JPH0695677A - Musical sound synthesizing device - Google Patents

Abstract

PURPOSE:To reduce a memory required for synthesis without deteriorating the quality of the musical sound synthesizing device used for an electronic musical instrument. CONSTITUTION:Every N samples of a series of waveform data of a musical sound are stored in a waveform memory 125, and data which index data regarding waveform data which are not stored in the waveform memory 125 as subordinate data to the waveform data stored in the waveform memory 125 are stored in an index memory 200. Then, vector data corresponding to difference data specified with waveform data outputted from the waveform memory 125 and the data stored in the index memory 200 are read out on the basis of the address output of an address generator 100 which changes corresponding to pitch information to calculate an output waveform, so a musical sound waveform can be decoded by the small memory.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a musical tone synthesizing device for an electronic musical instrument which synthesizes and outputs a tone color similar to that of a natural musical instrument with high quality.

[0002]

2. Description of the Related Art In recent years, a so-called musical tone synthesizing apparatus for electronic musical instruments has been introduced with a synthesizing method by digital technology, and the quality of synthesized speech has been remarkably improved. In particular, a method based on the so-called PCM method, in which the musical tone waveform of a natural musical instrument is stored as it is and is reproduced at a speed proportional to the pitch to be reproduced, is often used. Such a musical sound synthesizer, for example,
JP-A-52-121313, JP-A-51-934
No. 8 or Japanese Patent Laid-Open No. 5114015.

The above-mentioned musical tone synthesizer will be described below with reference to the drawings. FIG. 9 shows the configuration of a conventional musical sound synthesizer. In FIG. 9, 100 is an address generator, 120 and 121 are waveform memories, and 130 to 132.
Is an adder and 140 is a multiplier.

The operation of the tone synthesizer configured as above will be described below.

The waveform memory 120 can store waveform data such as musical tones, for example, piano, strings, bass drum, etc. Here, it is assumed that piano sounds are stored. Waveform memory 120
Storage capacity of 16MBit, waveform data of 16Bi
If t and the waveform address are 20 bits, when the sampling frequency Fs is 40 kHz, the waveform data for about 26.2 seconds is stored. In this conventional example, the waveform from the rising edge to the falling edge of the piano sound (A = 440 Hz) sampled under the above conditions is stored. The waveform memory 12
It is assumed that 1 stores the same waveform data as the waveform memory 120.

When the pitch information S is input, the address generator 100 cumulatively adds the pitch information S at a timing corresponding to the sampling frequency Fs to obtain an address A0.
~ Outputs A29. The correspondence between the pitch information S and the data of the addresses A0 to A29 is shown in FIG.

Of the addresses output from the address generator 100, the addresses A10 to A29 are supplied to the waveform memory 120 and the adder 130, and the addresses A0 to A9 are the multiplier 1
40. The addresses A10 to A29 increased by the value 1 in the adder 130 are supplied to the waveform memory 121.

Address A supplied to the waveform memory 120
10 to A29 are the waveform address WA0 in the waveform memory 120.
The waveform data Wi stored in the corresponding waveform address is output. On the other hand, the addresses A10 to A29 supplied to the waveform memory 121 and incremented by 1 are the waveform addresses W in the waveform memory 121.
The waveform data Wi + 1 used as A0 to WA19 and stored at the corresponding waveform address is output. The waveform data Wi and Wi + 1 are substantially adjacent waveform data, and the difference (Wi + 1−Wi) is calculated and output by the adder 131. The difference (Wi + 1−Wi) is multiplied in the multiplier 140 by using the addresses A0 to A9 as a decimal part, and then added to the waveform data Wi in the adder 132 and output as the waveform value W. By the above operation, the waveform data W
The linear interpolation value W by the addresses A0 to A9 is calculated for i and Wi + 1. Correspondence of the above respective data is shown in FIG. 9 and FIG.

When the pitch information S has a value of 1.0, the address generator 100 outputs S = 1.0 for each sampling frequency Fs.
Is cumulatively added, the values of the addresses A10 to A29 increase by 1 for each sampling frequency Fs. Therefore, the waveform address WA input to the waveform memories 120 and 121
0 to WA19 are sequentially incremented by 1 while being shifted by the value 1 by the adder 130. Waveform memory 12
When the output of the waveform memory 120 changes to W0, W1, W2, W3, ..., The output of the waveform memory 121 is W1, W2, W3, W4 ,.・
・ And it will change. The difference data output obtained by the adder 131 is eventually the multiplier 1 because the addresses A0 to A9 multiplied by the multiplier 140 are always 0.
The output of 40 is the value 0, and the output of the adder 132 is the waveform data W0, W sequentially read from the waveform memory 120.
Since 1, W2, W3, ... Are output as they are, the stored piano sound (A = 440 Hz) is output as a waveform.

When the pitch information S has the value 2, the output of the waveform memory 120 is W0, W2, W4, W6, ...
.. Since the output of the waveform memory 121 is W1, W3, W5, W7, ..., From the adder 132, the waveform data W0, W2, W4, W6 ,. Since it is output as it is, and the stored piano sound is read every other waveform data, the sound (880 Hz) of the pitch name A one octave higher is output.

The value 1.189 (frequency ratio) may be given to the pitch information S for outputting the sound with the pitch name C (523.25 Hz). At this time, the output of the waveform memory 120 is W0, W1, W2, W3, W4, W5, W7, W by the same operation as described above.
8, ..., The output of the waveform memory 121 is W1, W2, W3, W4,
W5, W6, W8, W9, ... In the initial state, that is, when the outputs A0 to A29 of the address generator 100 are 0, the output W0 of the waveform memory 120 is output from the adder 132 as it is, but when the pitch information S is added, the waveform memory 120 is output. Output value W1 and the output value W2 of the waveform memory 121
, The difference value (W2-W1) is calculated in the adder 131, and the value of the addresses A0 to A9 is 0.
Since it is multiplied by 189 and added to W1 in the adder 132, an interpolated value obtained by linearly interpolating between the waveform data W1 and the waveform data W2 by the value 0.189 of the addresses A0 to A9 is output. Become. Similarly, in the multiplier 140, the addresses A0 to A9 are sequentially set to 0.37.
8, 0.567, 0.756, 0.945, 0.134, ...
.. while changing to the corresponding difference data (W2-W
1), (W3-W2), (W4-W3), (W5-W4),
(W7−W8), ... And the corresponding waveform data W0, W1, W2, W3, W4, W5, W7,
... is added and output as an interpolation calculation value.

In the above description, the same waveform data is stored in the waveform memory 120 and the waveform memory 121. However, the waveform data Wi and Wi + 1 are separated into single data by using the known time division processing technique. The waveform memory 120 is read out by accessing the waveform memory 120 twice to make the storage capacity of the waveform memory the same amount as so-called sampling data, or the waveform memory 122 has a difference in advance in the musical tone synthesizer configured as shown in FIG. A method has been proposed in which the adders 130 and 131 are unnecessary by storing the calculated (Wi + 1−Wi), but the latter method is a waveform memory having a storage capacity equal to so-called sampling data. Besides that, a waveform memory for storing difference data was required.

When synthesizing a very long piano sound or a continuous tone such as strings, the last part of the waveform data stored in the waveform memory is repeatedly read to maintain the output sound. You can

[0014]

However, in the above configuration, even if the time-division processing is performed, a large amount of bits determined by multiplying the number of bits constituting one waveform data, that is, the number of quantized Bits and the number of waveform data. There is a problem that a waveform memory for storing data is required.

In order to change and output the output pitch of the waveform data stored in the waveform memory, the pitch information S is cumulatively added to obtain the waveform address of the waveform data to be read. Waveform data is not always read sequentially, known sequential compression technology,
That is, differential compression (DPCM) and adaptive differential compression (A
There is a problem that a technique such as DPCM) or a compression technique such as vector quantization cannot be applied.

In order to solve the above problems, it is an object of the present invention to provide a tone synthesizer capable of synthesizing a natural tone with a small amount of waveform memory.

[0017]

In order to achieve this object, the musical tone synthesizer of the first invention stores waveform data for every predetermined number N of a series of waveform data of musical tones. A waveform memory, a vector memory for storing a plurality of sets of vectors composed of N or more elements, a vector corresponding to difference data between the waveform data stored in the waveform memory and N waveform data in the vicinity, and a magnitude thereof. The index memory that stores the index data that specifies the depth corresponding to the waveform data that is stored in the waveform memory, the address generator that outputs the address that sequentially changes according to the pitch of the tone to be output, the address and the index The interpolation calculation unit that interpolates the elements of the vector memory corresponding to the data, and the waveform data and the interpolation calculation unit of the waveform memory corresponding to the address. Composed of an adder for adding and.

In order to achieve this object, the musical tone synthesizer of the second invention comprises a waveform memory for storing waveform data for every predetermined number N of data from a series of musical tone waveform data. A vector memory for storing a plurality of sets of vectors composed of N or more elements, a vector corresponding to difference data between the waveform data stored in the waveform memory and N waveform data in the vicinity thereof, and its magnitude are specified. The index memory that stores the index data corresponding to the waveform data stored in the waveform memory, the address generator that outputs the address that sequentially changes according to the pitch of the tone to be output, and the waveform memory corresponding to the address A first interpolation calculation unit for performing interpolation calculation on waveform data, and a first interpolation calculation unit for performing interpolation calculation on elements of a vector memory corresponding to addresses and index data. And interpolation calculation unit, composed of a first adder for adding the second interpolation computation unit output.

Furthermore, in order to achieve this object, a third
The musical sound synthesizer of the invention described above comprises a skip data generator for outputting time-varying skip data Ni, a waveform memory for storing waveform data for each skip data Ni among a series of waveform data of musical sounds, and a musical sound for output. An address generator that outputs addresses that sequentially change according to the pitch and skip data Ni, and an interpolation calculation unit that interpolates the waveform data output from the waveform memory corresponding to the addresses and skip data. Composed of.

[0020]

According to the structure of the first aspect of the invention, the address generator outputs addresses which sequentially change according to the designated pitch. Index data stored in the index memory that is read based on the address, and a value obtained by performing an interpolation operation on the plurality of elements that form the vector stored in the vector memory that is read based on the address and the index data by an interpolation operation unit, Since the value of the waveform data stored in the waveform memory that is read based on the address is added by the adder, the waveform data in which a series of waveform data of musical tones is stored every N pieces and the stored waveform data The output waveform is calculated and output from the vector data corresponding to a plurality of difference data regarding.

According to the configuration of the second invention, the address generator outputs addresses which sequentially change according to the designated pitch. A value obtained by interpolating the plurality of waveform data stored in the waveform memory based on the address by the first interpolation computing unit, and index data and address stored in the index memory that are read out based on the address. Since a plurality of elements constituting the vector stored in the vector memory read out based on the above are interpolated by the second interpolating section and added by an adder, a series of waveform data of musical tones is obtained. The output waveform is calculated and output from the plurality of waveform data stored for every N pieces and the vector data corresponding to the plurality of difference data regarding the stored waveform data.

According to the third aspect of the invention, the address generator outputs addresses which sequentially change according to the designated pitch and the skip data Ni. The skip data generator outputs skip data Ni that changes with time. The waveform data stored in the waveform memory, which is read out based on the skip data and the address, is interpolated by the interpolating unit to obtain the output waveform.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of a musical sound synthesizing apparatus according to a first embodiment of the present invention. In FIG. 1, 125 is a waveform memory, 133 to 136 are adders, 145 and 146 are multipliers, 200 is an index memory, 220 and 221 are vector memories, 250 is a comparator, and 260 is a gate. The address generator 100 has the same structure as the conventional example.

In the first embodiment, the predetermined interlace number (skip data) N of waveform data is 4 (2 of 2).
Squared). FIG. 2A shows waveform data obtained by sampling the same tone waveform as that of the conventional example, that is, a part of the waveform data Wi-N to Wi + N stored in the waveform memory 120 of the conventional example. The waveform data selected from the waveform Wi-N every N samples, that is, every 4 samples is shown in FIG. 2B, and these waveform data are stored in the waveform memory 125. For the waveform Wi-N, the address (i
Waveform data from -N + 1) to address i and waveform Wi-
The difference data from N is shown in FIG. 2C. Similarly, for the waveform Wi, the address (i + 1) to the address (i +
The difference data of N) is also shown.

It is assumed that the vector memory 220 stores a vector having four elements corresponding to the skip data N as shown in FIG. For example, the vector V
0 is composed of four elements (0.25, 0.50, 0.75, 1.00). For addresses 0 to 3, the last element is the head, and from the first element onwards. To (1.
00, 0.25, 0.50, 0.75). Further, the vector V1 is (0.80, 0.87, 0.
93, 1.00) of four elements, and addresses 4 to 7 with the last element as the head,
It is stored sequentially from the first element. And so on
The four elements forming the vector V11 have an address of 44
It is assumed that the last elements are stored in order from the first element to the first to the fourth elements. The vector memory 221 is the same as the vector memory 220.

It is assumed that the index memory 200 stores index data corresponding to the waveform data stored in the waveform memory 125. Waveform data W
The index data Ii corresponding to i is composed of vector index VIi and level data Li. The vector index VIi is composed of 4 bits to indicate the vectors V0 to V11. In the present embodiment, there are 12 types of vectors, but this can be increased or decreased as necessary. Since the number N of jumps in the waveform data is 4, the sampling frequency is 40k.
In Hz, the fluctuation of the frequency component of 10 kHz or more contained in the musical instrument sound is largely reflected in the fluctuation of the difference data as shown in FIG. 2 (c) described above. Since the frequency component of is very small, it is sufficient that there are about 12 types of vectors in this embodiment. That is, regarding the differential data in FIG. 2C, 4 of the addresses (i + 1) to (i + N)
The two difference data can be approximated by the vector V6, and the difference data from the address (i-N + 1) to i can be approximated by the vector V0, and thus correspond to the waveform data Wi-N and Wi. The vector indexes of the index data are 0 and 6, respectively. However,
It is very effective to increase the number of elements forming the index data in order to improve the waveform interpolation accuracy and reduce the noise caused by the interpolation.
Alternatively, the interpolation accuracy is improved to 16 or 1
As 024, the interpolation calculation itself between the index data elements can be omitted. Also, the level data Li
Determines the size for using the vector data designated by the vector index VIi as 12-bit unsigned data. That is, regarding the difference data in FIG. 2C, the level data of the index data corresponding to the waveform data Wi-N and Wi are respectively (Wi-Wi-
N) and (Wi + N-Wi).

The operation of the tone synthesizer configured as above will be described below. Similarly to the conventional example, the address generator 100 cumulatively adds the instructed pitch information S and outputs the addresses A0 to A29. Waveform memory 125
The number of samples of the waveform data (see FIG. 2 (b)) stored in the waveform memory is 1/4 (N = N) of the number of samples of the waveform data (see FIG. 2 (c)) stored in the conventional waveform memory 120.
4), the addresses A12 to A29 are stored in the waveform memory 12
5, and waveform data Wi is output. Addresses A12 to A are similarly applied to the index memory 200.
29 is supplied, and the vector index VIi forming the index data Ii and the level data Li are output.

The vector memory 220 has addresses A10 to A11 output from the address generator 100 as lower addresses and a vector index VIi output from the index memory 200 as upper addresses, which is a vector determined according to a total of 6 bits. Output the elements that make up. The vector memory 221 also outputs the elements forming the vector by the same operation, but the adder 134
The value obtained by adding 1 to the addresses A10 to A11 is used as the lower 2 bit address. Further, when the values of the addresses A10 and A11 are both 0, the comparator 250 outputs the value 0 and resets the output of the gate 260 to the value 0. When either of the addresses A10 and A11 is not 0, the comparator 250 outputs the value 0. The comparator 250 outputs the value 1, and the gate 260 outputs the value of the element of the vector output from the vector memory 220 as it is. Therefore, the value of address A10-A11 is (0 → 1
(→ 2 → 3), the gate 260 and the vector memory 221 receive (value 0 → first element of designated vector → second element → third element) and (designated vector), respectively. The first element → the second element → the third element → the fourth element) are output. These outputs are linearly interpolated by the adders 135 to 136 and the multiplier 145 which are configured similarly to the adders 131 to 132 and the multiplier 140 of the conventional example, and then are multiplied by the level data in the multiplier 146, Data corresponding to the linearly interpolated value of the difference data in FIG. 2C is further added to the waveform data output from the waveform memory 125 in the adder 133, so that the linearly interpolated waveform is finally obtained.

In the first embodiment, in order to facilitate understanding, the vector memory for storing the vector is two memories, vector memories 220 and 221. However, like the conventional waveform memory, the time division is performed. It is obvious that one vector memory 220 can be configured by processing.

Further, in the first embodiment, since the predetermined N is set to 4, the compression ratio is 1/4 as the waveform data.
Since the index data also requires the same storage capacity as the waveform data, it has a storage capacity that is half that of the conventional waveform memory 120, except for the vector memory.

In the above, the predetermined N is 4
However, depending on the type of musical sound, the data compression rate can be further increased by changing the value of N to another value. Further, in the first embodiment, the linear interpolation value is obtained by the vector, but it is not always necessary to perform the linear interpolation calculation, but a higher-order interpolation calculation may be performed. Although the interpolation value obtained from the vector element is added to the waveform data, the same effect can be obtained by performing the interpolation calculation after adding the vector element to the waveform data. deep. Furthermore, since the difference value of waveform data that are temporally close in a musical tone waveform is generally smaller than the word length that forms the waveform data, the word length of the elements that form the vector is smaller than the word length of the waveform data. Since it can be represented by, the storage word length of the vector memory can be reduced.

As described above, according to the first embodiment, the interpolated value is calculated from the waveform memory storing only N waveform data according to the predetermined N and the vector data specified by the index data. I decided to do so,
Tones can be synthesized with a small memory and a simple configuration.

The second embodiment of the present invention will be described below with reference to the drawings. FIG. 4 is a block diagram of a musical sound synthesizer according to the second embodiment of the present invention. In FIG.
125 and 126 are waveform memories, 130 to 136 are adders, 140, 145 and 146 are multipliers, 201 is an index memory, and 222 and 223 are vector memories. The address generator 100 has the same structure as the conventional example.

In the second embodiment, as in the first embodiment, the number of skips (skip data) N of the predetermined waveform data is set to 4 (2 squared). Waveform data obtained by sampling the same tone waveform as the conventional example, that is, a part of the waveform data stored in the waveform memory 120 of the conventional example Wi-N to Wi + N.
Is shown in FIG. Every N samples from the waveform Wi-N,
That is, the waveform data selected for every four samples is shown in FIG. 2B, but these waveform data are stored in the waveform memory 12
It should be stored in 5. Also, the waveform memory 125
Is the same as the waveform memory 126. FIG. 2D shows the values obtained by linearly interpolating the waveforms Wi-N and Wi, and the difference data between the waveform data from the address (i-N + 1) to the address i in FIG. 2A and the waveform Wi-N. However, similarly, the difference data between the linearly interpolated values of the waveform Wi and the waveform Wi + N and the waveform data of the address (i + 1) to the address (i + N) are also shown.

It is assumed that the vector memory 222 stores vectors V0 to V3 and V4 to V7 (not shown) having four elements corresponding to the skip data N as shown in FIG. For example, the vector V0 is (0.00,
It is composed of four elements (0.33, 0.67, 1.00), and for addresses 0 to 3, sequentially from the first element (0.00, 0.33, 0.67, It is stored as 1.00). Also, the vector V1 is (0.00, 0.5
0, 1.00, 0.50) and four elements are stored sequentially for the addresses 4 to 7 from the first element. Similarly, it is assumed that the four elements forming the vector V3 are sequentially stored from the first element for the addresses 12 to 15. The vectors V4 to V7 are assumed to be stored in the addresses 16 to 31 as data in which the signs of the vectors V0 to V3 are inverted. The vector memory 222 is the same as the vector memory 223.

It is assumed that the index memory 201 stores index data corresponding to the waveform data stored in the waveform memory 125. That is, the index data Ii corresponding to the waveform data Wi is composed of the vector index VIi and the level data Li. The vector index VIi is the vector V0 to V
It consists of 3 bits to indicate 7. In this embodiment, there are eight types of vectors, but this can be increased or decreased as needed. Since the number N of jumps in the waveform data is 4, the sampling frequency is 4
At 0 kHz, the fluctuation of the frequency component of 10 kHz or more contained in the instrument sound is largely reflected in the fluctuation of the difference data as shown in FIG.
The frequency component of 10 kHz or more included in the musical instrument sound is very small, and the difference between the waveform data stored in the waveform memories 125 and 126 and the interpolated waveform is obtained in order to obtain the difference data (see FIG. 2D). As a result, the elements at both ends of the vector are normalized so as to be converged in the vicinity of 0. Therefore, the types of vectors to be stored may substantially be about eight types in this embodiment. That is,
For the difference data in FIG. 2D, the address (i
The four difference data of +1) to (i + N) are vector V
2 and the address (i-N
Since the differential data from +1) to i can be approximated by the vector V0, the vector indexes of the index data corresponding to the waveform data Wi-N and Wi are 0 and 2, respectively. However, it is very effective to increase the number of elements forming the index data in order to improve the interpolation accuracy of the waveform and reduce the noise caused by the interpolation. Therefore, the number of elements is set to 8 or 16. It is also possible to improve the interpolation accuracy or omit 1024 and omit the interpolation calculation itself between the index data elements. Further, the level data Li is a 12-bit unsigned data, and determines the size of using the vector designated by the vector index VIi. That is, FIG.
Regarding the difference data in (d), the waveform data Wi-N
And the level data of the index data corresponding to Wi are smaller than (Wi-Wi-N) and (Wi + N-Wi), respectively.

The operation of the tone synthesizer configured as above will be described below. Similarly to the conventional example, the address generator 100 cumulatively adds the instructed pitch information S and outputs the addresses A0 to A29. Waveform memory 125
The number of samples of the waveform data (see FIG. 2 (b)) stored in the waveform data is 1/4 (N =
4), the addresses A12 to A29 are stored in the waveform memory 12
5 and the adder 130, and the waveform memory 126 has addresses A12 to A increased by 1 by the adder 130.
29 is supplied, and thereafter, by the same operation as in the conventional example, when the values of the addresses A12 to A29 are (i-1), the waveform data Wi-1 is output from the waveform memory 125, and the waveform memory 126 is output from the waveform memory 126. The waveform data Wi is output, the difference between the two waveform data is output from the adder 131, and the multiplier 140 performs multiplication with the addresses A2 to A11 as a decimal part, and then the adder 132 outputs the waveform data Wi. -
Since it is added with 1, the waveform memories 125 and 126 are eventually added.
The linear interpolation value Wp by the addresses A2 to A11 of the waveform data output from and is calculated.

Similarly, the addresses A12 to A29 are supplied to the index memory 201, and the vector index VIi forming the index data Ii and the level data Li are output.

On the other hand, the adders 134 to 136, the vector memories 222 and 223, and the multiplier 145 are constructed in the same manner as in the first embodiment, and the vector memory 222.
Is the address A10 output from the address generator 100.
~ A11 as the lower address, index memory 201
The elements constituting the vector determined according to the total 6-bit address having the vector index VIi output from the above as the upper address are output. Vector memory 22
3 also outputs the elements constituting the vector by the same operation, but the value obtained by adding 1 to the addresses A10 to A11 in the adder 134 is set as the lower 2 bit address. When the values of the addresses A10 and A11 are both 0, the value of the element of the vector data stored in the vector memories 222 and 223 is 0. Therefore, address A10 ~
As the value of A11 changes (0 → 1 → 2 → 3), the vector memories 222 and 223 respectively
(First element of designated vector → second element → third element) and (second element of designated vector → third element → fourth element → value 0) are output. These outputs are linearly interpolated by the adders 135 and 136 and the multiplier 145 which are configured similarly to the adders 131 and 132 and the multiplier 140 of the conventional example, and then multiplied by the level data in the multiplier 146, Data corresponding to the linearly interpolated value of the difference data of FIG. 2D is further added to the waveform data output from the waveform memory 125 in the adder 133, and the linearly interpolated waveform is eventually obtained.

In the second embodiment, in order to facilitate understanding, the waveform memory for storing the waveform data is the waveform memories 125 and 126, and the vector memory for storing the vector approximating the difference data is the vector memory. Although two memories 222 and 223 for respectively storing the same content of data are used, it is obvious that one waveform memory 125 and one vector memory 222 can be configured by time division processing in the same manner as in the conventional example.

The memory capacity required for the second embodiment is the same as the first embodiment, that is, the waveform memory and the index memory which are ¼ of those of the conventional example. Compared with this, it is possible to further reduce the types of vectors to be stored.

In the above, the predetermined N is 4
However, depending on the type of musical sound, the data compression rate can be further increased by changing the value of N to another value. Furthermore, since the difference value of waveform data that are temporally close in a musical tone waveform is generally smaller than the word length that forms the waveform data, the word length of the elements that form the vector is smaller than the word length of the waveform data. Since it can be represented by, the storage word length of the vector memory can be reduced.

As described above, according to the second embodiment, the index data and the interpolation value of the waveform data read from the waveform memory that stores only N waveform data according to the predetermined N are specified. Since the interpolation value is calculated from the elements forming the vector, it is possible to synthesize the musical sound with a smaller memory and a simple structure.

The third embodiment of the present invention will be described below with reference to the drawings. FIG. 6 is a block diagram of a musical sound synthesizer according to the third embodiment of the present invention. In FIG.
110 and 111 are waveform memories, 300 is a shifter, and 301
Is a skip data generator. The address generator 100, the adders 130 to 132, and the multiplier 140 have the same configuration as the conventional example.

In general, the rising part of a musical instrument sound, in particular, the rising part of a musical instrument sound of a damping system such as a piano contains a large amount of high frequency components. High frequency components are reduced. Therefore, by changing the interlacing number N of the predetermined waveform data in the conventional example or the first and second embodiments with time, it is possible to control the degree of follow-up of the compression technique to the sound waveform of the musical instrument.

FIG. 8A shows piano waveform data.
Further, FIG. 8B shows a change in the number N of skips of the waveform data (skip data) necessary for the waveform data of the piano.

In the waveform memory 110, with respect to the waveform data of the piano shown in FIG. 8A, the skip data N1 is set to 1, that is, all the waveform data up to time T1 are stored as the waveform data. From time T1 to time T2, the skip data N2 is set to 2, that is, one waveform data is stored for every two samples. Similarly, from the time T3 to the completion of the combining operation, the skip data N4 is set to 8 and the waveform data is stored. The waveform memory 111
Is the same as the waveform memory 110.

The skip data generator 301 is constructed as shown in FIG. In FIG. 7, reference numerals 230 and 231 are skip memories, 149 are multipliers, and 190 to 192 are adders.

The operation of the tone synthesizer configured as above will be described below. Skip data generator 30
1 is the address A output from the address generator 100.
Addresses A28 to A29 are input to the skip memory 230 by inputting 0 to A29. The skip data N1 to N5 of FIG. 8B are stored in the skip memory 230, and the skip memory 230 and the skip memory 231 are the same. When the skip data Nk corresponding to the input addresses A28 to A29 is output from the skip memory 230, the adder 19 is output from the skip memory 231.
The skip data Nk + 1 corresponding to the address increased by 1 by 0 is output. From the adder 191 (Nk + 1
-Nk) is output and the multiplier 149 performs multiplication with the addresses A0 to A27 as the fractional part, and the adder 19
Since Nk is added at 2 and the value of the higher 4 bits is output, the skip data Nk and Nk + 1 are eventually output as skip data Ni which is linearly interpolated by the addresses A0 to A27.

The shifter 300 includes a skip data generator 3
Since the input pitch information S is divided according to the value of the skip data Ni output from 01,
When the skip data has a value of 1, the pitch information S remains unchanged,
Further, when the skip data is 2 or 3, the pitch information S is shifted by 1 or 2 bits so as to be reduced and outputted. The pitch information S bit-shifted in the shifter 300 is the address generator 100.
However, the contribution of the pitch information S to the output address information is influenced by the bit shift amount by the shifter 300.

As described above, since the degree of influence of the pitch information S on the address output from the address generator 100 is controlled by the skip data Ni, the waveform memory 1
The waveform data output from the addresses A0 to A29 supplied to 10, 111 and their interpolated values become the waveform data stored for every Ni data and their interpolated values according to the skip data Ni.

In the third embodiment, in order to facilitate understanding, the waveform memories for storing the waveform data are the waveform memories 110 and 111, and two waveform memories for storing the same data are used. It is obvious that one waveform memory 110 can be configured by time division processing in the same manner as in the example.

Also in the third embodiment, as in the first and second embodiments, by the index data,
The waveform accuracy can be further improved by using vector data.

As described above, according to the third embodiment, the address output from the address generator is changed corresponding to the skip data Ni which changes with time. Since the waveform output can be obtained by interpolating the waveform data read from the waveform memory in which the waveform data of No. 1 is stored for each skip data Ni sample, it is appropriate to change the skip data according to the waveform of the musical tone. Data compression can be realized.

[0056]

As described above, the musical tone synthesizer of the present invention can generate a waveform memory in which a series of waveform data of musical tones is stored for every N samples and waveform data in which waveform data not stored in the waveform memory is stored in the waveform memory. By storing the data for indexing as dependent data in the index memory, the waveform data output from the waveform memory and the index memory are stored in the index memory based on the address output of the address generator that changes according to the pitch. Since the vector data corresponding to the difference data specified by the stored data is read and the output waveform is calculated, the musical tone waveform can be decoded with a small memory.

With respect to the waveform data stored in the waveform memory for every N samples, the value of N is changed with time according to the musical tone waveform, so that the sound quality of the waveform output is uniform over the entire period from the start to the end of the waveform output. Data can be compressed so that

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a musical sound synthesizer according to a first embodiment of the present invention.

FIG. 2 is an exemplary waveform diagram for explaining the operation in the first and second embodiments of the present invention.

FIG. 3 is an exemplary waveform chart for explaining the operation in the first embodiment.

FIG. 4 is a block diagram showing the configuration of a musical sound synthesizer according to a second embodiment of the present invention.

FIG. 5 is an exemplary waveform diagram for explaining an operation in the second embodiment.

FIG. 6 is a block diagram showing a configuration of a musical sound synthesizing apparatus according to a third embodiment of the present invention.

7 is a block diagram showing an internal configuration of a skip data generation unit in FIG.

FIG. 8 is an exemplary envelope waveform diagram for explaining the operation in the third embodiment.

FIG. 9 is a block diagram showing a configuration of a musical sound synthesizing device in a conventional example.

FIG. 10 is a schematic diagram showing an exemplary data structure in a conventional example.

FIG. 11 is a block diagram showing the configuration of a musical tone synthesizer to which a conventional example is applied.

[Explanation of symbols]

 100 Address generator 110-111, 120-126 Waveform memory 130-136, 190-192 Adder 140-149 Multiplier 200-201 Index memory 220-223 Vector memory 230-231 Skip memory 250-251 Comparator 260-262 Gate 300 Shifter 301 Skip data generator

Claims (6)

[Claims] 1. A waveform memory for storing waveform data for each predetermined number N of data from a series of waveform data of musical tones, and a vector memory for storing a plurality of sets of vectors composed of N or more elements. An index memory for storing index data for specifying a vector and a vector stored in the vector memory corresponding to difference data between the waveform data stored in the waveform memory and N waveform data in the vicinity thereof, and An address generator that outputs addresses that sequentially change according to the pitch of a musical tone to be played, and an interpolation operation that interpolates the elements of the vector output from the vector memory according to the index data of the index memory that corresponds to the address. Section, waveform data of the waveform memory corresponding to the address, and output of the interpolation calculation section Tone synthesis device comprising an adder for adding. 2. The musical tone according to claim 1, wherein the adder adds and outputs the waveform data output from the waveform memory and the vector element output from the vector memory, and the interpolation calculation unit performs interpolation calculation on the addition output. Synthesizer. 3. A waveform memory for storing waveform data for each predetermined number N of data among a series of waveform data of musical tones, and a vector memory for storing a plurality of sets of vectors composed of N or more elements. Storing the vector stored in the vector memory corresponding to the difference data between the data interpolated by the waveform data stored in the waveform memory and the N waveform data in the vicinity thereof and the index data specifying the size thereof An index memory, an address generator that outputs addresses that sequentially change according to the pitch of a musical tone that is output, and a first interpolation calculation unit that performs an interpolation calculation on the waveform data of the waveform memory that corresponds to the address. , The vector output from the vector memory according to the index data of the index memory corresponding to the address. A second interpolation operation unit for interpolation calculation elements of torque, the first and musical tone synthesizing apparatus constituted by a second interpolation operation unit adder for adding the output. 4. The musical tone synthesizing apparatus according to claim 1, wherein the word length of the element forming the vector stored in the vector memory is smaller than the word length of the waveform data stored in the waveform memory. . 5. A skip data generator for outputting skip data Ni that changes with time, a waveform memory for storing waveform data for each of the skip data Ni among a series of waveform data of musical sounds, and a tone sound to be output. An address generator that outputs addresses that sequentially change according to the high level and the skip data Ni, and an interpolation calculation that interpolates the waveform data output from the waveform memory corresponding to the addresses and the skip data A musical sound synthesizer composed of parts. 6. The musical tone synthesizing apparatus according to claim 5, wherein the value of the skip data Ni which changes with time increases with the passage of time.