JP3568972B2 - Voice synthesis method and apparatus - Google Patents

Description

[0001]
[Industrial applications]
The present invention relates to a speech synthesis method and apparatus for generating a synthesized speech from a character code string or prosodic information and a phoneme sequence.
[0002]
[Prior art]
2. Description of the Related Art Recently, various speech synthesizers have been developed which analyze a sentence mixed with kanji and kana, and synthesize and output speech information indicated by the sentence by a rule synthesis method. Such a speech synthesizer has begun to be widely used as a telephone introduction service in a banking business, a newspaper review system, a text-to-speech device, and the like.
[0003]
A speech synthesizer employing this type of rule synthesis method basically converts a voice uttered by a human into a predetermined unit, for example, CV (consonant, vowel, consonant), VCV (vowel, consonant). Vowels) and VCs (vowels, consonants), phonemic information obtained by analysis using a method such as LSP (line spectrum pair) analysis or cepstrum analysis is registered in a speech unit file, and the speech unit Synthesized voices are generated by generating synthesis parameters (phonological parameters and prosodic parameters) with reference to a file and generating a sound source and performing synthesis filtering processing based on these synthesized parameters.
[0004]
Conventionally, such a speech synthesizer requires dedicated hardware for processing in real time. The system configuration of this speech synthesizer is roughly divided into the following two types.
[0005]
In the first configuration, a host computer such as a personal computer (PC) converts a sentence mixed with Chinese characters into prosodic information and a phoneme sequence (language processing), and generates synthesis parameters, a sound source, and synthesis filtering using dedicated hardware. , D / A (digital / analog) conversion. On the other hand, the second configuration is such that all processes from generation of a kanji-kana mixed sentence to generation of speech are performed by dedicated hardware. In most cases, the dedicated hardware in any configuration is composed of an LSI or a synthesis LSI called a DSP (Digital Signal Processor) which performs a high-speed product-sum operation, and a general-purpose MPU (Microprocessor Unit).
[0006]
In such a dedicated device, a kanji-kana sentence is subjected to morphological analysis and dependency analysis using a word dictionary, and a phonetic symbol sequence composed of prosodic information and a phoneme sequence (prosodic information and phoneme sequences are represented by symbols and katakana. Stuff) are generated. A sequence of fundamental frequencies (prosodic parameters) is generated based on the prosodic information of the phonetic symbol sequence, phoneme parameters corresponding to the phonemes of the phoneme series are extracted from the speech unit file, and the phonemes are connected by connecting phonemes to generate phoneme parameters. You. At this time, if a symbol representing a pause section is included in the speech symbol string, both the prosody parameter and the phoneme parameter are set to “0” for the pause portion.
[0007]
The processing described above is performed by a general-purpose MPU (CPU), and the generated synthesis parameters (prosodic parameters and phonemic parameters) are input to the speech synthesis unit. A DSP or a synthesis LSI is used for the voice synthesis unit. In this voice synthesis unit, a DSP or a synthesis LSI inputs synthesis parameters for each frame period, performs sound source generation and synthesis filtering based on the parameters, and outputs the generated sound to the D / A converter for each sampling period.
[0008]
On the other hand, personal computers (PCs) and engineering work stations (EWSs) have also been able to perform the above processing by increasing their processing capabilities and by installing D / A converters, analog output units and speakers as standard. It is becoming possible to perform only software processing in real time. Recently, OSs (operating systems) capable of multitasking have begun to be adopted not only in EWS but also in PCs.
[0009]
However, most of such OSs do not guarantee real-time properties. That is, although there is no problem when the number of processing tasks other than voice synthesis is small, when the number of tasks increases, the CPU is used for processing other than voice synthesis, and voice synthesis may not be performed in real time.
[0010]
For these reasons, it is important to reduce the time required for speech synthesis as much as possible in order to synthesize speech in real time only by software processing under a multitasking OS such as PC or EWS.
[0011]
[Problems to be solved by the invention]
As described above, in the above-described conventional speech synthesizer, although the real-time property of the speech synthesizer is guaranteed in the dedicated device, a personal computer (PC) or an engineering work system employing an OS capable of multitasking is provided. In an apparatus realized by software processing of a station (EWS), speech synthesis was performed in real time when the number of processing tasks was small, but there was a problem in that real-time speech synthesis was not possible when the number of tasks was large.
[0012]
The present invention has been made in view of such circumstances, and an object of the present invention is to provide a speech synthesis method and apparatus capable of shortening the time required for speech synthesis and thereby performing speech synthesis in real time. .
[0013]
[Means for Solving the Problems]
The speech synthesizing method and apparatus according to the present invention focus on the fact that the pause unit only needs to output a signal (a signal having an output “0”) in which the synthesized speech to be output is silent and does not become sound. Excluding the pause symbol portion in the phonetic symbol sequence generated from the input character code sequence, generating a corresponding phonological parameter according to the phonological sequence and generating a prosodic parameter according to the prosodic information, this Pose sign Does not include partial parameters Phonetic and prosodic parameters From Synthetic filtering By processing Audio waveform data Produces Pose sign For the part generated Audio waveform data Based on the sampling frequency of an audio output device that outputs a corresponding audio from the pause data, pause data is generated separately so as to be silent for a time corresponding to the pause length. Synthetic filtering processing Generated by Audio waveform data To generate voice.
[0015]
[Action]
In the above configuration, A speech symbol string including a pause symbol is generated from the input character code string. Then, a phoneme parameter and a prosody parameter (synthesis parameter) are generated by excluding a pause symbol portion in the generated speech symbol string. This allows For live parts Synthetic filtering processing Is running Audio waveform data (Voice digital data) is generated, but for the pause part, Corresponding Contains no parameters From that , Synthetic filtering processing Is not executed. And instead, based on the sampling frequency of the audio output device, Pose sign Pause data is generated so that silence occurs only for the pause length of the part, Audio waveform data Is added to Therefore, this pause data is added Audio waveform data , The voice output including the pause section is correctly performed.
[0017]
As described above, according to the above configuration, the synthesis filtering that requires the longest time in the speech synthesis processing is performed only for the sound part, and for the silent part (pause part), the sound is performed during the corresponding time (period). Generate pause data for that time so that no signal I'll Therefore, the processing time of the speech synthesis can be shortened by the amount that the synthesis filtering of the silent section is not performed while outputting the speech including the silent section.
[0018]
【Example】
[First embodiment]
First, a first embodiment of the present invention will be described.
FIG. 1 is a block diagram showing a schematic configuration of the speech synthesizer according to the embodiment.
[0019]
The speech synthesizer shown in FIG. 1 includes an input unit 1 for inputting a character code string mixed with kanji or kana to be subjected to speech synthesis, and accent type, reading, and part of speech information for words and phrases to be subjected to speech synthesis. And a language processing unit 3. The language processing unit 3 analyzes a character code string input by the input unit 1 using the word dictionary 2 and performs language processing for generating a corresponding phoneme sequence and prosody information.
[0020]
The speech synthesis apparatus shown in FIG. 1 also includes a speech unit file 4 storing a cepstrum parameter group obtained by analyzing input speech in advance for each arbitrary speech unit, and a speech processing unit 3 for generating a speech segment file. A synthesis parameter generation unit 5 that generates a phoneme parameter (here, a cepstral parameter of a phoneme) according to the generated phoneme sequence and a prosody parameter according to the prosody information generated by the language processing unit 3, and a pause generation unit 6. Have. The pause generation unit 6 uses the pause data (D) representing “0” by the length based on the symbol indicating the pause section in the phoneme sequence and the prosody information given from the language processing unit 3 to the synthesis parameter generation unit 5. / A converter) is generated so that the output signal of the / A converter becomes “0”. This pause data is added to audio waveform data (audio digital data) output from the audio synthesizer 7 described below.
[0021]
The speech synthesis device shown in FIG. 1 also includes a speech synthesis unit 7 that generates a sound source and performs a synthesis filtering process to generate a synthesized speech based on the phoneme parameters and the prosody parameters generated by the synthesis parameter generation unit 5. , An audio device 8 and a speaker 9 for outputting sound. The audio device 8 includes a D / A converter that converts audio digital data output from the audio synthesis unit 7 and pause data (digital data) output from the pause generation unit 6 and added to the audio data into analog signals. A well-known configuration (not shown) such as an aliasing noise removal filter and a power amplifier is included.
[0022]
The speech synthesizer having the above configuration is realized by a personal computer (PC) or an engineering work station (EWS) that executes multitasking, and includes an input unit 1, a language processing unit 3, a synthesis parameter generation unit 5, The pause generating unit 6 and the voice synthesizing unit 7 (the sound source generation and filtering processing unit therein) are functional blocks realized by program processing (execution of a task for voice synthesizing process) of the CPU.
[0023]
Next, the operation of the speech synthesizer shown in FIG. 1 will be described with reference to the flowchart of FIG.
First, it is assumed that the input unit 1 has input a character code string containing kanji or kana to be subjected to speech synthesis, for example, "There will be a briefing session tomorrow. Thank you."
[0024]
The input unit 1 cuts out one sentence "Tomorrow will have a briefing." With a period "." (Step S1 in FIG. 2). One sentence cut out by the input unit 1 is given to the language processing unit 3.
[0025]
The language processing unit 3 collates the character code string (one sentence) cut out by the input unit 1 “Tomorrow there is a briefing session” with the word dictionary 2 and determines the speech synthesis target indicated by the input character code string. Find the accent type, reading, and part of speech information for the words and phrases that have become, determine the accent type and boundary according to the part of speech information, insert pause symbols, and convert to the form of reading of kanji kana mixed sentences, For example, a phonological sequence and a prosodic information (speech symbol string) of "A (Shi) awa.Setsukai <ga> / Arima (s) ... /" are generated (step S2 in FIG. 2). . Here, “＾” is an access position, “/” is a delimiter of an access phrase, “//” is the end of a sentence, “.” Is a symbol (pause symbol) indicating a pause section of 200 ms, () Represents a devoiced sound. The sampling frequency is 8 kHz.
[0026]
The speech symbol string “A (S) Tawa.Setsumekai <ga> / Arima (S) ... //” generated by the language processing unit 3 is given to the synthesis parameter generation unit 5.
[0027]
The synthesizing parameter generating unit 5 becomes, from the phonetic symbol string given from the language processing unit 3, “a (shi) tawa” which is a preceding accent phrase of the first appearing pause symbol (pause section symbol) “.”. A phoneme sequence is cut out, and cepstral parameters of a phoneme corresponding to the phoneme sequence are extracted from the speech unit file 4 to generate phoneme parameters (step S3 in FIG. 2). At the same time, the synthesis parameter generation unit 5 generates a prosody parameter according to the prosody information corresponding to the phoneme sequence.
[0028]
The synthesis parameter generation unit 5 provides the speech synthesis unit 7 with the phoneme parameter and the prosody parameter corresponding to the generated “A (Sh) Tawa”. Further, the synthesis parameter generation unit 5 gives the pause generation unit 6 one pause symbol “.” Following the “A (Shi) Tawa”.
[0029]
Conventionally, in the synthesis parameter generation unit 5, the phonetic parameters and the prosodic parameters of "A (shi) Tawa." Up to the pause symbol "." Note that “0” is set for both the phonological parameter and the prosodic parameter for the pause portion represented by the pause symbol, and the result is given to the speech synthesizer 7.
[0030]
When the speech parameter and the prosodic parameter (synthesis parameter) corresponding to “A (S) Tawa” are given from the synthesis parameter generation unit 5, the speech synthesis unit 7 inputs them and temporarily holds them. The voice synthesizing unit 7 generates a sound source and performs a digital filtering process in accordance with the input synthesis parameters of “A (Sh) Tawa”, thereby obtaining the voice digital data of “A (Sh) Tawa”. The input character code string "voice digital data shown in" tomorrow is "in the tomorrow will have a briefing. Thank you." (FIG. 2, step S4).
[0031]
The audio digital data generated by the audio synthesizer 7 is provided to the audio device 8. However, when the audio output by the audio device 8 is being performed, the generated audio digital data is given after the end of the audio output.
[0032]
On the other hand, the pause generation unit 6 generates 200 ms of pause data (“0”) based on one pause symbol “.” Following “A (S) Tawa” given from the synthesis parameter generation unit 5. Is generated (step S5 in FIG. 2). Here, since the sampling frequency is 8 kHz, the pause generator 6 generates 1600 pieces of pause data.
[0033]
When voice digital data (here, voice digital data of “A (S) Tawa”) is given from the voice synthesis unit 7 to the audio device 8, the pause generation unit 6 generates its own pause data (here, 1600 pause data corresponding to a 200 ms pause section indicated by one pause symbol “.” Following “A (S) Tawa” is given to the device 8 (step S6 in FIG. 2). As a result, 1600 pause data have been added after the audio digital data of “A (S) Tawa”.
[0034]
The audio device 8 converts the audio digital data of “A (S) Tawa” given from the speech synthesis unit 7 and the 1600 pause data given from the pause generation unit 6 and added to the data into D / The analog signal is sequentially converted into an analog signal by the A converter, and is output to the speaker 9 through the aliasing noise removal filter, so that the sound corresponding to “A (S) power” is output from the speaker 9 (FIG. 2 step). S7, S8).
[0035]
As described above, in the present embodiment, the pause unit focuses on the fact that the synthesized sound to be output is silent, and the audio device 8 only needs to output the signal of the output “0”. By generating pause data (“0” data) for a pause section based on the pause symbol “.” And excluding the synthesis parameter generation unit 5 from generating synthesis parameters, the speech synthesis unit 7 The composition filtering of the pause part is not performed. By doing so, the time required for the synthesis filtering in the voice synthesis unit 7 is reduced, and voice synthesis in real time becomes possible.
[0036]
In the related art, the synthesis parameter including the pause parameter is generated by the synthesis parameter generation unit 5 and given to the speech synthesis unit 7. Also, synthesis filtering was performed, and it took a long time for speech synthesis.
[0037]
Now, when the audio output from the speaker 9 by the audio device 8 is started, the synthesis parameter generation unit 5 determines whether or not the processing has been performed up to the end of one sentence indicated by “//” (step S9 in FIG. 2). ). If the end of one sentence has not been reached, as in this example, the following “Set-me-kai <ga> / Arima (s) ... /” is executed as described above from step S3 to step S8. Similar processing is performed. At this time, the pause generation unit 6 generates 8000 pieces of pause data corresponding to a 1000 ms (1 second) pause section indicated by five pause symbols “...
[0038]
Then, when the audio output corresponding to the above-mentioned “A (Sh) Tawa.” Is completed, the audio digital of “Set-me-kai <Ga> / Arima (S)” newly generated by the voice synthesizing unit 7 is output. The data and 8000 pieces of pause data generated by the pause generation unit 6 are sequentially supplied to the audio device 8 and used for the next audio output.
[0039]
At this time, since the processing of one sentence has been completed, the process proceeds from step S9 to step S10, and the control returns to the input unit 1.
In step S10, the input unit 1 determines whether or not the processing has been performed up to the end of the sentence. If the processing has not been completed as in this example, the process returns to step S1. In this step S1, the next sentence "Thank you." Is cut out by the input unit 1, and thereafter the same processing as described above is performed.
[0040]
By the way, in the present embodiment, in the example of "A (Shi) Tawa. Settsukaikai <ga> / Arima (S) ... //", six pause symbols, that is, 1.2 Given that there are pause symbols for seconds, if it takes only one second to generate, for example, one second of voice digital data by the synthesis filtering process of the voice synthesis unit 7, processing is performed for only 1.2 seconds of pause section. This means that the time has been reduced. However, the time required for pause data generation by the pause generation unit 6 is extremely shorter than the time required for synthesis filtering by the speech synthesis unit 7 and is ignored.
[0041]
In the first embodiment described above, the pause generator 6 is provided and the pause generator 6 generates the pose data. However, in general, a PC (personal computer) or an EWS (engineering work station) Has a structure in which a pause state is automatically continued unless audio digital data is written to the audio device (8). Therefore, a pause (pause section) can be generated without the pause generation unit 6, and the time can be arbitrarily controlled by setting a period during which nothing is written to the audio device (8). Further, a timer function is incorporated in the PC or EWS, and by using this function, a period in which nothing is written, that is, a pause section can be set.
[0042]
Therefore, a second embodiment in which a pause section is set using this timer function will be described below.
[Second embodiment]
FIG. 3 is a block diagram showing a schematic configuration of the speech synthesizer according to the embodiment. The same parts as those in FIG. 1 are denoted by the same reference numerals for convenience.
[0043]
The speech synthesizer shown in FIG. 3 differs from the speech synthesizer shown in FIG. 1 mainly in the following three points.
First, in place of the pause generation unit 6, a timer 11 and a timer setting unit 12 for setting a timer value to the timer 11 are used.
[0044]
Second, in FIG. 1, the pause symbol given from the synthesis parameter generator 5 to the pose generator 6 is given to the timer setting unit 12.
Third, the condition under which the audio digital data generated by the audio synthesizer 7 can be output to the audio device 8 is that the audio device 8 is not outputting audio and the timer value of the timer 11 is “0”. is there.
[0045]
The timer 11 is, for example, 1 ms per count (that is, the timer clock period is 1 ms), and does not become a value smaller than “0”.
The timer setting unit 12 is provided from the synthesis parameter generation unit 5 when outputting the ready signal 13 indicating that the audio device 8 is not outputting audio (the digital audio data can be input from the audio synthesis unit 7). The pause length indicated by the pause symbol (pause symbol string) is set in the timer 11.
[0046]
Next, the operation of the speech synthesizer shown in FIG. 3 will be described with reference to the flowchart in FIG.
First, as in the case of the first embodiment described above, it is assumed that the input unit 1 has input a character code string containing kanji or kana to be subjected to speech synthesis, "There will be a briefing session tomorrow. Thank you." I do.
[0047]
The input unit 1 cuts out one sentence "Tomorrow will have a briefing session" with a period "." (Step S11 in FIG. 4).
The language processing unit 3 collates the character code string (one sentence) cut out by the input unit 1 “Tomorrow there is a briefing session” with the word dictionary 2 and determines the speech synthesis target indicated by the input character code string. Find the accent type, reading, and part of speech information of the words and phrases that have become, determine the accent type / boundary according to the part of speech information, insert pause symbols, and convert to the form of reading kanji kana mixed sentences, The phoneme sequence and the prosodic information (speech symbol sequence) such as “A (Shi) Tawa. Setmekakai <Ga> / Arima (S) ... /” as described above are generated (FIG. 4). Step S12).
[0048]
The speech symbol string “A (S) Tawa.Setsumekai <ga> / Arima (S) ... //” generated by the language processing unit 3 is given to the synthesis parameter generation unit 5. Thus, the processing of step S13 described below is performed by the synthesis parameter generation unit 5 and the speech synthesis unit 7.
[0049]
That is, first, the synthesis parameter generation unit 5 determines, from the phonetic symbol string given from the language processing unit 3, “a (shi) ＾” which is a preceding accent phrase of a pause symbol (pause section symbol) “.” That first appears. The phoneme sequence “W” is cut out, and cepstral parameters of the phoneme corresponding to the phoneme sequence are extracted from the speech unit file 4 to generate phoneme parameters. At the same time, the synthesis parameter generation unit 5 generates a prosody parameter according to the prosody information corresponding to the phoneme sequence.
[0050]
The synthesis parameter generation unit 5 provides the speech synthesis unit 7 with the phoneme parameter and the prosody parameter corresponding to the generated “A (Sh) Tawa”. In addition, the synthesis parameter generation unit 5 provides the pause setting unit 12 with one pause symbol “.” Following the “A (S) Tawa”.
[0051]
The timer setting unit 12 inputs the pause symbol “.” Given from the synthesis parameter generation unit 5 and temporarily holds the pause symbol.
The speech synthesizing unit 7 inputs and temporarily holds the phoneme parameter and the prosodic parameter (synthesis parameter) corresponding to “A (S) Tawa” given from the synthesis parameter generation unit 5. The voice synthesizing unit 7 generates a sound source and performs a digital filtering process in accordance with the input synthesis parameters of “A (Sh) Tawa”, thereby obtaining the voice digital data of “A (Sh) Tawa”. The input character code string "digital data shown in" Tomorrow is shown "in" Tomorrow has a briefing session. Thank you. "
[0052]
When the voice synthesizing unit 7 generates the voice digital data of “A (S) Tawa”, the voice synthesizing unit 7 checks whether or not the audio device 8 is outputting voice by the ready signal 13 from the device 8 (FIG. 4). Step S14). If the voice is not being output, the voice synthesizing unit 7 proceeds to step S15. If the voice is being output, the voice synthesizing unit 7 determines that the voice is not being output (that is, the voice output ends and the ready signal 13 becomes true). wait.
[0053]
Here, since the audio device 8 is not outputting a voice, the voice synthesizing unit 7 proceeds to step S15. In step S15, the voice synthesizer 7 checks whether the value of the timer 11 (timer value) is “0”. If the timer value is “0”, the speech synthesizing unit 7 proceeds to step S16, and if not, waits for the value to become “0”.
[0054]
Here, since nothing is set in the timer 11, the timer value is “0”, and the voice synthesizing unit 7 proceeds to the next step S16. In step S16, the voice synthesizing unit 7 provides the audio device 8 with the audio digital data of “A (S) Tawa” generated in step S13 in order to output the voice from the speaker 9. As a result, in the same manner as in the above-described first embodiment, the output of the sound corresponding to “A (S) power” from the speaker 9 is started. At this time, the ready signal 13 from the audio device 8 is set to a false value indicating that audio is being output.
[0055]
Now, when the audio digital data of “A (S) Tawa” generated by the voice synthesizing unit 7 is given to the audio device 8, the synthesizing parameter generating unit 5 terminates one sentence indicated by “//”. It is determined whether or not the process has been performed (step S17 in FIG. 4). If the end of one sentence has not been reached as in this example, the speech synthesizer 7 and the audio device 8 are used for the following "Set-up <Ga> / Arima (s) ... /" Is performed in the same manner as in the above-mentioned "A (Shi) Power".
[0056]
In this step S13, the generation of the synthesis parameters of "set-up <ga> / arima (s)" and the generation of the audio digital data based on the synthesis parameters are performed, and the following five pause symbols ". Is given from the synthesis parameter generation unit 5 to the timer setting unit 12.
[0057]
The timer setting unit 12 receives the pause symbol “...” And holds it after one pause symbol “.” Subsequent to the previously held “A (S) Tawa”.
On the other hand, when the audio output of “A (S) Tawa” ends, the audio device 8 sets the ready signal 13 to true. Then, the timer setting unit 12 is the one following the oldest pause symbol (pause symbol string) at that time, that is, “A (Shi) Tawa”, among the pause symbols (pose symbol strings) held by the timer setting unit 12. The pause symbol “.” Is extracted, and the value “200” indicating the pause section 200 ms corresponding to the pause symbol “.” Is set in the timer 11. The operation of the timer setting unit 12 is realized on the PC or EWS by an audio output end interrupt process according to an interrupt (audio output end interrupt) generated in response to the audio output end.
[0058]
The timer 11 counts down every 1 ms.
When the voice synthesizing unit 7 generates the voice digital data of "Set-up <Ga> / Alima (S)" in the process of step S13, as described above, whether the audio device 8 is outputting the voice. It is checked whether or not it is (step S14 in FIG. 4).
[0059]
If the voice output for “A (S) Tawa” has already been completed, the voice synthesizing unit 7 proceeds to step S15, and if not completed, waits for completion.
Here, assuming that the audio output to “A (S) Tawa” has ended, the audio synthesizing unit 7 determines in step S15 whether the value of the timer 11 is “0”, that is, the audio output is It is determined whether or not a period set by the timer setting unit 12 (200 ms corresponding to one pause symbol “.” Subsequent to “A (S) power”) has elapsed after the end.
[0060]
Then, when the value of the timer 11 becomes “0”, that is, when the time of the pause section (200 ms) has elapsed after the end of the voice output, the voice synthesizing unit 7 generates the “Set-up <Gain>> / Arima @ (S) "to the audio device 8 to output audio from the speaker 9 (step S16 in FIG. 4).
[0061]
In this way, even if the audio output ends, the time until the value of the timer 11 becomes "0", that is, the pause section set by the timer setting unit 12 at the end of the audio output and corresponding to the pause section following the output audio. Until the time elapses, the output of the audio digital data to be output next is waited. The output of the audio device 8 is in a pause state from the end of the audio output until the next audio digital data is supplied, which is equivalent to generating a pause.
[0062]
In step S16, when the voice digital data of "Set-up <Ga> / Arima (S)" is provided from the voice synthesis unit 7 to the audio device 8, the processing of one sentence has been completed. Goes from step S17 to step S18, and the control returns to the input unit 1.
[0063]
In step S18, the input unit 1 determines whether or not the processing has been performed up to the end of the sentence. If the processing has not been completed as in this example, the process returns to step S11. In this step S11, the next sentence "Thank you." Is cut out by the input unit 1, and thereafter, the same processing as described above is performed.
[0064]
Also in the second embodiment described above, the pause unit is excluded from the target of the synthesis parameter creation by the synthesis parameter generation unit 5 so that the speech synthesis unit 7 does not perform the synthesis filtering of the pause unit. Therefore, the processing time related to speech synthesis can be reduced.
[0065]
Note that the present invention is not limited to the above-described embodiment. That is, in the embodiment, “0” is used as the pause data. However, depending on the specifications of the D / A converter in the audio device 8, the analog signal does not become “0” even if the “0” data is input. Therefore, it is not necessary to particularly limit the data to “0”, and digital data such that an analog signal becomes “0” may be used as pause data.
In short, the present invention can be variously modified and implemented without departing from the gist thereof.
[0066]
【The invention's effect】
As described above, according to the present invention, the data of the pause section is not included in the synthesis parameters (phonological parameters and prosodic parameters), and therefore, in the pause section, sound source generation and synthesis filtering requiring the longest time for processing are not executed. In addition, during the time (period) corresponding to the pause section, a signal that makes a sound is not output. Therefore, the speech filtering including the pause section is correctly performed, but the synthesis filtering of the pause section is not performed. The processing time required for speech synthesis can be shortened by the amount required, and speech synthesis can be performed by software processing under a multitask OS of a personal computer (PC) or an engineering work station (EWS). Thus, a great effect in practical use such as real-time operation can be obtained.
[Brief description of the drawings]
FIG. 1 is a block diagram of a speech synthesizer according to a first embodiment of the present invention.
FIG. 2 is a flowchart for explaining the flow of processing in the first embodiment.
FIG. 3 is a block diagram of a speech synthesizer according to a second embodiment of the present invention.
FIG. 4 is a flowchart for explaining the flow of processing in the second embodiment.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... input part, 2 ... word dictionary, 3 ... language processing part, 4 ... speech unit file, 5 ... synthesis parameter generation part, 6 ... pause generation part, 7 ... speech synthesis part, 8 ... audio device (speech output means) ), 9: speaker, 11: timer, 12: timer setting unit, 13: ready signal.

Claims (2)

In a speech synthesis method for synthesizing speech,
Generate a phonetic symbol string including a pause symbol from the input character code string,
A phoneme sequence that does not include the pause symbol portion is cut out by removing the pause symbol portion from the generated speech symbol sequence, a corresponding phoneme parameter is generated according to the cut phoneme sequence, and prosody information corresponding to the phoneme sequence is generated. Generates prosodic parameters according to
A phoneme parameters and prosodic parameters said generated from phoneme parameters and prosodic parameters do not include the parameters of pose symbol portion, and generates a speech waveform data by combining filtering process,
Based on the sampling frequency of an audio output device that outputs a corresponding audio from the generated audio waveform data, the pause symbol portion excluded from the generation of the phonological parameters and the prosody parameters is a time corresponding to the pause length. A speech synthesizing method characterized by separately generating pause data that causes silence, and adding the pause data to the speech waveform data generated by the synthesis filtering process to generate speech. In a speech synthesizer that synthesizes speech,
Language processing means for generating a speech symbol string including a pause symbol from the input character code string;
A phoneme sequence that does not include the pause symbol portion is cut out by removing the pause symbol portion from the generated speech symbol sequence, a corresponding phoneme parameter is generated according to the cut phoneme sequence, and prosody information corresponding to the phoneme sequence is generated. Synthesis parameter generation means for generating a prosody parameter according to
A phoneme parameters and prosodic parameter generated by the synthesis parameter generating means, the phoneme parameters and prosodic parameters do not include the parameters of the pause symbol portion, a voice synthesizing means for generating a speech waveform data by combining filtering process,
Audio output means for outputting a corresponding audio from the audio waveform data generated by the audio synthesis means,
Based on the sampling frequency of the voice output means, pause data is generated by the voice synthesis means so as to be silent for a time corresponding to the pause length of the pause symbol part excluded from the generation of the phonological parameters and the prosodic parameters. Pause generating means for adding to the generated audio waveform data,
A voice synthesizing apparatus, wherein a corresponding voice is output by the voice output unit from the voice waveform data to which the pause data is added by the pause generation unit.

Images