JPH0117600B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a speech synthesis device. A speech synthesis device that uses partial autocorrelation coefficients as feature parameters has a configuration as shown in FIG. In the figure, 1 is a sound source section, 2 is a digital filter, 3 is a parameter supply circuit, 4 is a digital-to-analog conversion circuit, 5 is an amplifier circuit,
6 is a speaker.

In this type of speech synthesis, voiced sounds are formed based on impulses with a predetermined pitch period, and unvoiced sounds are formed based on random noise (white noise). Accordingly, the sound source section 1 includes an impulse generation circuit 1A and a random noise (random pulse) generation circuit 1B. Note that FIG. 3 shows an example of an impulse output from the impulse generating circuit 1A. In the figure, the horizontal axis t is time.

By the way, the vowel part in the audio waveform is 1KHz.
Since major energy exists in the first formant region below, relatively high calculation accuracy is required for this low frequency component when extracting the partial autocorrelation coefficient. In order to improve this problem, a method can be adopted in which a partial autocorrelation coefficient is extracted from a difference signal of an input waveform (in analog terms, a differentially processed signal).

In this case, integration processing is required when synthesizing speech using the extracted partial autocorrelation coefficients.
Such an integration process can be performed, for example, by providing an integration circuit including a resistor and a capacitor on the output side of the digital-to-analog conversion circuit 4.

However, in this case, the following problem occurs.

That is, it is relatively difficult to integrate a circuit including a capacitor such as an integrating circuit into an integrated circuit (IC). Therefore, for example, when circuits 1 to 5 shown in FIG. 1 are integrated into an IC, resistors, capacitors, etc. are required as external components of the IC.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a speech synthesizer that does not require a capacity for integral processing.

This invention provides that the filter used for synthesis is linear, and that by appropriately setting the sound source waveform accordingly, integration processing can be performed without using an integrating circuit consisting of resistance and capacitance. This work is being done with a focus on what can be done. That is,
In this invention, a waveform obtained by integrating impulses, which has been conventionally used as sound source information, is used as sound source information.

Now differential processing D(Z)=1−kZ ^-1 , integral processing I
If (Z)=1/(1-kZ ^-1 ) and 0<k1, then
The integral waveform i(t _o ) (hereinafter referred to as the integral sound source) is expressed by the following formula:
(1) and (2).

i(t _o )=1/2πjI(Z)Z ^n-1 dZ =1/2πjZ ⁿ /Z−kdZ ………(1) ∴i _o =i(t _o )=k ⁿ (n=0, 1 , 2...) ......(2) i _o in equation (2) is the first term 1 as shown in Figure 4,
This is an example of a geometric ratio of common ratio k.

Here, if k of D(Z) and k of I(Z) are set to different values (k ₁ for D(Z) and k ₂ for I(Z)), the overall characteristic is expressed by the following equation (3). ), (4).

F(Z) = D(Z)・I(Z) = (1-k ₁ Z ^-1 )/(1-k ₂ Z ^-1 ) ......(3) f _o = 1/2πjF(Z)Z ^n-1 dZ=1/2πj [1-k ₂
−k ₁ /Z−k ₂ ]Z ^n-1 dZ=(k ₂ −k ₁ )k ₂ ^n-1 ………(4) This is the differential output of the coefficient k ₁ of the integral output {k ₂ ⁿ }. I can think about it. Therefore, if k ₁ > k ₂ , the whole shows differential characteristics, and if k ₁ < k ₂ , it shows integral characteristics, and the frequency characteristics of the spectrum outline depend on the relative magnitude relationship of k ₁ and k ₂ . It is clear that the slope of can be controlled.

A signal as shown in equation (2) can be generated, for example, by an impulse generating circuit and an appropriate arithmetic circuit or arithmetic operation that receives the output thereof.
However, when such calculations are used, there is a problem that the circuit configuration becomes complicated.
Therefore, according to the present invention, although not particularly limited, the above signals are set in advance in a read-only memory (ROM).

FIG. 2 is a block diagram of a speech synthesizer according to an embodiment of the present invention. In the figure, 1 is a sound source section;
2 is a digital filter, 3 is a parameter supply circuit, 4 is a digital-to-analog conversion circuit, 5 is an amplifier circuit, and 6 is a speaker. Circuits 1 to 5
is converted into an IC using known IC technology.

The sound source section 1 includes an integral sound source 1A and a random noise (random pulse) generation circuit 1B.

The integral sound source 1A is composed of a read-on memory _A1 and a readout control section _A2 . The read-only memory _A1 has a plurality of storage addresses selected by the read control unit _A2 . In the read-only memory _A1 , data corresponding to the levels of the signals _i1 , _i2 , _i3, etc. shown in FIG. 4 are written in order from the first address.

Readout control unit A ₂ is parameter supply circuit 3
Each address of the read-only memory is sequentially selected starting from the first address.

Although not particularly limited, in this example,
In order to make the configuration of the control unit _A2 relatively simple and to simplify the control, the address of the read-only memory is fixed regardless of the signal cycle (pitch cycle) supplied from the parameter supply circuit ₃ to the control unit A2. It is changed sequentially at fixed time intervals determined by the configuration of the control unit _A2 .

The control unit _A2 is configured to instruct the leading address of the read-only memory _A1 every time a signal is supplied from the parameter supply circuit 3. Control part A ₂
is also configured to output sequential address signals only once in one pitch period. This ensures that the data corresponding to i ₁ , i ₂ , and i ₃ in FIG. 4 are generated only once in each pitch period.

In the speech synthesizer shown in FIG. 2, the parameter signal supplied from the parameter supply circuit 3 causes the above-mentioned data signal outputted from the sound source 1A in the case of a voiced sound, and the data signal output from the circuit 1B in the case of an unvoiced sound.
Random noise output from each is taken into the digital filter 2. A predetermined calculation is performed by the digital filter 2, and the result of the calculation is supplied to the digital-to-analog conversion circuit 4, where it is converted into an analog signal. This analog signal is supplied to the speaker via an amplifier circuit 5 provided as necessary.

According to the present invention, as described above, since the integration process is performed by appropriately configuring the sound source, there is no need for an integration circuit consisting of a resistor and a capacitor. Therefore, the speech synthesis device of the present invention is suitable for integration into an IC.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a speech synthesis device using partial autocorrelation coefficients, FIG. 2 is a block diagram of a speech synthesis device according to an embodiment of the present invention, FIG. 3 is an impulse waveform diagram, and FIG. 4 is a block diagram of a speech synthesis device using partial autocorrelation coefficients. , is a waveform diagram of an integral sound source signal. 1...Sound source section, 2...Digital filter, 3
...Parameter supply circuit, 4...Digital-to-analog conversion circuit, 5...Amplification circuit, 6...Speaker.

Claims (1)

[Scope of Claims] 1. In a speech synthesis device using a partial autocorrelation coefficient extracted for a difference signal of an input waveform as a feature parameter, a plurality of impulses whose level is sequentially decreased within each pitch period during voiced speech are provided. A speech synthesis device characterized in that the signal is sound source information. 2. The sound source information is transmitted to a sound source unit comprising a read-only memory and a control unit that sequentially forms address information of the read-only memory so as to output the plurality of impulse signals from the read-only memory in each pitch period. 2. The speech synthesis device according to claim 1, wherein the speech synthesis device is formed by: