JPS593496A - Fundamental frequency control system for rule synthesization 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 [Technical Field of the Invention] The present invention relates to a speech synthesis device using a regular synthesis method, and is capable of obtaining sufficient performance through simple control.

[Prior art to the invention]

There are roughly two types of output methods for audio output devices. 1
One is to prepare all the words or sentences to be output,
It outputs them sequentially in a desired combination,
When there are many types of sentences to be output, there is a drawback that a huge amount of storage capacity is required.In contrast, the output method that is the subject of this application parameterizes the phonological units of natural speech into only the necessary types. This is a rule synthesis method that synthesizes arbitrary words or sentences from those phonological units.In this case, not only do you arrange each phonological unit, but you also control their pitch and amplitude to make more natural speech. be able to. .

[Problems with conventional technology]

In the conventional method, in order to obtain a smooth time-varying pattern of the fundamental frequency close to that of human speech, the fundamental frequency is set using complex means such as function approximation at short time intervals of about 10 m5ec. 1, which is performed using numerical values, because from the perspective of human hearing, a better approximation can be obtained by operating according to the logarithm of the frequency. However, the so-called speech synthesis LSI 1li, which creates an audio signal according to this fundamental frequency information, requires input of the C fundamental period (reciprocal of the frequency), and therefore it is necessary to convert the logarithm of the frequency into an m period. However, there is a problem in that it requires a complicated circuit and conversion processing time.

Furthermore, when connecting two phoneme units, if the final amplitude of the first phoneme unit and the initial amplitude of the second phoneme unit are different, it is necessary to interpolate the difference so that the difference between them is smooth. This amplitude information is generally given as a parameter in a logarithmic representation with further information compression, but in order to interpolate it, it must be expressed as a simple integer, then interpolated, and then returned to the original buff meter format and input to the speech synthesis LSI. However, there is a problem in that the circuit and processing time required for this increase are large.

[Purpose of the invention]

←Invention this month valve→ The purpose of the present invention is to solve these problems and provide a speech synthesis control method that can perform high-speed processing with a simple circuit and obtain sufficient performance.

[Structure of the Invention] In order to solve the above-mentioned drawbacks, the present invention first provides information for controlling the fundamental frequency as a rough approximation, and then smoothes the information (digital information) using a digital filter.
I am trying to obtain a clarity that is close to that obtained by giving a detailed approximation.

In addition, a table lookup method is adopted to convert this fundamental frequency information into parameters based on the pitch period, which simplifies the hardware. Achieve ease of gender change.

In addition, in the interpolation process of amplitude information, the amplitude parameter is treated as a pure binary number and interpolated without converting it to a negative integer value, thereby achieving high-speed and high-performance interpolation process. .

[Embodiments of the invention]

FIG. 1 is an explanatory diagram of the rule-based method, taking as an example the case where the word "Yamagata" is synthesized. There are several ways to form phonological units, but here we use the so-called V
It uses a combination of CV (vowels, consonants, vowels).

``YAMAGATAJ is made up of five phonetic units ``YAJ''.

It is obtained by connecting "FAMA", "AGA", "EATA", and "A" ((a) in the same figure). Since each unit can be connected using the same vowel, natural connections can be obtained relatively easily.

FIG. 6B shows fundamental frequency information (logarithmic expression), indicating that "ma" has an accent.

Figure (c) shows a pinch (periodic) expression of the fundamental frequency, where the logarithmic expression is a straight line, but the period is a curved line.

FIG. 2 is a block diagram of one embodiment of the present invention, in which 1 is means for manually inputting words/sentences to be synthesized using character codes, 2 is means for converting character strings into sequence information of phoneme units (VCV), and 3
4 is a means for specifying accent and intonation patterns, 5 is a prosodic table that gives parameters according to intonation, etc. 6 is a means for disassembling and arranging the phoneme unit sequence into V and C sequence information, 6 is a means for specifying accent and intonation patterns, 6 is a means for giving parameters according to intonation, etc. means for creating a chronological pattern of connection times between each phoneme;
7 is a digital filter.

8 is a conversion table; 9 is a file in which each phoneme unit (VCV) is parameterized and stored at a sampling period of, for example, 10 m5ec; 10 is a means for combining VCV parameters; and 11 is a means for synthesizing an audio signal according to the parameters. , 12 are speakers.

When finding the fundamental frequency information (logf) by indexing the phoneme table 5 using intonation information, etc., conventionally, in order to obtain a sufficiently smooth Iogf curve, the information in the phoneme table 5 of 2 is indexed at a period comparable to the above-mentioned sampling period. Either the information in Table 5 was prepared in detail, or the information in Table 5 was changed and instead a relatively complicated predetermined function was used to interpolate between them.

On the other hand, in the present invention, the information in table 5 is, for example, 10
It is made coarser every 0 m5 e c, and interpolation is performed using a simple function such as linear interpolation. Instead, the information is smoothed by the digital filter 7.

FIG. 3 shows an embodiment of the digital filter 7, in which the difference is taken manually or by a subtracter 31 from the delayed output Y (Z-'), yz-i, and the amplifier 32 multiplies it by eight times (1 ＞
a>0) L, and YZ 1 are added to it by an adder 33 and output. Output YVi] Sampling period (for example, 10 m5ec) is delayed by the delay element 34 and sent to the subtracter 31
．． It is fed back to the adder 33. The transfer characteristic of this circuit is , which is a pole when Z = 1-a, and when mapped to the S plane, Z = e "πfT. Here, f is the polar frequency and T is the sampling period. Now, a=174.
If T=0.01 seconds, then f--UA'n(1-a)÷
4.6Hy, i.e. cut ono 2πT frequency 4.611z, -6db10ct low pass,
It becomes a filter.

If such a dirk is used, even a linear approximation like the solid line in Fig. 1(h) will result in a smooth Iogf like the dotted line in Fig. 1(h).
It becomes information.

Furthermore, a conversion table 8 is used to convert the logarithmic expression information of the fundamental frequency obtained in this way into period information. FIG. 4 shows an example of a table, using a ROM (read only memory). For example, 8-bit binary numbers (0 to 255) 10 are given to the human input, whereas the output is 7-bit binary numbers. (111 to 21) 1o is output in base number.

According to such a table conversion method, it is possible to convert the device into a device with arbitrary characteristics simply by converting or switching the ROM. For example, changing from a male voice to a female voice can be easily performed.

Further, amplitude interpolation between each phoneme unit is performed as follows.

Note that each phoneme unit is stored in the VCv file 9 as a time-series set of various parameters every 10 m5ec.

Amplitude information is also one of the parameters. Figure 1 (a
) The amplitudes of the same vowel "A" are generally different, such as at the end of [YAJ and at the beginning of 1'-A M AJ , and when connecting them, it is necessary to perform interpolation to connect them smoothly. The amplitude information in the parameters is generally expressed as shown in the table below.

In other words, a decimal integer is expressed as a floating point binary number, and the first digit below the decimal point is always "1°", so it is omitted to compress the number of bits.For example: The final amplitude of a certain sound (H), < rammeter is 1'1000
1”, the first amplitude parameter of the next phoneme is °’10
101", conventionally each] (parameter was set to an integer value of "10")
” and °°20” (However, linear interpolation between them is taken and the old); it is returned to the parameter format.

In contrast, in the present invention, the parameter format is treated as a pure binary number and linear interpolation is directly performed. That is, in the above example, the difference between "10001" and "10101" is 00100", and if linear interpolation is performed between four points, it is sufficient to add a difference of "00001" to each point.

Therefore, as my ex-husband said, the parameter format is '10001'.
``1'10101'' t is interpolated in steps of ``00001''. If you convert this to decimal and display, +
110", 121",'14"l, 16".

It becomes 1'20°', and it can be seen that it is an exponential interpolation. This method not only eliminates the need for unnecessary conversion, but also has the advantage of superior auditory characteristics. This is because human hearing senses that even if frequency, amplitude, etc. change logarithmically, it only feels like a linear change. is superior.

An example of an interpolation circuit is shown in FIG. 5, and this circuit may be considered to be within the VCV coupling unit 10.

The final amplitude parameter of the previous phonetic unit is stored in the register 41, and the first amplitude parameter of the next phonetic unit is stored in the register 42.
, and the subtracter 43 calculates the difference by treating each as a pure binary number. On the other hand, means 6 for determining the connection time between phonological units
Divider 44 divides the difference by the connection time (number of sample points) from
The adder 45 adds the quotient to the previous final amplitude value.

〔Effect of the invention〕

As described above, according to the present invention, by performing simple processing with a simple circuit, it is possible to obtain characteristics equivalent to or better than the conventional ones, and it is effective in reducing the cost of the audio output device.

[Brief explanation of the drawing]

Fig. 1 is an explanatory diagram of a rule synthesis method, Fig. 2 is a schematic block diagram of an embodiment of the present invention, Fig. 3 is a block diagram of an embodiment of a digital filter, and Fig. 4 is a block diagram of an embodiment of a conversion table. , FIG. 5 is a block diagram of an embodiment of the interpolation circuit. Figure 1 Akira λ Mouth

Claims (1)

[Claims] In a rule synthesis method that parameterizes and stores multiple phoneme units at a predetermined sampling period, connects the desired phoneme units, and synthesizes speech by giving a time-varying pattern of fundamental frequency information, the fundamental frequency information is used as an address. A basic frequency control method in a rule synthesis method, characterized in that the indexed table Vcj converts the information into another parameter format.