JPS6059398A - Voice 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 (Technical Field) The present invention relates to a speech synthesizer that reads information in the waveform region of a phoneme waveform from a storage area and synthesizes speech.
The present invention relates to a speech synthesizer that performs addition to superimpose DPCM-encoded symmetric phoneme waveforms and obtains speech output.

(Prior Art) When a human voice is a voiced sound, it is emitted when airflow from the lungs becomes a periodic impulse flow through the vocal cords and resonates with a hollow resonator called the vocal tract.

Figure 1 shows the waveform of a human voice.

As shown in Figure 1, the human voice has almost the same waveform repeated in a period called ``pitch 1''.

This is because, as mentioned earlier, the human voice is emitted by a quasi-periodic impulse flow, and the pitch period is equal to the interval of this impulse flow.

When attempting to synthesize such speech, various methods can be used depending on whether the information of the speech is stored in a format such as "shioto".

One method is to store the waveform of one pitch period of voice (this is called a phoneme waveform) in a storage area for various voices and use it as control information, and output voice by connecting the phoneme waveforms. There is a method to do this.

Figure 2 shows a configuration using this method.

1 is a storage area for storing phoneme waveforms of 1 period and 1 period for various voices, 2 is a storage area for control information such as pitch and amplitude multiplication for connecting phoneme waveforms,
3 is a '3 synthesis section that connects speech segments.

Speech is synthesized by the synthesizer 3 by connecting the voices stored in the storage area 1 using the control information stored in the storage area 2. However, in order to create a highly natural synthesized sound, the pitch of the voice, The size of the must be controlled in an articulated manner. The loudness of the voice is controlled by multiplying the phonemes in the storage area by a certain amount using amplitude multiplication information in the storage area. Furthermore, the pitch of the voice is controlled by pitch information in the storage area 2. However, the phoneme waveform in the storage area 1 does not necessarily match the pitch given by the control information 9 based on the length of the pitch period of the voice when the phoneme waveform is extracted.

Therefore, if the pitch of the control information is shorter than the speech segment length, the rear end of the speech segment is cut off, and if the pitch of the control information is longer than the speech segment length, the last value of the speech segment is extended and the control information is A phoneme with the same pitch length as . (See Figure 3) However, if the phoneme waveform stored in storage area 1 is cut in the middle, if the phoneme waveform is not sufficiently attenuated, discontinuity will occur between the phoneme waveform and the connected phoneme waveform, which will affect the synthesized sound. It has the disadvantage of causing shadows, and conversely, when the phoneme waveform is lengthened, the spectrum of the phoneme waveform is deformed, resulting in severe deterioration of sound quality.

(Objective and Summary of the Invention) The object of the present invention is to solve these drawbacks, and to solve the above problems, it is an object of the present invention to solve the above problems. By time-multiplexing code reproduction,
In addition, the adder used for superposition calculation of symmetric phoneme waveforms is ADPC.
By sharing this with the addition in the M-code regenerator, a speech synthesizer for synthesizing natural and high-quality speech is realized with a simple circuit configuration, and will be described in detail below.

(Premise of the Invention) As mentioned before, speech is produced by the resonance of the impulse flow of the vocal cords, and this can be replaced by an electric circuit.

In other words, speech is considered to be a combination of overlapping output waveforms of a resonant filter (hereinafter referred to as phoneme waveforms) corresponding to one impulse emitted by an excitation circuit corresponding to the vocal cords. This will be explained using FIG. 4.

IO in FIG. 4 is an impulse train generated by the excitation circuit. Here, the time interval between each impulse is the pitch period interval. 11 is a synthesized audio output waveform obtained by driving a resonant filter with the in-sense train generated by the excitation circuit. 12 is an impulse P of the impulse train generated by the excitation circuit.
This is a phoneme waveform when the resonant filter is driven by 1. Similarly, 13 to 17 are phoneme waveforms when the resonant filter 4 is driven by in/ya ruses P2 to P6. Since the drive impulse train lO emitted by the excitation circuit is the addition of impulses P1 to P6, according to the superposition theorem, the synthesized speech output waveform 11 is the sum of each phoneme waveform 12 to 174.
It is obtained by adding .

The phoneme waveform 12 shown in FIG. 4 is O before time point t1. After the time point t1, it has a property of attenuating with the passage of time and becoming 0 at an infinite time point.

In fact, in the case of a speech waveform, the phoneme waveform is considered to be almost O when 16 milliseconds have passed from the excitation point (1,).

Therefore, the reproduction process of the phoneme waveform 12 is performed at the excitation time point (tl
/ to 16 m') seconds is sufficient.

However, in order to start the reproduction process of the phoneme waveform 13 from the excitation time point t2, multiple reproduction processes are required because the reproduction process of the phoneme waveform 12 has not yet been completed. The required waveform reproduction multiplicity n is given by the following equation (1). In the case of normal voice, n of about 4 is sufficient. So n = 4
This will be explained below.

Next, as a code for reproducing the phoneme waveform, an ADPCM code with good encoding efficiency is used, and an even symmetric waveform that requires half the amount of information is used as the phoneme waveform.

A regenerator for reproducing symmetrical phoneme waveforms from a series of ADPCM codes has been proposed in Japanese Patent Application No. 56-1.85489, so a detailed explanation will be omitted.

In differential decoding processing, such as reproduction processing of AlTlPCM codes, when the reproduction processing is terminated after 16 milliseconds (128 sample periods), the final output value is held.

Therefore, the phoneme waveform conditions (conditions) are 0 before the excitation time point (condition 2). In order to make it 1 (0 at the infinite time point), the initial reproduction value must be O. It is necessary that the reproduction final output value is O.

Since the present invention deals with even symmetrical waveforms, if the initial reproduction value is O, the final reproduction value will be 0, which satisfies the condition.

(Example of the invention) FIG. 5 shows one embodiment of the present invention.

Here, the term "channel" or r ch J is used for numbering purposes for processes that are time-divisionally processed within one sample period.

Each part in FIG. 5 is as follows. 100 is the first
Ch input careno star, 101 is the second channel input register, 1
02 is the 3rd channel input register, 103 digits is the 4th channel input register, 104 and 105 are the selectors, and 106 is the ADPC.
The register 107 that stores the M i data Ln moves the pointer f3 using the value of the register 106 as an address. D
108 is an adder/subtractor, 109 is a selector, 110 is a selector, 111 is a first
ch pointer register, l12 is the second channel pointer register, 113 c is the third channel pointer register, 114 is the fourth channel, ]? Increnostar, 115 is selector, 1
16 is a pointer limiter that limits the pointer value Pn to a specific range and outputs it as a pointer limiter value ~ Pn'; 117 is a quantization step; quantization memory stores the step value X; 11B is a shift register; 119 is a Adder/subtractor, 120 is a register, 121 is a register for outputting synthesized speech, 122 is an output terminal, 135 is a first channel down count, 136 is a second channel
137 is the 3rd channel down counter, 138 is the 4th channel down counter, and 137 is the 3rd channel down counter. ch gounkaunk, 139 is the 1st c
b decoder, 140 is the second channel decoder, 14 is the third
ch decoder 142 is the fourth .ch decoder. ch decoder, 143
144 is a BUSY signal of each channel, 144 is a LEFT signal indicating reproduction of the left half of the symmetrical waveform reproduction, 145 is a start signal for waveform reproduction, and 150 is a controller.

The controller 150 is the first channel input renostar 10 described above.
0, second channel input register 101. ..., 3rd ch
It is connected to each circuit such as the decoder 14 and the fourth channel decoder 142 to control their operation, but the connections are omitted in FIG. 5 to avoid complicating the diagram.

FIG. 6 shows data used in the present invention to reproduce an even symmetrical phoneme waveform, and the pointer initial value DPn.

64 ADPCM codes Ln+, Ln2. Ln3
．． - + Consists of Ln64. Patent application 56-1
．． As stated in 85489, 16 milliseconds (128
In order to reproduce an even symmetrical waveform with sample period), 64 A
DPCM code is required. These 64 ADPCM codes are reproduced in forward and reverse directions to reproduce 128 phoneme waveforms.

Initial value data T) of the pointer P is data stored in the pointer at the start of symmetrical waveform reproduction.

Further, FIG. 7 shows the relationship between the playback processing and time assigned to each channel.

In Fig. 7, 200 is the composite output waveform, 20I is the chi V
The phoneme waveform reproduced by the reproduction process assigned to channel 1, 202 is the activation signal S T for channel 1,
20.9 is the BUSY signal BUSYI for channel 1,
204 is the LEFT Shingetsu LEFTI for channel 1, 205, 209, 213 are the phoneme waveforms reproduced by the reproduction processing assigned to channels 2, 3.4, 206
．． 210, 214 are activation signals ST of channel 2.3.4
2. ST3. ST4.207, 211.215 are BUSY signals for channel 2°3.4 BUSY 2, BUS
Y3, BUSY4.

:! 08.212, 216 is LE for channel 2, 3.4
TF signal i, EFT 2, LEFT 3, LE
FT, 4 signals.

Embodiments of the present invention will be described in detail with reference to FIGS. 5 to 7.

At time point t1, a starting signal 81'1 is applied to the controller 150 from the outside and the channel] is activated as shown in FIG.
The reproduction process of the phoneme waveform corresponding to 2 is started.

? -, the controller 150 starts the down counter 135
Set the value to 28. This value is 16 mIJ seconds 1
is the corresponding value. (Sampling period of output waveform is 125
(If it is a microsecond, 16 milliseconds is a 1:8 sampling period.) The decoder 139 outputs the output of the down counter 135 from
, outside I, ``°1'' is output as the BUSY 1 signal. If the output of the down counter 135 is still 65 or more, 1'' is output as the LBFT 1 signal. (Refer to Figure 7) Hereafter, data for waveform reproduction is read from the input register 10θ every sampling period (125 microseconds).
The symmetrical waveform reproduction processing unit 135 based on the ADPCM code is 1
It is reduced by The 64 ADPCM codes are played back in order from the beginning in the forward direction, and when it is finished, that is, after the 65th playback processing, the LEFT 1 signal becomes "0", and the 64 ADPCM codes are played back in order from the end.
Perform reverse playback processing. In this way, when the 12828th raw processing is completed, the down counter 135 becomes O, and the BUSY ] signal becomes "0".

Table 1 shows that the controller 150 inputs input registers (100, 100,
ADPCM code Ln stored in 101, IO2, 103), output by shift register 118), and LE
This figure shows the ADPCM waveform reproduction calculation processing (excluding pointer movement calculation) performed in the digital camera 120 and the adder/subtractor 119 using the FT signal 144.

Similarly, at time point t2, an activation signal ST2 is externally applied to the controller 150, and the reproduction process of the phoneme waveform of channel 2 (corresponding to FIG. 4, 13) is started.

At this time, the down counter is 1,? 6 is set to the value 128, and the BUSY 2 signal becomes "1". Thereafter, the same reproduction process as in channel 1 is performed, and after 128 sampling periods, the down counter 136 becomes 0 and the BUSY 2 signal becomes "o".

From time point t3 onwards, in channel 3, time point t4
From channel 4, time point t.

Similar control is performed for the 4-i channels.

In this way, reproduction processing is performed using the four channels and a synthesized audio output waveform is output at each sampling period, but the reproduction processing of each channel is performed in a time-division manner between one sampling time point. Here, the time relationship of processing within one sampling period is shown in Fig. 8(a), and its flowchart is shown in Fig. 8(b).
Shown in (c).

The processing within one embedding period is divided into five cycles, which are sequentially processed. It should be noted that most of the circuits required for the ADF'CM playback process processed in each cycle are shared, and among the components shown in Figure 6, all components that do not have a channel number added to their name are used in each cycle. These are commonly used in common areas, and these are collectively referred to as "shared parts."

(Cycle 1) The waveform reproduction processing assigned to channel 1 is performed by input register 100, pointer register 11, and down counter 1.
35, is performed using the decoder 139 and the shared part, and the result is stored in the register 120.

(Cycle 2) The waveform reproduction process assigned to channel 2 is input to the current register 101, pointer register 112, and down counter 1.
36, is performed using the decoder 140 and the shared part, and the result is stored in the register 120.

(Cycle 3) The waveform reproduction process assigned to the top and bottom 3 is carried out through the shared part of the input screen 102, pointer screen 113, down counter J37, and decoder 141, and the result is stored in the register 120. Be it.

(Cycle 4) The waveform reproduction processing assigned to channel 4 is performed by the input register register 103, pointer register 114, and down counter 1.
38, is performed using the decoder 142 and the shared part, and the result is stored in the register 120.

(Cycle 5) The value of the /nostar 120 is stored in the register 121 and used as PCM data for outputting synthesized speech.

The control in the above five cycles is shown in FIG. 8(b).

<c)・'D It is shown in the flowchart.

As described above, in this embodiment, the symmetrical phoneme waveform reproduction processing using the ADPCM code applied to each channel is performed by time-division processing using the constituent elements for each channel and the commonly used shared part. are doing. Also, a sequential addition X means (register 120) used for waveform reproduction of each channel and channel.
The adder/subtracter 119) also serves as an addition means for superimposing the phoneme waveforms, and has the advantage of being realized by a small circuit configuration. Furthermore, by changing the activation interval of each channel, the pitch period of the synthesized audio output can be easily changed. Furthermore, since one phoneme waveform starts from O and ends at I)O, it also has the advantage that no deterioration in sound quality such as waveform discontinuity due to addition of each phoneme occurs.

(Effects of the Invention) As described above, according to the present invention, it is possible to configure a speech synthesis circuit that can control the pitch period and outputs high-quality synthesized speech with a simple circuit configuration.

[Brief explanation of the drawing]

Figure 1 shows the speech waveform, Figure 2 is a block diagram of a conventional speech synthesizer, Figure 3 shows the waveform used to change the pitch, and Figure 4 shows the superposition of phoneme waveforms. FIG. 5 is a diagram showing one embodiment of the present invention, FIG. 6 is a diagram showing the f-event format of the phoneme waveform, and FIG. 7 is a diagram showing the phoneme waveform and S
Figure 8(a) shows the time relationship between the T times signal BUSY signal and LEFT signal. Figure 8(a) shows the time relationship of processing within one sample period time. Figures 8(b) and (c) show the time relationship of the T times signal BUSY signal and LEFT signal. 3 is a flowchart showing processing within one sample cycle time. 100... 1st. Channel entry Karen Star, 101...2nd
ch human power register, 102...3rd ch input renostar, 1θ3 4th. channel input register, 104... selector, 105... selector, 106... register, 107...
Monthly 9 Ink movement amount memory, 108... Addition/subtraction device, 10
9 selector, 110... selector, t711... 1st c
h pointerless star, 112...2nd ch pointerless star, 113, 3rd ch7]? Increnosta, 114
4th channel pointer register, 115 ・Selector, 1
16...Pointer limit, 117...Quantization memory,
118...Shift register, 119...Addition/subtraction unit, j
20・Register, 121...Register, 122・
・Output terminal, 135...1st channel down counter, 13
6... 2nd channel down counter, 137... 3rd channel
Down counter, 138...4th. Channel down counter, 139...1st channel decoder, 140...2nd c
h decoder, 14th 3rd channel decoder, 142...
4th channel decoder, 143・BUSY signal, 1.44・
...LEFT signal, 145...Start signal, 150...
controller. Patent Applicant: Oki Electric Industry Co., Ltd. Figure 6 (c) 1 Wallet, ν 3' Leakle 4 1-1-1 ( ), H, 1982 Patent Application No. 167534 2ヅKomei's Name Speech Synthesizer 6 Contents of Correction Details of the Attachment Contents of Correction (1) In the 6th line of page 10 of the specification, ``128 pieces'' is corrected to ``1128 sampling periods''. (2) In the same book, page 10, line 7, “Initial value data DI
) Correct "J" to "Initial value data DPn".

Claims (1)

[Claims] Corresponding to a plurality of channels, unit data including an I ink initial value and a plurality of PCM codes are manually inputted into a reproducing means, and an even symmetrical phoneme waveform is reproduced for each channel with a pitch cycle shift. In a speech synthesizer that adds the phoneme waveforms of each channel using an adding means and outputs the addition result as a synthesized sound, the phoneme waveform of each channel has a certain time length from the start of playback, and the initial value and final value are O. The waveform is evenly j-symmetrical, and the first-stage reproduction means is used for each of the first and second waves. 2 times for each channel within the sample time.
1. A speech synthesis device which performs reproduction processing in a time-division manner, and wherein said addition means shares a renostar and an adder/subtractor used in the Ail recording reproduction means and accumulates phoneme waveforms. vessel.