JPS63211000A - 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 (Industrial Application Field) The present invention is a device for synthesizing speech (including musical tones) in a sample value system, that is, a time-discrete system, and is a speech synthesis device whose circuit configuration can be freely changed. Regarding.

(Prior art) Conventionally, in a sample value system (or digitally)
As a speech synthesizer that generates sample values of speech waveforms,
It consists of a sound source circuit that generates a resource waveform based on sound source information such as pitch, voiced or unvoiced, and a digital synthesis filter whose frequency characteristics are controlled based on spectral information. The sound source information and spectral information in such speech synthesizers, which are known to synthesize speech by filtering them using It is possible.

An example of such a speech synthesizer can be found in the technical magazine of NEC Corporation! NEC Technical Report Vo1.34, N1110. p,
This is shown in Suzuki et al.'s "Speech synthesis LSIJ using formant as a parameter (first conventional example)" published in 92-97.The speech synthesis circuit inside this LSI performs periodic A voiced sound source circuit 201 that generates an impulse train and an unvoiced sound source circuit 20 that generates a pseudorandom number sequence
2 and five cascade-connected digital resonance circuits (digital filters) 203.

A second conventional example is the Journal of Acoustic Society of America.
f Acoustical 5ociety of A
merica.

? ) t') Journal of the Force Acoustics Society) Vol, 67, k3.
Dee, H., and Kratt (D, H.) published on p. 971-995.

Klatt)'s paper ``Software for Academia/Parallel Formant Synthesizer (S
ofware for a cascade/par
allel forn+antsynthesizer
) There is something shown in J. In this example, in addition to the cascade-connected resonant circuits as in the first conventional example, there are also parallel-connected resonators, and the sum of their outputs is calculated to generate the sound. It is something that is synthesized.

As a third conventional example, a patent application filed in 1982 by Fushikida and Mitome
050595 r Speech Synthesizer", which has a synthesis filter whose pole-zero circuit connection is variable. The connection relationship is determined based on the input speech synthesis information.

(Problems to be Solved by the Invention) In the first conventional example, since the structure of the speech synthesis circuit is simple, it is possible to implement it into a one-chip LSI. There was a problem that it was not possible to synthesize voices.

The second conventional example has a more complex structure than the first conventional example, and can accommodate several articulation styles depending on how the parameters are controlled, resulting in a relatively good synthesized sound. is obtained. However, there are known problems such as unpleasant clicking sounds occurring when the parameter values change suddenly.

In order to solve this problem, the configuration of the speech synthesis circuit must be replaced with a different one, and since there are many types of parameters required for synthesis, it is not suitable for compressing the amount of information.

As described above, each conventional speech synthesis device has its own merits and demerits, and must be selected depending on the purpose. In addition, the structure of the speech synthesis circuit cannot be changed. On this point, in the third conventional example, the connection of the pole-zero circuit of the synthesis filter can be changed, so it is suitable for speech with certain articulation styles, but it is suitable for controlling the amplitude of the sound source and filtering. There was a problem in that fine control of the resonant frequency was not easy. Therefore, there was a limit to the compression of the amount of information and the improvement of the clarity of plosives and voiced fricatives.

An object of the present invention is to provide a speech synthesis device which allows the configuration of a speech synthesis circuit to be easily and freely changed even during speech synthesis, and which is also suitable for LSI implementation.

(Means for Solving the Problems) The speech synthesis device of the present invention is a device that operates in a sample value system, and converts a sound source waveform generated by a sound source circuit into a digital filter (in the sample value system) and an filter. ) in a type of device that synthesizes speech by filtering it, a voiced sound source circuit.

Components of the speech synthesis circuit such as an unvoiced sound source circuit, an impulse resource circuit, a digital filter, and an adder 1 multiplier, a memory that stores frequency characteristic parameter values of the digital filter, and a signal register that stores output signals of each component. , with the output terminal of the signal register as an input terminal, and a signal to the digital filter, the input terminal of the multiplier and multiplicand data of the adder 9 multiplier, the input terminal of the frequency characteristic parameter value memory of the digital filter, and the output terminal of the synthesized speech. An exchange switch whose output terminal is a line, a memory that stores connection information indicating the mutual connection relationship between the exchange switches, and a connection between the input and output of the exchange switch according to the connection information in the connection information memory. i11 control means.

(Function) In the present invention, a voiced sound source circuit, an unvoiced sound source circuit, and a digital filter are components of a speech synthesis device.

Memory and adder for its frequency characteristic parameter values.

Although it is similar to a conventional speech synthesizer in that it has a multiplier, these components are not wired together in a fixed manner, but instead have a signal register for temporarily storing the output of each component and its signal. It differs from the first and second conventional examples in that an exchange switch is provided between the register and each component. Further, the third conventional example differs from the present invention in that although the connection of the synthesis filter can be changed, the connection to the multiplier cannot be changed.

In the present invention, it is not only possible to simply change the circuit configuration of the synthesis filter, but also the memory that stores the values of the frequency characteristic parameters of the digital filter, such as the pole-zero frequency and the bandwidth, and the input terminal of the multiplier. It is now possible to send signals. This makes it possible, for example, to use the impulse response of one digital filter as the frequency characteristic parameter value of another digital filter, or to control the signal amplitude using the output signal of the digital filter. Amplitude modulation etc. can be easily realized with side groups of few parameters.

(Example) Next, the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention, in which 101 is a control circuit, 102 is a voiced sound source circuit, 1
03 is a silent sound source circuit, 104 is an impulse sound source circuit, 1
05 is a filter array, 106 is a frequency characteristic parameter value memory, 107 is an adder array, 108 is a multiplier, 1
09 is a signal register array, 110 is an exchange switch, 1
11 is a connection information memory, and 112 is an output register.

The voiced sound source circuit 102 generates a periodic sound source waveform for each sampling period in accordance with a control signal that instructs the generation of a sound source waveform based on resource information such as pitch information and amplitude information sent from the fIII1m circuit 101 via a signal line 123. A circuit that generates sample values and sends them to the signal register array 109 via signals @ 135.

The unvoiced sound source circuit 103 is connected to the signal line 12 from the control circuit 101.
This circuit generates a pseudorandom number sequence according to the amplitude information and control signal sent via the signal line 136 and sends it to the signal register array 109 via the signal line 136.

The impulse sound source circuit 104 is a circuit that generates a single impulse according to the M control signal sent from the control circuit 101 via the signal line 125 and sends it to the signal register array 109 via the signal line 137.

A filter array 105 is a set of digital filters with variable coefficients, and the coefficients of each filter are connected from a frequency characteristic parameter value memory 106 to signal lines 140 and 141.
It is calculated from frequency characteristic parameter values such as the pole-zero frequency, bandwidth, and gain given through the . Furthermore, at the start of synthesis, etc., the value of the internal delay register is reset according to the control signal.

Then, according to the control signal, the designated filter filters the signal sent from the exchange switch 110 via the signal line 156 or 157, generates a sample value for each sampling period, and generates a sample value for each sampling period.
to the signal register array 109 via the signal register array 109. In this example, there are two input signal lines, two signal lines for transmitting frequency characteristic parameter values, and two output signal lines to the filter array 105, but the number depends on the number of filters in the filter array 105. It will be decided accordingly.

The frequency characteristic parameter value memory 106 stores frequency characteristic parameter values such as pole-zero frequency, bandwidth, gain, etc., and receives the signal from the ttsn circuit 101! I
According to the control signal sent through 127, the data is sent to the filter array 1 through signals 914G and 141.
Send to 05. This frequency parameter value is determined by the control circuit 10
1 and through exchange switch 11G,
Filter output signals may be provided via signal lines 154 and 155.

The adder array 107 is a set of adders that calculates the sum of signals, and the adder array 107 is a set of adders that calculates the sum of signals. , replacement switch 110
Calculate the sum of signals sent via signal @ 158 or 159, and send the result to signal line 142 or 143.
to the signal register array 109 via the signal register array 109. In this case as well, like the filter array 105, the number of input/output signal lines is determined by the number of adders.

Multiplier 108 calculates the product of the multiplier multiplicands input via signals ill 16G and 161 according to a control signal sent from control circuit 101 via signal line 129, and outputs the product to signal line 144.

The signal register array 109 is a set of registers that stores sample values of signals, and according to the control signal sent from the control circuit 101 via the signal line 130, the signal register array 109 is a set of registers that stores sample values of signals. The number of registers that store sample values of the signal to be sent and send them to the exchange switch 11G via signal lines 145 to 152, respectively, if instructed, is determined according to the number of filters, adders, or frequency parameter value memories. It is something. .

The exchange switch 11G connects the signal line 13 from the control circuit 101.
1, the sample value of the signal input from a specified input terminal, that is, one of the signal lines 145 to 152, is sent to a specified output terminal, that is, one of the signal lines 153 to 159. This is a switch that outputs . The number of input terminals and the number of output terminals of the exchange switch are determined depending on the number of filters, adders, or memories of frequency parameter values.

The connection information memory 111 receives a signal @ from the control circuit 101.
133, which stores connection information representing connection relationships between individual components, and sequentially sends the stored connection information to the control circuit 101 via a signal line 134 in accordance with a request from the m control circuit 101. send out.

The output register 112 is a circuit that temporarily stores the signal sent from the exchange switch 11G according to instructions from the control circuit 101, and outputs the signal to the signal line 122 in accordance with the timing of the sampling period.

As a result, even if the configuration of the speech synthesis circuit is changed and the time to generate one sample of synthesized speech fluctuates,
It is possible to output audio in exact synchronization with the sampling period.

Speech synthesis is performed as follows. First, signal line 12
1 to the M control circuit 101 is inputted, then connection information indicating the configuration of the speech synthesis circuit is inputted, and the control circuit 101 sends this to the connection information memory and stores it.

Thereafter, whenever new connection information is input, the contents of the connection information memory are rewritten each time.

Next, sound source information and spectrum information for each predetermined time, that is, values of parameters for controlling the sound source circuit and digital filter, which are the constituent elements of the speech synthesis circuit, are input. The control circuit 101 transmits this information to the voiced sound source circuits 102 . Silent sound source circuit 103. Impulse sound source circuit 104 and frequency characteristic parameter value memory
106 to control these components, and
While reading the connection information from the connection information memory 111, it is sent to the exchange switch 110 to control the flow of signals.

In this way, the sound source waveforms generated by the three types of sound source circuits 102, 103 and 104 are filtered by several digital filters, their sum is calculated by several adders, or their product is calculated by a multiplier. A synthesized speech is generated by calculation, etc., and is finally stored in the output register 112.

At this time, some of the signals are used not as synthesized speech but as frequency characteristic parameters of an external digital filter.

Finally, a control signal is sent from the control circuit 101 to the output register 112 via the signal line 132 at every sampling period, and sample value data of the synthesized voice is output from the signal line 122.

(Effects of the Invention) As described above, according to the present invention, even during speech synthesis, the configuration of the speech synthesis circuit can be dynamically changed by providing connection information from the outside. Furthermore, since the impulse response of one digital filter can be used as the frequency characteristic parameter value of another digital filter, changes in formants over time, amplitude modulation of sound source waveforms, etc. can be easily realized by controlling a small number of parameters.

Therefore, when used for speech synthesis based on rules, complex circuits can be realized, and the optimal circuit configuration can be realized for each phoneme according to the articulation style. Also, fine control of the sound source amplitude is possible, so it is natural and clear. This has the effect of being able to synthesize a variety of voices.

Furthermore, by simplifying the configuration of the speech synthesis circuit and generating various parameters using impulse responses of filters, the present invention can be applied to analysis and synthesis for the purpose of information compression.

Although it has the feature of being able to realize speech synthesis circuits with various configurations, the speech synthesis device of the present invention is actually composed of only basic filters, exchange switches, etc. It is also suitable, and cost reductions can be expected through mass production.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, and FIG. 2 is a block diagram showing an example of a conventional speech synthesis device. In the figure, 101 is a control circuit, 102 is a voiced sound source circuit, 103 is an unvoiced sound source circuit, 104 is an impulse sound source circuit, 105 is a filter array, 106 is a frequency characteristic parameter value memory, 107 is an adder array, 108 is a multiplier, 109 is a signal register array, 110 is an exchange switch, 111 is a connection information memory, 112 is an output register, 2
01 represents a voiced sound source circuit, 202 represents an unvoiced sound source circuit, and 203 represents a digital resonant circuit.

Claims (1)

[Claims] In a speech synthesis device that operates in a sample value system and synthesizes speech by filtering a sound source waveform generated by a sound source circuit with a digital filter: voiced sound source circuit, unvoiced sound source circuit, Components of a speech synthesis circuit including an impulse sound source circuit, a digital filter, an adder, and a multiplier; a memory for storing frequency characteristic parameter values of the digital filter; a signal register for storing output signals of each component; An input terminal is connected to the output terminal, and a signal line is connected to the input terminal of the multiplier and multiplicand data of the digital filter, the adder and the multiplier, the input terminal of the frequency characteristic parameter value memory of the digital filter, and the output terminal of the synthesized voice. an exchange switch serving as an output end; a connection information memory for storing connection information indicating their mutual connection relationship; and means for controlling the connection between the input and output of the exchange switch according to the connection information in the connection information memory. Characteristic speech synthesizer.