JPS6186799A - speech synthesizer - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a speech synthesis device that converts each word data constituting input document data into a speech signal using a synthesizer and utters it. The present invention relates to a speech synthesis device that can continuously utter adjacent word data according to the timing.

[Prior Art] A speech synthesis device that uses a synthesizer to convert document data such as English input from the outside via a transmission line into an audio signal and outputs speech corresponding to this document data from a speaker is, for example, It is structured as follows. That is, a voice memory is formed in a storage unit such as a ROM to store voice analysis data such as utterance frequency values, frequency component values of voice waveforms, and duration values of each frequency corresponding to each word data. There is. In addition, the input document data is decomposed into each word data using space symbols etc. inserted between each word data, and the voice analysis data corresponding to this word data is sequentially read out from the word memory and sent to the next synthesizer. A voice control section composed of a one-chip microcomputer or the like is provided. Then, when the voice analysis data of each word data is input, the synthesizer synthesizes an actual voice signal based on the above-mentioned multi-value of this voice analysis data and sends it to the speaker. Then, a voice corresponding to the input document data is uttered from the speaker.

[Problems to be Solved by the Invention] However, the speech synthesis device configured as described above has the following problems. That is, the voice control section generally determines the end of voice production for one word data by detecting an end signal added to the end of each voice analysis data. Then, after a certain pause time has elapsed from the time when this end signal was detected,
Start processing the next word data. Therefore, as shown in FIG. 9(a), when three word data are consecutive, the above-mentioned pause time To occurs between each uttered word. This pause time To is essential for distinguishing each word when speaking normal words.

However, when pronouncing a composite word such as a number, if fllTa is present during the pause, it becomes difficult to hear and the words become unnatural. To give a specific example, as shown in FIG. 9(b), when [the fortyone] is uttered, [the] and [f] are uttered.

A pause time To is inserted between [forty] and [one] in addition to between [forty] and [one].

In order to avoid the above problem, it is possible to calculate speech analysis data for all synthetic words such as numbers as one word and store it in the above-mentioned speech memory. Storing voice analysis data increases the storage capacity of a voice memory formed in a ROM or the like, causing a problem that the entire device becomes very expensive.

The present invention has been made based on the above circumstances, and its purpose is to make it possible to utter words that are adjacent to each other in sequence, such as in the case of compound words, etc.
To provide a speech synthesis device that is easier to hear and sounds more natural without increasing the storage capacity of a ROM or the like by removing the pause time that exists between uttered speech.

[Means for Solving the Problems] The speech synthesis device of the present invention includes a utterance time memory that stores each utterance time corresponding to each word data, and a timer that measures the elapsed time from the utterance start time of each word data. means, a first word memory for storing word data to be uttered first, and a second word memory for storing word data to be uttered next to the word data stored in the first word memory,
When it is determined that a pair of word data stored in the first and second word memories is to be uttered continuously, the timer measures the time while uttering the voice for the word data stored in the first word memory. The voice utterance corresponding to the word data stored in the second word memory is started when the elapsed time becomes immediately before the end of the utterance time corresponding to the word data being uttered stored in the generation time memory. This is what I did.

[Operation] With the speech synthesis device configured in this way, each word data constituting the document data inputted from the outside is sequentially stored in the first and second word memories. If the adjacent word data stored in the word memory is word data to be continuously uttered, the word data stored in the second word memory is stored in the second word memory immediately before the end of the utterance of the word data stored in the first word memory. Voice production corresponding to the word data is started. Therefore, these two words will be heard consecutively.

[Example] An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a speech synthesis device according to an embodiment. That is, 1 in the figure indicates various storage units.

This is a voice control section formed by a microprocessor that has a built-in I10 board and performs various arithmetic processing, and this voice control section 1 includes a ROM2. A synthesizer 3 that converts each input voice analysis data into an actual voice signal, and a counter 4 that measures the utterance time.
is connected. A clock signal CLK is inputted to this counter 4 from the outside, and an output signal is sent to the audio control section 1 as an interrupt signal INT. In addition, the synthesizer 3 stores the voice analysis data corresponding to each word data.
A speaker 7 is connected via an audio memory 5 formed of OM and a low-pass filter 6.

As shown in FIG. 2, the utterance time memory RO1 formed in the ROM2 stores an index IND for each word data displayed in a combination of each character code such as an ASCII code, and an index for this word data. A number N corresponding to the utterance time of the corresponding voice and an address A in the voice memory 5 in which voice analysis data corresponding to this word data are stored are respectively stored.

In addition, as shown in FIG. 3, in the voice memory 5, for each address A, a voice analysis such as a voice frequency value corresponding to each word data, a frequency component value of a voice waveform, a duration value of each frequency, etc. Data D is stored.

Furthermore, as shown in FIG. 4, the storage section formed of RAM or the like in the voice control section 1 includes a data buffer PRAI that temporarily stores document data input from the outside via the I10 port. A first word memory RA2 that stores data related to word data to be uttered first. A second word memory R that stores data related to word data to be uttered next.
A3° and first and second word memories RA2. R.A.
A flag area RA4 is formed to store a continuous utterance flag indicating that the pair of word data stored in 3 is to be uttered continuously. Each word memory RA2. In RAa, areas RA5, . RA6 and an area RA7°RAS for storing addresses Al and A2 are formed.

When the voice synthesizer is powered on and various initial processes are completed, the voice control section 1 is configured to execute the voice production process according to the flowchart of FIG. 5. That is, when the initial processing is completed, the document data processing shown in FIG. 6 is executed at Pl. In FIG. 6, first, the number N2 indicating the utterance time of the area RA6 of the word data to be uttered next stored in the second word memory RA3 of the storage unit and the address A2 of the area RA8 are read out.
An area RA5. for storing the utterance time and address of the word memory RA2. Each is stored in RA7. Next, the word data to be uttered next is read from the data buffer RA1, this word data is sent to the utterance time memory ROI shown in FIG. 2, the utterance time memory RO1 is searched, and the utterance time of the corresponding word data is searched. number N and address A indicating
is read out and stored in each area RA6 and RAS of the second word memory RA3. First and second word memories RA2
．． When the storage of each utterance time Nl, N2 and each address At, A2 of each word data in RA3 is completed, it is determined whether or not a pair of word data stored in each word memory should be continuously uttered. [−
], or the flow of the text. Then, only when it is determined that the word data should be continuously uttered, the continuous utterance flag in the flag area RA4 is set to 1.

Note that each word data is stored in the data buffer RA1 by interrupt processing. That is, when document data is input from the outside to the 110 port via the transmission path, the space symbol, [, ] symbol, [,
] symbols, etc. are detected, decomposed into word data, and sequentially stored in the data buffer RAI.

When the document data processing is completed at Pl of the main routine in FIG. Send A1 to synthesizer 3. Synthesizer 3, which received address A5,
The address A1 is sent to the audio memory 5, the audio analysis data D stored at the corresponding address A1 is read out, and preparations are made to output an audio signal corresponding to this audio analysis data D.

Next, in P2, check the contents of the flag area RA4, and if the continuous voice flag is not set to 1, after confirming that the audio signal is not being sent from the synthesizer 3 to the low-pass filter 6, output the audio signal to the synthesizer 3. At the same time as sending a start command, a countdown start command is sent to the counter 4.

Then, an audio signal is input from the synthesizer 3 to the speaker 7 via the low-pass filter 6, and the speaker 7 emits audio corresponding to the word data stored in the first word memory RA2. At the same time, a number N indicating the vocalization time set in the counter 4 in synchronization with the input of the clock signal CLK
The count number K is subtracted. Then, when the count value K reaches O, an interrupt signal INT is sent to the audio control unit 1.
is input. When this interrupt signal INT is input, it is determined that the vocalization of the word data stored in the first word memory RA2 has ended, and a predetermined pause IIf! After flTo has passed, the process returns to Pl and prepares to utter the next word data.

If the continuous utterance flag in the flag area RA4 is set to 1 at P2, the first and first word memories RA2.
It is determined that the two word data stored in RA 3 will be continuously uttered, and a command to start outputting an audio signal is sent to the synthesizer 3 and a command to start countdown is sent to the counter 4 in the same manner as described above. When it is detected that the count value K of the counter 4 has reached a specific value KM near 0 indicating the time immediately before the end of the utterance time, the utterance operation that is about to end is immediately stopped. After the continuous voice flag is reset to O, the process immediately returns to Pl without pausing the pause time To.

If the speech synthesizer is configured in this way, it will be as shown in Fig. 7 (
As shown in a) and (b), when it is determined that a pair of adjacent word data stored in the first word memory RA2 and the second word memory RA3 should not be continuously uttered, the continuous utterance flag is set. is reset to 0, and a pause time To is inserted between the utterance of the first word and the utterance of the second word. On the other hand, if it is determined that continuous utterance is required,
When the continuous utterance flag is set to 1 and there is a pause between the utterance of the first word and the utterance of the second word, f! ITn is removed and the final part of the utterance of the first word is cut;
Voicing of the second word begins immediately.

In the specific example of FIG. 8, the utterance of the first word [this] in [this is] and the utterance of the second word [i S
A regular pause FRBTo is inserted between the utterance of ``FRBTo''. On the other hand, the utterance of the first word [fOrtV] and the second word [one] in the compound word [fortyone]
Not only is there no pause time To between the utterance of [fortyl], but the final consonant of [y] in [fortyl] is erased,
the first consonant or vowel of one] is uttered. Therefore, the compound word [fOrt"l/One] is easy to hear and sounds natural without any discomfort.

Further, by performing program control in this manner, the overall cost of the speech synthesis device does not become particularly high.

[Effects of the Invention] As explained above, according to the present invention, when adjacent word data such as a composite word needs to be uttered consecutively, the pause time that exists between the uttered voices is removed. That's what I do. Therefore, the storage capacity of ROM etc.
It is possible to provide a speech synthesizer that is easier to hear and sounds more natural without increasing the amount of noise.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a speech synthesis device according to an embodiment of the present invention, FIG. 2 is a diagram showing a speech time memory formed in the ROM of the device, and FIG. 3 is a diagram showing a voice memory of the device. 4 is a diagram showing the memory formed in the audio control section of the device, FIGS. 5 and 6 are flowcharts showing the operation of the device, and FIGS. 7 and 8 are diagrams showing the operation of the device. FIG. 9 is a diagram showing each voice uttered by a conventional speech synthesizer. DESCRIPTION OF SYMBOLS 1... Audio control part, 2... ROM, 3... Synthesizer, 5... Audio memory, 6... Low pass filter, 7... Speaker, ROl... Voice time memory,
RAl...data buffer, RA2...first word memory, RAS...second word memory, RA4...
Flag area. Applicant's representative Patent attorney Takehiko Suzue Figure 1 Figure 2 Figure 3 Figure 4 Figure 6 Figure 7 (a) (b) Figure 8 (a) (b) (a) (b)

Claims (1)

[Claims] Speech analysis data corresponding to each word data of text data composed of a plurality of word data inputted from the outside is sequentially read from the speech memory formed in the storage unit and converted into an audio signal by a synthesizer to generate the text data. A speech synthesis device that outputs a voice corresponding to from a speaker, a speech time memory that stores each utterance time corresponding to each word data, and a clock means that clocks an elapsed time from a utterance start time of each word data. , a first word memory that stores word data to be uttered first, a second word memory that stores word data to be uttered next to the word data stored in this first word memory, and these first word memories. and a determining means for determining whether or not to continuously utter the pair of word data stored in the second word memory, and when the determining means determines that the pair of word data is to be continuously uttered, the first word When the elapsed time measured by the timer during the utterance of the voice corresponding to the word data stored in the memory is immediately before the end of the utterance time corresponding to the word data being uttered and stored in the generation time memory; and means for starting speech production corresponding to the word data stored in the second word memory.