JPS61144700A - Sound source generator - 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 device for generating a vocal cord sound source waveform in a device for synthesizing speech and the like.

[Prior art and its problems]

2. Description of the Related Art Conventionally, there has been known an apparatus for synthesizing speech, etc., which synthesizes speech by inputting a vocal cord sound source waveform to a filter that realizes the transfer characteristics of the vocal tract.

Conventionally, periodic vocal cord sound source waveform (hereinafter referred to as sound source waveform)
There are known devices that generate an impulse train for each pitch period of the voice, and devices that generate the impulse train by filtering it using a filter that approximates the spectrum of the sound source waveform for l pitches. ing. As a conventional example of this, for example, L.R. Rabiner (L.
,R.

Rabiner) and R W Chef y-
(R, W, 5chafer) co-authored the document ``Digital Processing of Speech Signals (Digital Processing of Speech Signals)''
5peech Signals)”.

However, the sound source waveform obtained by this conventional example is
There is a problem in that synthesized sounds of good quality cannot necessarily be obtained because the waveforms are different from the actual sound source waveforms.

As a second conventional example, one is known in which the sound source waveform for one pinch is stored in a memory and the sound source waveform is repeatedly used for each pinch cycle. As an example of the sound source waveform for one pitch in this case, the one given by the following equation is shown in the above-mentioned document.

According to this second conventional example, it is known that a relatively good synthesized sound can be obtained, but when changing the shape of the sound source waveform, it is necessary to store each sound source waveform, so it requires a lot of time. The problem was that it required memory.

Furthermore, if the sound source waveform represented by equation (1) is to be obtained by calculating equation (1) at each time, there is a drawback that multiple calculations are required.

[Purpose of the invention]

An object of the present invention is to improve the above-mentioned drawbacks of the conventional ones, and to provide a sound source waveform generation device that can obtain a sound source waveform with a relatively good synthesized sound with a small number of calculations, and that can easily change the shape of the waveform. It is about providing.

[Structure of the invention]

A first aspect of the present invention is a device for generating a periodic sound source waveform for a device for synthesizing speech, etc., which includes a pulse generation circuit with variable amplitude, pulse width, and pulse generation period, and a recursive digital filter with variable coefficients. and a non-linear digital filter having means for passing only non-negative value signals to the output section of the adder, and the non-linear digital filter is provided with a non-linear digital filter having means for passing only non-negative value signals to the output section of the adder, and the non-linear digital filter is The present invention is characterized in that the pulse waveform generated by the pulse generation circuit is filtered by the nonlinear digital filter while switching the coefficients of the digital filter.

By further changing the waveform shape of the sound source waveform obtained by such a sound source waveform generation device, it is possible to generate a sound source waveform suitable for synthesizing desired speech.

Therefore, according to the second aspect of the present invention, in a device for generating a periodic sound source waveform of a device for synthesizing speech, etc., there is provided a pulse generating circuit with variable amplitude, pulse width, and pulse generation period, and a cyclic waveform with variable coefficients. The digital filter includes a non-linear digital filter having means for passing only non-negative value signals to the output section of an adder, and a linear digital filter with variable coefficients, and the filter is a - While switching the coefficients of the nonlinear digital filter in synchronization with the rising and falling points of the pulse waveform,
The pulse waveform generated by the pulse generation circuit is filtered by the nonlinear digital filter, and the output of the nonlinear digital filter is further filtered by the linear digital filter.

〔Example〕

FIG. 1 is a block diagram showing a first embodiment of the present invention.

This sound source waveform generation device includes a pulse generation circuit 1 whose amplitude, pulse width, and pulse generation period are variable, a nonlinear digital filter 2 whose coefficients are variable for filtering pulses from this pulse generation circuit, and a coefficient of the nonlinear digital filter 2. The pulse generating circuit 1 is composed of a switch 3 for switching the pulse generating circuit 1 and a control circuit 4 for controlling the pulse generating circuit 1 and the switch 3.

FIG. 2 is a diagram for explaining an example of the sound source waveform generated by the present embodiment. FIG. 2(a) shows the pulse generated by the pulse generation circuit 1, and FIG. 2(b) shows the generated This shows the sound source waveform. Although both are depicted as temporally continuous signals, they are actually discrete signals for each sampling period.

The sound source waveform in FIG. 2(b) has a rise time of N for one pitch and a fall time of N2. N
l and N2 both represent the number of samples,
The actual time length is Nl or N2 depending on the sampling period.
It is the value multiplied by

The pulse generation circuit 1 generates the pulse shown in FIG. 2(a), sends this pulse to the nonlinear digital filter 2 via the signal line 5, and simultaneously outputs a synchronization signal indicating the rising and falling points of the pulse. via line 6 to switch 3. The amplitude of the pulse generated by this pulse generating circuit 1,
The nine-cycle pulse width (see FIG. 2(a)) can be arbitrarily set by a control signal sent from the control circuit 4 via the signal line 7.

The nonlinear digital filter 2 is a second-order cyclic digital filter, and includes two delay elements 8.9, two multipliers 10 and 11, and an adder 12. The output section is provided with a means (hereinafter referred to as a nonlinear element) 13 for passing only signals of non-negative values so as to be nonlinear. The output of the nonlinear element 13, that is, the sound source waveform shown in FIG.

The switch 3 selects the signal line 15 and the signal line 16 from the control circuit 4 at the rising edge of the pulse in synchronization with the synchronization signal sent from the pulse generation circuit 1 via the signal line 6, and At the time of falling, signal line 1 from control circuit 4
7 and the signal line 18, the information from the signal line 15 or the signal line 17 is sent to the first multiplier 10 inside the nonlinear digital filter 2 via the signal line 19, and the information from the signal line 16 or the signal line 18 is sent to the first multiplier 10 inside the nonlinear digital filter 2. The information is sent to the second multiplier 1 inside the nonlinear digital filter 2 via the signal line 20.
Send to 1.

The control circuit 4 transmits pulse amplitude and pulse width 1 to the pulse generation circuit 1 according to control information input from the signal line 21.
Information for controlling the cycle is sent, and two values each of the coefficients a and b of the two multipliers 10 and 11 of the nonlinear digital filter 2 are sent to the signal lines 15, 16, 17, and 18.
is sent to switch 3 via.

In the sound source waveform generating device having such a configuration, the pulse generating circuit 1 generates pulses as shown in FIG. 2(a) in accordance with a control signal from the control circuit 4. As shown in FIG. The pulse generating circuit 1 causes the switch 3 to select the signal lines 15 and 16 when it detects the rising edge of the pulse, and causes the switch 3 to select the signal lines 17 and 18 when it detects the falling edge of the pulse.

A coefficient a to the first multiplier 10 is supplied to signal lines 15 and 17, and a coefficient a to the second multiplier 11 is supplied to signal lines 16 and 18. These coefficients a and b have values determined by the following equations (2) and (3).

(a) Coefficient from the time the pulse rises to the time it falls, that is, while the pulse is a non-zero value; (B) After the time the pulse falls, until the next rise, that is, while the pulse is zero.

In equations (2) and (3), N and N2 are the rise time and fall time of the sound source waveform for one pitch shown in FIG. 2(b), and in this example, the time N + is fJ
Z Set to a value equal to the pulse width of the pulse in figure (a,).

The nonlinear digital filter 2 synchronizes with the pulses from the pulse generation circuit 1 and uses the coefficients a and b determined by equations (2) and (3) as the coefficients of the multiplier 1O211, thereby achieving the result shown in FIG. The sound source waveform shown in b) is connected to signal line 1.
Get 4. This sound source waveform is the second waveform shown in equation (1).
It can be seen that the waveform is extremely similar to the sound source waveform of the conventional example.

According to this embodiment, the shape of the sound source waveform for one pinch can be freely controlled by controlling only the coefficients a and b determined by equations (2) and (3), the pulse amplitude, and the pulse width. Furthermore, by setting the value of R in equations (2) and (3) to a value other than 1, the shape of the waveform can be changed even more widely. These points are superior to the second conventional example, and the amount of calculation is comparable to that of the simplest configuration in the first conventional example.

Note that if the value of R in equations (2) and (3) is fixed to 1, the value of the coefficient is -1 in both equations (2) and (3), so the second The multiplier 11 is no longer necessary, and the configuration can be made even simpler.

FIG. 3 is a block diagram showing a second embodiment of the present invention. In this embodiment, a linear digital filter 22 is provided after the nonlinear digital filter 2 of the embodiment shown in FIG.
is further added so that the control circuit 23 sends coefficient values to the linear digital filter 22 via the signal line 24 in accordance with control information input from the signal line 21. The rest of the structure is similar to the embodiment shown in FIG. 1, and therefore, the same elements as those in the embodiment shown in FIG. 1 are designated by the same numbers.

According to such a sound source waveform generation device, the linear digital filter 22 filters the output signal of the nonlinear digital filter 2 according to the value of the coefficient sent from the control circuit 23, and outputs the output from the signal line 25 as a sound source waveform. Output. For example, the frequency characteristic of the linear digital filter 22 is 6
If the coefficients are set so that the low-pass filter has a cutoff characteristic of dB10ct, a sound source waveform suitable for nasal sounds, for example, can be generated. As described above, according to this embodiment, by further changing the shape of the waveform obtained by the nonlinear digital filter 2, it is possible to generate a sound source waveform suitable for synthesizing desired speech.

〔Effect of the invention〕

As described above, according to the present invention, it is possible to generate a sound source waveform that provides a relatively good synthesized sound with the same amount of calculation as the first conventional example.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of the first embodiment of the present invention, FIG. 2 is a diagram for explaining the sound source waveform generated by the embodiment of FIG. 1, and FIG. An example of a pulse waveform generated by the circuit, FIG. 2(b) is a diagram showing an example of a generated sound source waveform, and FIG. 3 is a block diagram of a second embodiment of the present invention. ' 1... Pulse generation circuit 2... Nonlinear digital filter 3... Switch 4.23... Control circuit 8.9... Delay element 10.11... Multiplier 12... ... Adder 13 ... Nonlinear element 22 ... Linear digital filter l6 Shimatanaru Figure 2

Claims (2)

[Claims] (1) A device for generating a periodic sound source waveform for a device for synthesizing speech, etc., which includes a pulse generation circuit with variable amplitude, pulse width, and pulse generation period, and a cyclic digital filter with variable coefficients, and an adder. a non-linear digital filter having means for passing only non-negative value signals through the output section of the non-linear digital filter, and the coefficients of the non-linear digital filter are changed in synchronization with the rising and falling points of the pulse waveform generated by the pulse generating circuit. A sound source waveform generation device characterized in that the pulse waveform generated by the pulse generation circuit is filtered by the nonlinear digital filter while switching the pulse waveform generated by the pulse generation circuit. (2) A device for generating a periodic sound source waveform for a device for synthesizing speech, etc., which includes a pulse generation circuit with variable amplitude, pulse width, and pulse generation period, and a cyclic digital filter with variable coefficients, and an adder. A non-linear digital filter is provided with means for passing only non-negative value signals at the output section of the circuit, and a linear digital filter with variable coefficients is provided. filtering the pulse waveform generated by the pulse generation circuit by the nonlinear digital filter while synchronously switching the coefficients of the nonlinear digital filter, and further filtering the output of the nonlinear digital filter by the linear digital filter. Characteristic sound source waveform generation device.