JP3603008B2 - Speech synthesis processor - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a speech synthesis processing device that utters a read sentence.
[0002]
[Prior art]
Japanese Patent Laying-Open No. 8-83270 discloses a configuration in which a speech synthesizer can specify speech-related data in text data to automatically change the speech from the composite data and output speech. . This is because the user specifies emotional speech data for text data in a speech synthesis process that tends to be a read-aloud tone without undulations, so that simulated speech with muffled emotions Is automatically uttered.
[0003]
[Problems to be solved by the invention]
However, in the above-described conventional method, a sentence whose speech tone changes frequently has a lot of display data relating to the speech tone change, which is complicated, and may be difficult to see.
[0004]
SUMMARY OF THE INVENTION It is an object of the present invention to improve the tone change rule so that it is easier to see and handle.
[0005]
[Means for Solving the Problems]
According to the present invention, an input sentence including a voice synthesis sentence to be uttered and voice quality switching information for designating and switching the voice to be uttered is sequentially read to extract a voice synthesized sentence and voice quality switching information, and the extracted voice synthesized sentence is subjected to voice quality switching. In a speech synthesis processing device that performs speech synthesis processing based on information and utters with a specified voice quality,
It has voice quality switching history storage means for sequentially storing the extracted voice quality switching information, and the read speech synthesis sentence utters based on the most recently stored voice quality switching information,
The voice switching information includes voice release information for releasing the voice switching information indicating the designated voice quality, when reading the voice quality release information, releases the voice switching information the most recently stored, and later sentences, the most A speech synthesis processing device characterized by uttering based on previous voice quality switching information stored one time before newly stored latest voice quality switching information .
[0006]
Further, the voice quality cancellation information of the present invention includes a line feed code.
Further, the voice quality cancellation information of the present invention is characterized in that it includes a period code or a reading code.
[0007]
According to the present invention, the input sentence includes a speech synthesis sentence and voice quality switching information, reads the input sentence in order, reads out the speech synthesis sentence and the voice quality switching information, and stores the extracted voice quality switching information in the voice quality switching history storage means. Keep it. Then, the read voice synthesis sentence is uttered based on the voice quality switching information stored most recently. Thus, the voice is changed and the voice is changed according to the voice quality switching information inserted in the input sentence.
According to the present invention, the voice quality switching information includes voice quality release information, and when the voice quality release information is read, the voice quality switching information most recently stored in the voice quality switching history storage unit is canceled and released. Thus, in the following, it is uttered based on the previous rounds of voice switching information stored in the voice switching history storage means. Therefore, for example, when the standard voice quality setting is first performed, the latest voice quality switching information stored most recently in the voice quality switching history storage unit is set at the beginning of the part where the voice quality is changed, and the voice quality release information is set at the end of the sentence. Thus, the subsequent sentence can be returned to the original standard voice quality, which is the previous voice quality switching information stored in the voice quality switching history storage unit one time before the most recently stored voice quality switching information .
[0008]
Further, in the present invention, by setting a code originally inserted into an input sentence, such as a line feed code or a punctuation mark, in voice quality cancellation information, display data relating to a tone change is displayed in a complicated manner as in the above-described related art. Can be prevented, and it can be prevented from being difficult to see.
[0009]
Further, the voice quality switching information of the present invention includes a question mark or an exclamation mark.
[0010]
Further, the present invention is characterized in that, when the question mark or the exclamation mark, which is the voice quality switching information, is read, the word of the immediately preceding speech synthesis sentence is uttered in the voice quality associated with the voice quality switching information.
[0011]
According to the present invention, a question mark or an exclamation mark is used as voice quality switching information. For example, when a question mark is read, the ending of the word immediately before the question mark is set or the word immediately before the exclamation mark is generated in a surprised tone. By doing so, it is possible to utter in a natural speech tone without inserting special speech tone change data.
[0012]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an embodiment of a speech synthesis processing device according to the present invention will be described in detail with reference to the accompanying drawings. The speech synthesis processing device of the present embodiment is realized by an information processing device such as a personal computer or a portable information terminal.
[0013]
FIG. 1 is a block diagram showing a schematic configuration of a speech synthesis device 1 of the present embodiment. The apparatus includes a CPU (central processing unit) 11, a ROM (read only memory) 12, a RAM (random access memory) 13, a storage device 14, a dictionary 15, an input unit 16, a display unit 17, a sound processing unit 18, and a speaker 19. Consists of
Next, an outline of the present apparatus will be described. The storage device 14 stores a sentence to be subjected to speech synthesis, and the user inserts sound quality switching information into the sentence using the input unit 16 and the display unit 17 to edit the sentence, and performs speech synthesis sentence and voice quality switching. Create input text with information. The specific method of inserting the voice quality switching information will be described in detail with reference to FIGS.
[0014]
The speech synthesis program is stored in the ROM 12, and the kanji reading and accent information are registered as data in the dictionary 15. The CPU 11 reads the input sentence from the storage device 14 in accordance with the program stored in the ROM 12, performs inflection with the sound processing unit 18 based on the data stored in the dictionary 15, and performs speech synthesis with the specified voice quality, Speak from speaker 19.
[0015]
FIG. 2 is a block diagram showing the configuration of the speech synthesis process of the speech synthesis device 1. The processing unit 22 sequentially reads out the input sentences 21 including the voice quality switching information from the storage device 14, performs voice synthesis processing, and utters a voice 28 with voice quality switching.
[0016]
The processing unit 22 includes a font sound quality correspondence storage unit 23, a voice synthesis sentence temporary storage unit 24, a voice quality switching history storage unit 26, and a voice synthesis processing unit 27. These are generated in the RAM 13 at the time of execution.
[0017]
The font voice quality correspondence storage means 23 stores fonts and voice quality switching information in association with each other, as shown as an example in FIG. For example, pictograms that resemble the robot's face shown in the font column are associated with voice quality switching information of “make a robot voice”. The period "." Is associated with voice quality cancellation information, and "!" And "?" Are voice quality switching information that utters the sentence before the appearance of each font with a surprised voice or a questionable voice. Is associated with.
[0018]
Although not normally displayed, a line feed code or a reading mark “,” included in text data may be set as voice quality release information.
[0019]
Next, a speech synthesis processing method will be described. The input sentence 21 is read and processed one character at a time. When the read font is not associated with the font voice quality correspondence storage unit 23, the read font is temporarily stored in the speech synthesis sentence temporary storage unit 24 as text data to be a speech synthesis sentence. The associated voice quality switching font is converted into the corresponding voice quality switching information 25, and the voice quality information is stored in the voice quality switching history storage unit 26 according to the voice quality switching information 25.
[0020]
In this way, a series of input sentences is read, and voice quality switching information is sent to the speech synthesis processing unit 27 together with the speech synthesis sentences stored in the speech synthesis sentence temporary storage means 24, and is uttered as speech 28.
[0021]
4 to 6 show the storage format and operation of the voice quality switching history storage unit 26. The voice quality switching history storage means operates like a stack, and voice quality switching information is pushed onto the stack in the order of reading, and popped off the stack by voice quality release information.
[0022]
The operation of the sound quality switching history storage means 26 will be described with reference to FIG. The sound quality switching history storage means 26 stores sound quality switching information in the order of “voice of an alien”, “voice of sumo wrestling”, and “voice of a grandmother” from the bottom, and stores the voice synthesis sentence temporary storage means. It is assumed that the information stored in 24 is reset and nothing is stored. Then, the input sentences are sequentially read out, and the speech synthesis sentences are accumulated as text data in the speech synthesis sentence temporary storage unit 24 until the sound quality switching information appears.
[0023]
The state 41 in FIG. 4 indicates a state when voice quality switching information "make a robot voice" appears. In the following description, the voice quality switching information may be abbreviated as voice quality information. When the voice quality switching information appears, the voice synthesis sentence stored in the voice synthesis sentence temporary storage means 24 together with the voice quality information of "make a grandmother's voice", which is the last information stored in the voice quality switching information history storage means 26, is output. The data is sent to the synthesis processing unit 27, and the voice synthesis sentence is uttered in the voice of the grandmother. Then, the voice quality information of "make a robot voice" is stored at the end of the sound quality switching history storage means 26. By doing so, the text after the voice quality switching information can be uttered in the voice quality of the robot.
[0024]
FIG. 5 shows processing when voice quality release information is given as voice quality switching information. When voice quality release information is given from the state of 51, the sentence stored in the voice synthesis sentence temporary storage means 24 together with voice quality information of "make a robot voice" which is the latest information accumulated so far. Is sent to the voice synthesis processing unit, and is uttered by the voice of the robot. Then, the voice quality information "make a robot voice" stacked on the top of the stack is deleted from the stack (52). By doing so, the sentence to be uttered next is replaced with the original voice quality information (that is, the previous voice quality switching history stored in the voice quality switching history storage unit 26 one time before the latest information “make a robot voice”). Information) , "Grandmother's voice".
[0025]
Also, as shown in FIG. 3, by setting the period “.” As the voice quality release information, the voice quality switching information stored in the voice quality switching history storage unit 26 is extracted one sentence at a time, and the voice quality switching history is stored for each sentence. Voices can be sequentially uttered based on the voice quality switching information stored in the storage unit 26.
[0026]
When the line feed code is set as voice quality cancellation information, a paragraph can be uttered as a group of speech synthesis sentences. When the reading point “,” is set as voice quality cancellation information, a sentence separated by a reading point can be used. It can be uttered as a group of speech synthesis sentences.
In addition, when the voice quality information of "make a robot voice" is stacked on the stack at 42 in FIG. 4, a plurality of, for example, two voice quality information pieces are stacked, thereby invalidating the voice quality release information immediately thereafter, and two sentences with the specified voice quality. It is also possible to utter.
[0027]
FIG. 6 shows a process when the immediately preceding sentence voice quality switching information is provided. As shown in FIG. 3, immediately before sentence voice quality switching information is associated with “!” And “?”, And in the state of 61 in FIG. 6, “surprising voice” associated with “!” When the sentence voice quality switching information of “make” is given, the voice information of “surprised voice” is added to the voice information of “grandmother's voice”, which is the last information loaded, and the speech synthesis text is added. The sentence is sent to the speech synthesis processing section together with the sentence stored in the temporary storage means 24, and is uttered with the surprised voice of the grandmother. The voice quality information of this "surprised voice" is not accumulated in the voice quality switching history means 26, and the voice quality switching history means 26 maintains the state as it is (62).
[0028]
FIG. 7 is a flowchart showing the operation of the present invention. As described above, the input sentence including the voice quality switching information is read out one character at a time, and is processed one character at a time according to the flowchart shown in FIG.
[0029]
First, it is determined whether or not the read character is voice quality switching information (step S701). In the case of voice quality switching information, utterance is made based on voice quality information stacked on the top of the voice quality switching history storage unit 24 as shown in FIG. (Step S702). Then, after deleting the speech synthesis sentence stored in the speech synthesis sentence temporary storage means 24, the input switching information is loaded on the voice quality switching history means 26, and the process returns to the original processing (step S703).
[0030]
If it is not voice quality switching information, it is determined whether it is voice quality release information (step S704). If it is voice quality release information, the latest voice information stored at the top of the voice quality switching history storage means 24 as shown in FIG. The voice is uttered using the voice quality information (step S705). On deleting the speech synthesis sentence speech synthesis Bun'ichi time storage means 24 deletes the voice information stacked on top of voice switching history storage unit 24, return to the original process, i.e. from the latest voice switching information it uttered by the voice switching information stored in one time before (step S706).
[0031]
If it is not the voice quality release information, it is determined whether or not it is the immediately preceding sentence voice quality switching information (step S707), and if it is the immediately preceding sentence voice quality switching information, it is stored at the top of the voice quality switching history storage means 24 as shown in FIG. The voice is then uttered in a voice quality obtained by adding the immediately preceding sentence voice quality switching information to the voice quality information, the information in the voice synthesis sentence temporary storage means 24 is deleted, and the process returns to the original processing (step S708).
[0032]
If it is not the immediately preceding sentence voice quality switching information, it is temporarily stored as normal text data in the speech synthesis sentence temporary storage means 24 (step 709), and thereafter, it is determined whether or not all the sentences have been completed (step 710), and the process is terminated. If not, return to the original processing. When the processing is completed, the voice synthesis sentence stored in the voice synthesis sentence temporary storage means 24 is uttered with the voice quality information loaded on the top of the voice quality switching history storage means 24, and the process is terminated (S711).
[0033]
8 to 14 show display examples when an input sentence is created by inserting voice quality switching information into a sentence to be subjected to speech synthesis processing. When the whole sentence is uttered with the specified voice quality, first, as shown in FIG. 8, the cursor 100 is placed at the head of the sentence. Next, as shown in FIG. 9, a menu display is displayed at that location, and a desired voice quality is selected there. By doing so, the associated font 101 stored in the font voice quality correspondence storage means 23 is inserted at the cursor position as shown in FIG. In this manner, the input sentence in which the voice quality switching information is inserted only at the head of the sentence is uttered with the specified voice quality.
[0034]
As other setting states, the standard voice quality setting is specified for the entire text, and the period code "." Is associated with the voice quality release information. As described above, the voice quality switching information is inserted only at the beginning of the text. If it does, only the first sentence, "Suddenly, we will gather at 5 o'clock today," will be uttered in the voice quality specified by the voice quality switching information, and then in the standard voice quality setting become. Such setting of the standard voice quality may be set so as not to be released by the voice quality release information and stored in the voice quality switching history storage unit 26, for example.
[0035]
FIG. 11 shows a procedure for switching the voice quality only in the section designated by the user. The cursor 100 is placed at the beginning of the section where the voice quality is to be switched (FIG. 11), and the end point is specified according to the existing text document section specifying means such as pressing the cursor key while pressing the shift key (FIG. 12). While the section is specified, the menu display is opened and the desired voice quality is selected (FIG. 13). By doing so, the font 141 associated with the voice quality switching information is inserted at the beginning of the voice quality switching, and the font 142 associated with the voice quality release information is inserted at the end point.
In this case, the beginning of the sentence, "Suddenly, but I will contact you." Is uttered in the voice quality of the standard voice quality setting, and the voice quality switching font 141 is provided before the next "Note!" For this reason, the voice information is switched to the corresponding voice quality information until "Caution!-Please confirm by telephone." Next, since there is a font 142 associated with the voice quality release information, the setting of the voice quality switching font 141 in front of the "note" is released, and the subsequent text is replaced with the corresponding voice quality switching information 141 or earlier. Will be uttered in the standard voice quality setting which is the voice quality of the user.
[0036]
As another condition, for example, when the voice quality switching font “switch to the voice quality of the robot” is specified before the voice quality switching font 141, the voice quality release font 142 and thereafter are uttered in the voice quality of the robot.
[0037]
【The invention's effect】
According to the present invention, by setting the voice quality cancellation information, it is possible to return to the original voice quality and utter. By associating a code such as a line feed code or a punctuation mark originally inserted into the text data as the voice quality release information, the display is not complicated and the display is easy to see.
Also, by changing the voice quality of the word immediately before the question mark or the exclamation mark is attached, the content can be conveyed reliably, and when listening, it is possible to encourage a person to listen carefully.
[Brief description of the drawings]
FIG. 1 is a schematic diagram showing a speech synthesis processing device 1 according to an embodiment of the present invention.
FIG. 2 is a block diagram illustrating a speech synthesis process.
FIG. 3 is an example of a storage format of a font voice quality correspondence storage unit 23;
FIG. 4 is a schematic diagram showing an example of an operation of the voice quality switching history storage means 26 when voice quality switching information is given.
FIG. 5 is a schematic diagram showing an example of an operation of the voice quality switching history storage means 26 when voice quality cancellation information is given.
FIG. 6 is a schematic diagram showing an example of an operation when the immediately preceding sentence voice quality switching information is provided to the voice quality switching history storage means 26;
FIG. 7 is a flowchart for explaining the operation of the present invention.
FIG. 8 is a schematic diagram showing an example of display content when voice quality switching information is inserted.
FIG. 9 is a schematic diagram showing an example of display contents when voice quality switching information is inserted.
FIG. 10 is a schematic diagram showing an example of display contents when voice quality switching information is inserted.
FIG. 11 is a schematic diagram showing an example of display contents when voice quality switching information is inserted.
FIG. 12 is a schematic diagram showing an example of display content when voice quality switching information is inserted.
FIG. 13 is a schematic diagram showing an example of display contents when voice quality switching information is inserted.
FIG. 14 is a schematic diagram showing an example of display content when voice quality switching information is inserted.
[Explanation of symbols]
1 Speech synthesis processor 11 CPU
12 ROM
13 RAM
14 storage device 15 dictionary 16 input unit 17 display unit 18 sound processing unit 19 speaker 21 input sentence 22 processing unit 23 font voice quality correspondence storage means 24 voice synthesis sentence temporary storage means 25 voice quality switching information 26 voice quality switching history storage means 27 voice synthesis processing Part 28 Voice 100 Cursor 101, 102 Font 142 associated with voice quality switching Font associated with voice quality cancellation

Claims (5)

The input speech including the voice synthesis sentence to be uttered and the voice quality to be uttered and the voice quality switching information for switching are sequentially read to extract the voice synthesis sentence and the voice quality switching information, and the extracted voice synthesized sentence is based on the voice quality switching information. In a voice synthesis processing device that performs voice synthesis processing and utters with a specified voice quality,
It has voice quality switching history storage means for sequentially storing the extracted voice quality switching information, and the read speech synthesis sentence utters based on the most recently stored voice quality switching information,
The voice switching information includes voice release information for releasing the voice switching information indicating the designated voice quality, when reading the voice quality release information, releases the voice switching information the most recently stored, and later sentences, the most A speech synthesis processing device that utters speech based on previous voice quality switching information stored one time before newly stored latest voice quality switching information . The speech synthesis processing device according to claim 1, wherein the voice quality cancellation information includes a line feed code. The speech synthesis processing device according to claim 1, wherein the voice quality release information includes a punctuation code or a reading code. The speech synthesis processing device according to claim 1, wherein the voice quality switching information includes a question mark or an exclamation mark. 5. The speech synthesis according to claim 4, wherein when the question mark or the exclamation mark, which is voice quality switching information, is read, the word of the immediately preceding voice synthesis sentence is uttered in the voice quality associated with the voice quality switching information. Processing equipment.

Images