sets the speaking rate “5” to (
S7) is a command that specifies a speaking speed of "7" and generates the voices "namagome" and "namatamago", respectively, at this speed.

[Problems to be Solved by the Invention] However, in the conventional example described above, since the speaking speed that can be specified with each command is fixed, there are the following drawbacks.

(1) When you want to uniformly increase or decrease the speaking speed of the entire text,
That is, when it becomes necessary to make the entire text speak faster or slower, all speed control commands in the text must be rewritten.

(2) If the correspondence between the speed control command and the fixed speed is not linear, the development speed of the entire text cannot be uniformly increased or decreased.

These drawbacks will be explained using the example shown in FIG. If you try to have the text in Figure 2 (a) uttered quickly as a whole, the second
As shown in figure (b), command (S3) is changed to (S4)G
, :, Also, (S5) must be rewritten as (S6), and (S7) must be rewritten as (S8). Similarly, if the text in FIG. 2(a) is to be uttered slowly throughout, each text must be rewritten as shown in FIG. 2(C). Also, in this case, the command (S3
), (S4), (S5), and (S6) are respectively 3.3.4. O,,5,0,5,8(
97 seconds), the relative speed of command (S5) to command [S3) is 5.0/3.3 = 1.52
Therefore, the relative speed of command (S6) with respect to command (S4) is 5.8/4.0 = 1.45.

Therefore, when the text in Fig. 2 (a) is transcribed into speech, the relative speed between the sounds "Namamugi" and the sounds "Namagome" and the second
The sound “Namamugi” when the text in Figure (b) is transcribed into voice
The relative speed of the text and the voice ``namagome'' differs, and the entire text cannot be uttered uniformly and quickly.

The present invention was made to solve the above problems, and
It is an object of the present invention to provide a speech synthesis device that can uniformly increase or decrease the speaking speed of synthesized speech.

[Means for Solving the Problem] In order to achieve the above object, the speech synthesis device of the present invention has the following features:
It has the following configuration. That is, an input means for inputting a text including command information specifying a speech rate, an analysis means for analyzing the input text, and a generation means for generating a speech pattern using a predetermined rule based on the analysis result of the analysis means. and output means for outputting the speech pattern produced by the generation means according to the speech rate, the speech synthesis device comprising speed control means for controlling the speech rate as a relative value to a preset standard speed. Equipped with

Preferably, the standard speed of the speed control means is set by command information written in the input text.

More preferably, the standard speed of the speed control means is set by a switch.

Still preferably, the standard speed of the speed control means is set by the command information or switch, and
It is characterized by preferentially using one of them.

[Operation] In the above configuration, a text including command information specifying a speech rate is input, and after analyzing the text, a voice pattern is generated using a predetermined rule based on the analysis result.

Then, it operates to output the voice pattern in accordance with a speaking speed that is a relative value to a preset standard speed.

[Embodiment] Hereinafter, a first preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

Explanation of the structure (Fig. 1)> Fig. 1 is a block diagram showing the structure of the speech synthesis device in the first embodiment. A ROM that is a central control unit that takes charge of control and stores the processing procedures of the CPU, which will be described later.
It also includes a RAM used as a work area. 3 is a unit audio file, which stores unit audio for synthesis in the form of audio parameters. A rule file 4 stores rules for analyzing input text and generating a time series of audio parameters. 5 is an auxiliary memory, which stores system parameters and variables such as standard speed. 6 is an input buffer that stores input text. 7 is an output buffer that stores the generated audio parameter time series.

A synthesizer 8 synthesizes speech using the speech parameters sent from the central control section 2. A D/A converter 9 converts the audio synthesized by the synthesizer 8 into an analog signal. Further, 9 is an output terminal for outputting an analog audio signal.

Explanation of operation outline> In the above configuration, when text is input from the input terminal 1, the central control unit 2 controls it and sends it to the input buffer 6.
6 Next, the central control unit 2 extracts the text stored in the input buffer 6 one character at a time from the beginning, and if they are part of a sentence, the central control unit 2 extracts the text stored in the input buffer 6 one character at a time, and if they are part of a sentence, reads the analysis rules and analysis dictionary stored in the rule file 4. Perform text analysis using

Furthermore, according to the synthesis rules also stored in the rule file 4, appropriate unit voices stored in the unit voice files are selected and interpolated and connected to generate a voice parameter time series corresponding to the input text. Next, the generated audio parameter time series is temporarily stored in the output buffer 7, and audio parameters are extracted one set at a time from the output buffer 7 and sent to the synthesizer 8 in response to a parameter transfer request from the synthesizer 8. Then, the synthesizer 8 generates a digital audio signal from the audio parameters sent from the central control unit 2 and outputs it to the D/A converter 9. The D/A converter 9 receives this signal, converts it into an analog signal, and sequentially outputs it from the output terminal 10.

On the other hand, if the character string taken out from the input buffer 6 is a command, the command is analyzed and the analysis result is stored in the auxiliary memory 5. Note that the command in this example is “(
)” is a string of letters and numbers surrounded by symbols,
Two types of commands are defined: control commands and system commands.

(1) Control Command First, the control command is a character string consisting of one uppercase English letter and a number, and specifies the speaking speed, volume, strength of accent, etc. For example, (S5) is a command that specifies a speaking rate of "5", and (A4) is a command that specifies an accent strength of "4".

(2) System command Next, the system command is a character string consisting of four lowercase letters and a number, and is used to set various standard values of the speech synthesis system.For example, (sspe80) is the standard value for the speaking speed. is set to 80 samples.

Explanation of speaking speed (Figure 3) Among the commands mentioned above, the ones that are relevant to this embodiment are the speaking speed designation command (S) and the speaking speed standard value setting command (s).
s e p). These two commands control the number of audio samples per Irl frame j. The relationship between the number of audio samples and the speaking speed will be explained below.

The audio parameters stored in the unit audio file 3 of this embodiment are the unit audio uttered by a human being at a sampling frequency of 1.
The data was sampled at 0 KHz, analyzed with a frame shift width of 10 m5, and converted into audio parameters, as shown in FIG. 3(a). Also, a set of audio parameters output to the synthesizer 8 expresses the characteristics of a vocal waveform with a duration of 10 m5, and this 10 m5 audio section includes 10000X (10/1000) = 100 audio samples. ing. Therefore, when the synthesizer 8 uses this set of audio parameters and generates 100 audio samples, the output audio will have the same duration as the original audio and, as a result, will have the same speaking rate as the original audio. I will have it.

On the other hand, as shown in FIG. 3(b), when the synthesizer 8 generates 80 audio samples using a set of audio parameters sent from the central control unit 2, the generated The duration of the voice will be 0.8 times longer, and the speaking speed will be 1
．． It becomes 25 times.

Furthermore, as shown in FIG. 3(C), when 120 voice samples are generated, the duration length of the generated voice is 1.2 times greater, and the speaking speed is 0.83 times greater.

That is, in this embodiment, in response to the speaking rate designation command,
The speech rate is controlled by determining the number of samples per set of voice parameters for a standard speech rate value and sending this to the synthesizer 8. Then, the synthesizer 8 stores this number of samples, and uses the set of audio parameters sent from the central control unit 2 to generate audio samples for this number of samples. Thereafter, when the generation of a predetermined number of audio samples is completed, a request is made to the central control unit 2 to transfer the next audio parameter.

By repeating the above process, synthesized speech having a speaking rate corresponding to the speaking rate designation command is generated.

In addition, the voice speed designation command (S) in this embodiment is based on the formula (
1), the number of audio samples p per Irl frame j
(n) is determined.

p (n) = kXa (n) = (1) However, a (n) is determined by the following conditions.

2.25-0.25n (1≦n≦5) 1.75-0.
15n (6≦n≦9) Here, k is the standard value of the speaking rate expressed in units of samples, and the system command (ss
pe).

<Explanation of processing procedure (Figs. 4 to 6)> Next, the process when the text shown in Fig. 4 (a) is input from the input terminal 1 will be explained below according to the flowchart shown in Fig. 5. do.

First, in step Sl, when text is input, the text is stored in the input buffer 6, and step S
Analyze the input text in step 2. Next, it is determined whether the character string analyzed in step S3 is the above-mentioned command. If it is a command, the process advances to step S4, and it is determined whether it is a command to set a standard value of speaking rate.

Note that in this embodiment, the character string of the input text is “(ss
peloo)" and indicates the standard speaking rate setting command, the process advances to step S5, processes the following number string "100", and sets the numerical value 100 as the standard speaking speed in step S6. It is stored in the auxiliary memory 5.

Next, in step Sll, the delimiter of the character string is checked, and if it is not a delimiter, the process returns to step S2. Then, in step S2, the character string (S3) is processed, and since S is a speech rate designation command, the process advances to step S7. In step S7, according to equation (1),
Find the number of audio samples p (n) per rl frame j.

In this case, since n=3, p (n) is 2.25
-0.25X3 = 1.5, and since k = 1oo, p (n) = 150 (samples).Therefore, in step S8, a numerical value of 150 is set as the speaking rate and stored in the auxiliary memory 5. do. Also, step Sll
Now, as shown in FIG. 4(a), since the character string continues, the process returns to step S2. Then, the following character string "namamugi" is processed in the same way, and in step S9, a speech parameter time series corresponding to the speech "namamugi" is generated. Then, in step SIO, audio speed information is added to the beginning and stored in the output buffer 7.

At this time, the speaking speed stored in the auxiliary memory 5 is “15
0'', the numerical value 150 is added to the beginning of the audio parameter time series.

Thereafter, the synthesizer 8 requests the central control unit 2 to output voice parameters at intervals of the number of samples of the speech rate standard value set initially. In response to this output request from the synthesizer 8, in step S12, a set of audio parameters stored in the output buffer 7 is extracted and output. However, when speech rate information is added,
The synthesizer 8, which always sends the speech rate information to the synthesizer 8 and subsequently sends out a set of voice parameters, stores the information sent from the central control unit 2 if it is the speech rate, and continues to send it. The voice parameters are received and used to generate voice samples as many as the number of samples indicated by the stored speech rate information.

In this way, for the character string "Namamugi", the sound "Namamugi" is [
It is generated at a rate of 150 samples per 1'l frame J. In the same way, the character string "(S5)" is processed, a value of 100 is set as the speech rate, and the voice "namagome" is generated for the character string "namamugi" at a rate of 100 samples per rl frame j. Ru. Then, the string “(S7)
When "raw egg" is processed, the audio "raw egg" is generated at a rate of 70 samples per 11 frames j.

Through the above processing, the input text (a) shown in FIG.
The results of processing (b) and (c) are shown in Figure 6. The standard speaking rate set as a result of command processing is (a
) is 100 samples, (b) is 80 samples, and (C) is 120 samples, which are different, but the number of samples generated as a result of processing the command (S3) and the number of samples generated as a result of processing the command (S5) are different. The ratio is 1.5 in both cases.

Similarly, when comparing the command (S7) and the command (S5), the ratio is always constant and is 0.7.

Therefore, according to this embodiment, by simply changing the standard speaking rate setting command in the text, the speaking speed of voices generated by the same text can be uniformly increased or decreased. The correspondence relationship with is not affected by whether or not it is linear.

[Other Embodiments] Next, a second embodiment of the present invention will be described in detail with reference to the drawings.

<Description of the configuration (Figures 7 to 9)> As shown in the configuration diagram of Figure 7, a switch circuit 11 is added to the configuration of the first embodiment, allowing the user of the speech synthesis device to The speed of speech can be controlled using a switch.

The switch part of the switch circuit 11 has the appearance shown in FIG. 8, and increases the speech rate from the scale "1" to the scale "5", and decreases the voice speed from the scale "-1" to the scale "-5". reduce

Further, the switch circuit 11 includes a rotary switch 111 and an encoder 112, as shown in FIG. Then, the position information of the rotary switch 111 is encoded by the encoder 112 and sent to the central control section 2. Here, if the position of the scale of the switch is m and the output of the encoder is i, the relationship between the two is shown by equation (2).

i=m+5...(2) However,
m=-5,-4,...-1, O, l=5.

Therefore, the output i of the encoder is an integer value from "0" to "10".

(Figures 10 and 11) In the above configuration, when text is input from the input terminal 1, it is first stored in the input buffer 6. Next, the utterance rate is stored in the auxiliary memory 5. The initial value "0" is stored as the standard value, and the initial value "5" is stored as the voice speed control information. Then, the input text stored in the input buffer 6 is processed character by character from the beginning, and the following processing is performed depending on whether the content is part of a sentence or a command. .

First, if the content of the text taken out from the input buffer 6 is a command, its numerical part is digitized and stored in the corresponding location in the auxiliary memory 5.

For example, if the character string is (S3), the numerical value 3 is stored in the variable area n in the auxiliary memory 5, and the character string is (sspe
120), the value 1 is stored in the variable area k in the auxiliary memory 5.
Stores 20.

On the other hand, if the content of the text taken out from the input buffer 6 is part of a sentence, the same processing as in the first embodiment is performed. This process includes unit audio file 3 and rule file 4.
is used to generate an audio parameter time series and store it in the output buffer 7. At this time, the variable area k in the auxiliary memory 5 is checked, and if its inner bound is a value other than "O", from this value and the value of the variable area n in the auxiliary memory 5, the first embodiment The number of audio samples b (n) per fl frame J is determined using the explained equation (1). Then, this value is added to the beginning of the generated audio parameter time series, and if the content of k is "0", a negative sign is added to the value of the variable area n in the auxiliary memory 5. This process analyzes the sentences in the input text and generates the speech parameter time series, and if the speech rate standard value k has already been set to a value other than "0", the speech rate control information n is used to generate the speech parameter time series. The number of audio samples per "frame" of the audio parameter time series can be immediately determined. However, if the content of the voice rate standard value is "O", it means that only the voice rate control information n is added as rate information to the generated voice parameter time series. For example, the variable area in the auxiliary memory 5 is
In the state shown in FIG. 0 (a), "180" is calculated as the number of audio samples p (n) per "one frame" and added to the beginning of the generated audio parameter time series. In addition, in the state shown in FIG. 10(b), a value -3'', which is the content of the speech rate control information n with a negative sign, is added to the beginning.

Next, when the audio parameter time series stored in the output buffer 7 is transferred one set at a time in response to a parameter transfer request from the synthesizer 8, the following processing is performed depending on the code of the speech rate information added to the beginning. conduct.

If the speaking rate information is a positive value, it indicates the number of audio samples per rl frame j as described above, so it is immediately transferred to the synthesizer 8, and subsequent audio is processed at this number of samples, that is, at the speaking rate. generated.

Additionally, if the speaking rate information is a negative value, it means that the standard speaking rate was "O" at the time of the analysis process.
A standard speech rate value is calculated using the output i from the switch circuit 11, and the number of voice samples p (
i, n) are determined by equation (3).

However, n is a value obtained by removing the negative sign from the speaking rate information added to the beginning of the voice parameter time series, and indicates the value of the numerical part of the speaking rate control command.

p(L,n)=k(i)Xa(n)
= (3) where a (n) = 2.25-0. 25n (1≦n≦5)1.75-
0. 15n (6≦n≦9) Also, k(i) = 20X (10-i) (0≦i≦5) 10X
(15-i) (6≦i≦10) For example, if the starting information of the audio parameter time series is “-3” and the scale position m of the speaking rate setting switch is “4”, then from equation (2) i=9 and n=3, so from equation (3) p (i,
n) = 90, the numerical value 90 is sent to the synthesizer 8 as the number of audio samples per l frame, and the audio generated thereafter has this speaking rate.

That is, for an input text that does not have a standard speaking rate setting command in the text, the standard speaking rate can be set using the low reswitch 111, but if there is a standard speaking rate setting command in the text, the standard speaking rate setting command in the text can be set. The text after that does not depend on the position of the rotary switch 111,
Audio is generated according to the standard values set by the command.

The above processing will be explained using an example of input text shown in FIG.

When the text shown in FIG. 11(a) is input from the input terminal 1, it is stored in the input buffer 6 and then sent to the central control unit
First, the character string "(S3)" is processed as a speech rate control command, and the numerical value 3 is stored in the variable area n in the auxiliary storage device 5. Next, the character string "Namamugi" is analyzed, a speech parameter time series is generated, and stored in the output buffer 7, with speech rate information added to the beginning. The variable area in the auxiliary storage device S has an initial value of “O”.
Therefore, the speech rate information at this time is in the variable area n
The value is "-3", which is the value obtained by adding a negative sign to the speech rate control information stored in It is taken out and transferred to the synthesizer 8. At this time, the speech rate information added at the beginning is processed, but in this case, since it has a negative value, the output i of the switch circuit 11 is checked.

As a result, for example, the low reswitch 111 is
In the state shown in b), position information m=5, and n=
3, so the number of audio samples per rl frame is 75.
is transferred to the synthesizer 8, and the speech generated thereafter has this speaking rate.

That is, the voice "Namamugi" is generated at a rate of 75 samples per fl frame j. Similarly, the character string “(S5
) Namamugi” is processed and the voice “Namagome” is generated at a rate of 50 samples per Irl frame j, and the character string “(S7
) is processed and the voice ``raw egg'' is generated at a rate of 25 samples per fl frame.

That is, the text content is converted into audio and output at a speed 1.5 times faster than the normal speed.

Next, when the text shown in FIG. 11(b) is input from the input terminal 1, it is stored in the input buffer 6, and then the central control unit 2 first processes the character string (sspeloo) as a command. , the numerical value “100” is stored in the variable area k in the auxiliary storage device 5.

Next, the character string "(S3)" is processed as a command, and the numerical value "3" is stored in the variable area n in the auxiliary storage device 5. Subsequently, the character string "Namamugi" is analyzed as part of the sentence, and a time series of audio parameters is generated and stored in the output buffer 7. At this time, speaking rate information is added to the beginning, but in this case, the value in the variable area in the auxiliary storage device 5 is "100" and not "O", so the value in the variable area n is used together with the expression According to (1), the number of audio samples P (n) per rl frame j is immediately calculated, and the result is added as speech rate information.

That is, in equation (1), =100. Since n = 3, P (n) = 150, and the value 1 is used as speech rate information.
50 is added to the beginning.

Next, upon receiving a parameter transfer request from the synthesizer 8,
The audio parameters are extracted one set at a time from the output buffer 7 and output to the synthesizer 8. At this time, the speech rate information added at the beginning is processed, but in this case, the positive value "1" is used.
50'', it is directly output to the synthesizer 8 as the number of audio samples per ``1 frame''. Thereafter, the generated speech follows this number of samples, so the speech "Namamugi" is generated at a rate of 150 samples per rl frame j.

Similarly, the character string "(S5) Raw rice" is processed, and the speech "Namagome" is generated at a rate of 100 samples per rl frame. Also, the character string “(S7) raw egg” is processed,
The voice "Namatamago" is generated at a rate of 5o samples per Irl frame j.

That is, regardless of the output of the switch circuit 11, the voice is generated and output at the speed described in the text.

As explained above, according to the above-mentioned embodiment, the following effects can be obtained.

(1) The voice speed setting switch allows the user of the synthesizer to listen to the voice synthesized and output at his/her preferred speed.

(2) Since the speed is increased/decreased by the voice speed setting switch uniformly for the entire text, the speed control performed by commands within the input text is faithfully reproduced.

(3) If you insert the speaking rate standard value setting command into the input text and set a value other than "0" as the standard speaking rate, it will always be written in the text regardless of the position of the speaking rate setting switch. Since the voice is generated and output at exactly the same speed, important messages can be reliably conveyed.

[Effects of the Invention] As explained above, according to the present invention, a standard value for speech rate is provided, and this standard value can be set by a command written in the text or by an external switch. This has the effect of uniformly increasing or decreasing the speed.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing a first embodiment of the present invention, Fig. 2 is a diagram showing input text explaining a conventional technical example, and Figs. 3 (a) to (C) are samples of voice patterns. FIG. 4 is a diagram showing input text in the first embodiment. FIG. 5 is a flowchart showing processing in the first embodiment. FIG. 6 is a diagram showing processing results in the first embodiment. , FIG. 7 is a configuration diagram showing a second embodiment of the present invention, FIGS. 8(a) and (b) are external views of a rotary switch, FIG. 9 is a configuration diagram of a switch circuit, and FIG. 10 ( a) and (b) are diagrams showing the internal state of the auxiliary memory, and FIG. 11 is a diagram showing a processing example of the second embodiment. In the figure, 1... input terminal, 2... central control unit, 3...
・Unit audio file, 4... Rule file, 5...
Auxiliary memory, 6... Input buffer, 7... Output buffer, 8... Synthesizer, 9... D/A converter, 10...
- Output terminal, 11... Switch circuit. -G, Man Figure 3 Figure 4 Figure 11 Cause of rejection 8th (0) Figure 8 (b) Figure 9 (b) Figure 10

Claims (4)

[Claims] (1) An input means for inputting text including command information specifying the speaking speed, an analysis means for analyzing the input text, and a generation method for generating a speech pattern using a predetermined rule based on the analysis result of the analysis means. and output means for outputting the speech pattern produced by the generating means according to the speech rate, the speech synthesis device comprising: a speed control means for controlling the speech rate as a relative value to a preset standard speed; A speech synthesis device characterized by comprising: (2) The speech synthesis device according to claim 1, wherein the standard speed of the speed control means is set by command information written in the input text. (3) The speech synthesis device according to claim 1, wherein the standard speed of the speed control means is set by a switch. (4) The speech synthesis device according to claim 1, wherein the standard speed of the speed control means is set by the command information or a switch, and one of them is used preferentially.