JPS58158694A - Voice generation system - 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 Object of the Invention The present invention relates to a sound generation method in a device equipped with a sound generation mechanism capable of generating a predetermined sound based on digital information. Specifically, PCM-pPARc
It relates to a voice modification method in a device that has a voice synthesis mechanism using OR codes.

Prior Art Conventional analog recording (
Methods based on digital technology have been devised and put into practical use for memory) and playback (readout) editing methods, but compared to analog recording methods, they are less compatible with digital equipment and have less controllability. While it has the advantage of being easy to use,
Compared to storing information in an analog state, the information storage efficiency is lower, and as a result, the number of words that can be generated is difficult. It was economically impossible to prepare separate voice generation parameters (high, low, etc.).

OBJECT OF THE INVENTION The object of the present invention is to provide a device that is equipped with a sound generation mechanism based on digital technology and capable of generating voices, by adding 16 decorations to the original voice of the voiced generation mechanism. Another object of the present invention is to economically and additionally provide such a device having a sound generation mechanism that accommodates equivalently a large number of sounds.

All voice synthesis methods using digital technology convert the base voice from analog to digital.

After layer conversion, a predetermined compression method (differential pulse encoding, f
(LM-type predictive coding, etc.) to reduce the amount of data (this will be referred to as the parameter of the voice). η At the time of voice synthesis, the parameter is decoded using an algorithm reverse to that used for encoding, and then converted from digital to analog to obtain a desired audio signal. The parameter corresponds to the phonological period.

It must have a variable length data amount and be continuously supplied. Therefore, the sound generation mechanism generally stores in advance a parameter string corresponding to a sound section of a desired length in a dedicated storage mechanism, and reads out the parameter string from the parameter mechanism using a dedicated control mechanism 5, and The system is configured to decode the data and supply it to a speech synthesis mechanism that obtains a speech signal. Therefore, in order to distinguish and generate sounds with different nuances even if they are phonetically synonymous (for example, skin and river), conventionally the results of encoding each sound are used as phonological parameters separately. It was difficult to economically obtain sounds with different pronunciation forms because they needed to be stored in a memory device. (Data of 1,200 to 9,600 bits per second is required) In the present invention, the generation of sounds that are phonetically equivalent only with different pronunciation forms is independent of the encoding algorithm (even if there is a relationship). At least.

This was done by focusing on the fact that this can be easily achieved by adding operations to the decoding results or the audio during the decoding process.

By specifying the address of the storage mechanism containing the parameter string corresponding to the basic phoneme and also specifying the pronunciation form, it is possible to obtain phonemes with different nuances.

Furthermore, in consideration of generating a meaningful sentence by linking the phonemes with different nuances obtained in this way, it has a second storage mechanism of finite size that stores the address of the parameter string and its modification information, It is configured to read parameter addresses and modification information from this second storage mechanism.

The secondary storage mechanism may take various forms depending on the purpose of the device.

can take a position. For example, in the case of a device whose sole purpose is to generate audio, it can be a FIFO-format pipeline register (memory), and in the case of a device that is a character display device that partially generates audio, it can be used as a character display device. You can apply a display glue frenosie buffer to this. In either case, the data stored in the buffer is delimited by a special code indicating that the data pertains to the parameter address and its axis information, or by specific information bits that make up one character of the buffer. configured to be identified by

Embodiments of the Invention Hereinafter, embodiments of the present invention will be explained with reference to FIG. The figure shows a device in which a voice generation mechanism based on a parameter synthesis method is added to a character display device, and the device is configured to output sounds along with the character display to indicate abnormal conditions or messages that should particularly call the operator's attention. Buffer structure and related sound-electric synthesis.

It shows the connections between the mechanisms. In the figure, block 1 is a shared buffer mechanism for storing a character display device buffer, voice parameters and their modification information. Block 2' is a voice generation section using a parameter synthesis method, and block 3 is a voice modification section. The buffer 4 has a bit length sufficient for one word to represent a display character or a voice parameter address and its modification information. Furthermore, it has a control bit 5, which is used for identification with a display character code or audio parameter address.

When reading the buffer 4, the control bit 5 is placed in register B, and the information part of the word is allocated to the parameter address & modification information register 7 or the character code & display attribute information register 8, respectively. The data stored in register 8 is given to character generator 9 and displayed as characters on display 10.

and it will be displayed. On the other hand, the voice parameter address and its modification information stored in the register are sent to the parameter storage section 11 and the voice modification section 15, respectively. The address sent to the parameter storage unit 11 is a parameter register containing the contents of the address.

12 and controls the reference pitch generator 13 and multistage digital filter 14 according to the parameter definition.

control and obtain a digital value corresponding to the corresponding voice. The voice modification section 15 uses the output from this digital filter 14? After receiving and modifying it according to the modification instruction given in register 7, the D/A converter & low-frequency wave transducer 1
6, the modified voice is output from the speaker 17.

Depending on the content of the modification, the voice modification section may be used as a reference pitch generator 15.
It can also be configured to be incorporated into the multi-stage digital filter section 14 and to modify the basic parameters.

FIG. 2 shows an example of the configuration of the voice modification section. The figure shows a configuration for modifying the volume.

21 is a shift register that stores volume information given as modification information. 22 is a shift register that similarly stores serial audio information output from the audio synthesizer 2.

21 is programmed so that only 1° bit in the register is 1. Also, in the rightmost cell position, %111
A detector 25 is provided. After storing a predetermined sample value in the raster 22, the shift register 20M is transferred to the shift register 20M by a common lock CLK for the shift registers 21 and 22.
Shift in the 5B direction (to the right in the figure). This shift is 11
This continues until the kite 1 is detected by the kite 1 detector 25.

Ru. The 22 data overflowed by the shift are stored in the overflow register 24. When 11 is detected by the detector 23, the CLK is stopped and the shift operation is completed.

The values held in the shift register 22 and overflow register 24 are collectively converted into an analog signal by a D/A converter 25. With this configuration, the sound pressure is doubled per shift, making it possible to adjust the volume in 6 dB increments.

It is.

Effect of the invention According to the present invention, based on limited phonemes, A.

It is possible to economically obtain the effect of accommodating a wide variety of phonemes.

[Brief explanation of drawings]

Fig. 1 is a block diagram of a device equipped with a sound generation mechanism showing an embodiment of the present invention, and Fig. 2 is a block diagram showing details of the sound modification section shown in Fig. 1'. ...Shared buffer mechanism, 2..Speech generation section, 3..Speech modification section. 4... Buffer, 7... Parameter address & modification information register, 8...
...Character code & display attribute information register, 9...Character generator, 10...Display device, 11...Parameter storage section, 12...Parameter register, 13...Reference pinch generator, 14...・Multi-stage digital filter, 15...Speech modification unit, 16...D/A converter & low-frequency wave transducer, 17...Speaker.

Claims (1)

[Claims] A storage mechanism that stores generation parameters (for example, a PCM code string or a pARcOR code string) of a voice generated according to a predetermined procedure for synthesizing a predetermined voice, and a parameter output from the storage mechanism to synthesize a predetermined voice. Speech synthesis mechanism (e.g. PCM decoder or PARCOR
A second storage mechanism is provided for storing the address of the storage mechanism in which the audio parameters to be generated are stored, and the second storage mechanism stores the parameters of the storage mechanism. A voice generation method characterized in that voice modification information is stored together with an address, and when voice is generated, the generated voice is modified based on the modification information.