JPS59160194A - 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 synthesis device. For example, the shape of the waveform is X (N+ 1) = K - The present invention relates to a technique that is effective for a voiced sound whose source is a waveform represented by (1)=1 (hereinafter referred to as an integral waveform).

[Background technology]

In conventional speech synthesis devices in which the sound source of voiced sound is an integral waveform, the integral characteristic (K) is fixed.

The inventors of the present application have discovered that there is a close relationship between the integral characteristic of the above-mentioned integral waveform and the sound quality of synthesized speech. Generally speaking, if you increase the value of the above-mentioned integral characteristic (K) for voice data of the same person, the synthesized voice will become soft (sound with suppressed high frequency components), and the above-mentioned characteristic (K )
If you reduce the value, the sound will become clearer (higher frequencies are emphasized).

On the other hand, the original sound differs greatly depending on gender, and there are also individual differences. Therefore, the inventor of the present application considered making the above-mentioned integral characteristic in the speech synthesis device variable and setting it to an integral characteristic according to the original sound.

[Purpose of the invention]

An object of the present invention is to provide a speech synthesis device that improves sound quality.

The above and other objects and novel features of this invention include:
It will become clear from the description of this specification and the accompanying drawings.

[Summary of the invention]

A brief overview of typical inventions disclosed in this application is as follows.

In other words, 1/2" data of the contents of the register whose initial value is set by the pinch cycle is formed, and 1/2" data is subtracted from the contents of the above register, and this is used as new data of the above register. It consists of a manually operated arithmetic circuit and uses this circuit loop to perform a plurality of arithmetic operations within one pitch period to form an arbitrary variable integral waveform.

〔Example〕

FIG. 1 shows a block diagram of one embodiment of the invention. Each circuit block surrounded by a broken line in the figure and a ROM (read-only memory) in which audio data is written are formed on a semiconductor substrate such as a silicon single crystal using known semiconductor integrated circuit manufacturing technology. be done.

The speech synthesis semiconductor integrated circuit device ICI forms synthesized speech using the well-known PARCOR method, although it is not particularly limited. Semiconductor integrated circuit device IC for speech synthesis of this embodiment
I is a voiced sound generator 4. A sound source section consisting of an unvoiced sound (noise) generator 5, a K stack 6, a digital filter 7, and a D/
It is composed of an A converter 8, a parameter interpolation circuit 3, a variable make decoding circuit 2, a serial/parallel converter 1 constituting an interface, and the like.

When the starting address of the voice to be generated from an appropriate control device (not shown) is applied to the voice data ROM (IC2), the voice data (characteristic parameter values) are extracted. ) is processed as follows.

The audio data (feature parameters) is
The parameter decoding circuit 2 decodes the data through the parameter decoding circuit 2.

Further, in the parameter interpolation circuit 3, the feature parameter values of the previous frame are interpolated every 2.511 Is, and smoothing is performed for discrete changes in each feature parameter.

Then, the pitch information is transmitted to the sound source section, the amplitude information and the PARC
The OR coefficient controls the digital filter 7. As a result, the original voice is restored and synthesized, and is output via the D/A converter 8 and external speaker SP.

FIG. 2 shows a block diagram of one embodiment of the voiced sound generator 4. As shown in FIG. In this embodiment, the following circuit is constructed in order to make the integral waveform as a voiced sound sound source liquid-resistant.

An initial value is set in the shift register SR according to the pinch cycle, and the value is successively changed by the next circuit loop. That is, the output of the shift 1 register SR is supplied on the one hand to the arithmetic circuit FA as human power. On the other hand, although not particularly limited, it is converted into a value of 1/2 to 1/16 by the delay circuit DL. This delay circuit DL is constituted by a D-type flip-flop, although it is not particularly limited, and delays the contents of the shift register SR, in other words, by delaying the contents by -1 pint to 4 bits.
Forms a value of 1/2 to 1/16. The signal is then selectively input to the arithmetic circuit FA through the multiplexer MPX. This arithmetic circuit FA includes the shift register S
The value inputted through the multiplexer MPX is subtracted from the value of R, and the output is input again to the shift register SR. The circuit loop described above performs a plurality of calculations within one pitch period.

The value of the shift register SR changed in this way is
At a predetermined timing, the data is taken out to the U-nochi circuit FF and outputs data constituting an integral waveform.

The operation of the voiced sound generator 4 of this embodiment will be explained with reference to FIG.

In the same figure (a), 1/4 is output by multiplexer MPX.
The case where the value of is selected is shown.

1 above from the initial value xO manually input to the shift register SR
By repeating the operation of subtracting the value of /4 and inputting it to the shift register SR, the value xO as shown in FIG.
It is possible to form data showing an integral characteristic such as ~x3...

The data xO to x3 . . . are taken out through the launch circuit F'F and output.

FIG. 6B shows a case where 1/8 is selected by the multiplexer M P X.

1 above from the initial value xO input to the shift register SR
By repeating the operation of subtracting the value of /8 and inputting it to the shift register SR, the value x as shown in FIG.
Data showing integral characteristics such as O-x 3... can be generated.

In the above embodiment, a case has been described in which the acquisition period of the launch circuit FF and the calculation period of the circuit loop are made to match. However, by speeding up the calculation period to 1/2, for example, By subtracting by 4+1/8, it is also possible to obtain an integral waveform of a cheek with more Kvi.

〔effect〕

(1) Through the above calculation operation, it is possible to obtain an integral waveform in which the contents of the shift register SR are successively changed, and its characteristics can be made variable by switching the multiplexer MPX, so that it can be adjusted to suit individual differences in the original voice. By forming the original synthesized speech based on the optimal integral characteristic, it is possible to improve the sound quality of the synthesized speech.

(2) According to (1) above, it is possible to produce synthesized speech with a quality that matches the preference of the user of the speech synthesis device, so that it is possible to improve the functionality.

(3) Since it can be configured with a digital circuit, it can be built into the semiconductor integrated circuit device ICI for speech synthesis.

Although the invention made by the present inventor has been specifically explained based on Examples above, this invention is not limited to the above Examples (although it is possible to make various changes without departing from the gist of the invention). For example, in addition to the above-mentioned delay circuit, a circuit that forms the 1/2n value of the contents of a shift register includes a circuit such as a barrel shifter that reads out the shift register in parallel and performs a downshift of n points. In this case, n-bit shift operation can be performed using one clock, so high-speed operation can be achieved.

Further, the switching operation of the multiplexer MPX for obtaining the integral waveform may be such that the value to be subtracted is made variable by switching the switching operation each time one piece of data x1 to xn is obtained. Such an operation makes it possible to greatly increase the variety of integral waveforms that can be obtained.

[Application field]

The speech synthesis method may be any method that restores the original speech using characteristic parameters, such as LS'P (line spectrum pair). As described above, the present invention can be widely used in speech synthesis devices that use integral waveforms as voiced sound sources.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the present invention, Fig. 2 is a block diagram showing an embodiment of the voiced sound generator, and 31g1(a) and (b) show its operation. FIG. 3 is a characteristic diagram showing an integral characteristic that is beyond description. 1. Interface, 2. Parameter decoding circuit, 3. Parameter interpolation circuit, 4. Voiced sound generator, 5
・・Noise generator, 6・・K stack, 7・・Digital filter, 8・・D/A converter, ROM・・Read only memory, SR・・Shift register, FF・・Launch circuit, FA・・Arithmetic circuit, DL・・Delay circuit, M
PX... Multiplexer agent Akio Takaran Figure 1 Figure 2 Figure 3 (t2-)

Claims (1)

[Claims] 1. In a speech synthesis device that uses an integral waveform as a voiced sound source, the sound source has a register whose initial value is set according to a pitch period, and the contents of this register form 1/2". A circuit loop consisting of a circuit, an arithmetic circuit that subtracts the above 1/2'' from the contents of the register and manually inputs it to the register, and a circuit loop that performs calculations multiple times within a pinch period, and extracts the contents of the register at a predetermined timing. including a launch circuit,
A speech synthesis device characterized in that it generates a variable integral waveform from the output of a launch circuit. 2. The speech synthesis device according to claim 1, wherein the speech synthesis device uses a partial autocorrelation coefficient as a characteristic parameter.