JP2003202882A - Method and device for sound synthesis - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To generate sound data with a necessary pitch not through pitch conversion using a digital filter having a large arithmetic load. <P>SOLUTION: A device for sound synthesis has a short-cycle sampling sound element storage part 1 which performs sampling for each of sound elements in sampling cycles shorter than reference sampling cycles and holds amplitude values corresponding to short-cycle sampling points obtained by the sampling as short-cycle sampling sound elements, an output sound element generation part 2 which obtain short-cycle sampling sound elements to be processed from the held short-cycle sampling sound elements, obtains waveform data for each of short-cycle sampling points corresponding to waveform data acquisition cycles specified with input pitch information for the sound elements, and reproduces the waveform data as waveform data for each of the reference sampling cycles, and a phase control processing part 3 which determines from which short-cycle sampling point a start is made by taking the consistency to waveform data at the end of the last sound element into consideration. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sound synthesizing method and a sound synthesizing apparatus for generating a sound waveform such as voice or musical instrument sound.

[0002]

2. Description of the Related Art There are roughly two methods for sound synthesis for generating sound waveforms of voices and musical instrument sounds. One of them is a method of synthesizing a sound waveform by defining some sound source model and controlling its parameters, which is called rule synthesis. Also,
Another method is to record the synthesized voice or the actual sound of the instrument and process / edit the waveform data to generate the desired sound. This is called the parameter editing method or recording editing method. There is.

For the former rule composition, various methods have been proposed, from a simple method with a small amount of calculation to a method with an extremely large amount of calculation to obtain a physical model of a sound source strictly, but satisfactory results are obtained. Generating quality sound requires a great deal of computation, and current technology often does not provide a sufficient physical model.

On the other hand, in the latter parameter editing method, various necessary sounds are terminated in advance, and sound data generated from the recorded sounds (this is called a sound element) is held. Then, by connecting these sound pieces appropriately, the necessary sound can be easily generated.

However, in this parameter editing method, combinations such as sound generation of musical instruments such as electronic pianos in which there are many combinations of sounds that are played simultaneously and text synthesis of human voices are performed. In the case of innumerable numbers, it is difficult to hold the data of all combinations of them. In such a case, it is necessary to generate necessary sound from the held sound element.

[0006] Here, timbre (frequency component), sound volume, pitch (pitch of pitch), and duration (sound speed) are important in generating a necessary sound. In the case of the parameter editing method, the timbre does not need to be considered because it is included in the sound element. Also, the volume and duration are easy to control. On the other hand, it is generally not easy to generate sound data of various pitches without impairing quality and timbre information.

In sound synthesis processing of the conventional parameter editing system, when data of different pitches are reproduced, a method of preparing those data in advance and a method of performing pitch conversion from a small number of sound segments to obtain a required pitch There is a way to generate data.

[0008]

In the former method of preparing data of various pitches in advance, it is necessary to store data of all necessary pitches in a memory,
There is a problem that the amount of memory consumption increases and the cost of hardware increases.

Further, in order to generate sound data of a required pitch by pitch conversion from the latter sound segment, a conversion method such as digital filter processing is generally used, but this digital filter processing is a calculation. There is a problem in that a CPU with a small processing capacity has too large an arithmetic load because the amount is extremely large and a lot of processing time is required.

Therefore, the present invention makes it possible to generate sound data of a required pitch from a single sound segment without using a pitch conversion method that requires a large amount of calculation such as digital filter processing, and a plurality of sound data can be generated. With the purpose of providing a sound synthesizing method and a sound synthesizing device, it is possible to obtain a high-quality synthesized voice with less noise by performing a phase control that smoothly connects the connected parts when connecting the sound elements. There is.

[0011]

In order to achieve the above-mentioned object, the first invention of the sound synthesizing method of the present invention is a sound for which waveform data for each preset sampling cycle is obtained. In a sound synthesizing method for generating sound data of an arbitrary pitch by using a plurality of pieces and converting the pitch of the sound pieces, for each sound piece, 1 / of the reference sampling period is used. By sampling at a short sampling period of n (n> 1), waveform data corresponding to each short-period sampling point (short-period sampling point) is acquired and held as a short-period sampling sound segment. In addition, from the stored individual short-period sampling sound segments, the short Obtain a sampling sound segment, acquire waveform data for each short-period sampling point corresponding to the waveform data acquisition period specified by the input pitch information, for the acquired short-period sampling sound segment, and acquire it. The waveform data is output as the waveform data for each of the preset reference sampling periods.

In the first invention of such a sound synthesizing method, when acquiring the waveform data for each short period sampling point corresponding to the waveform data acquisition period designated by the input pitch information, which waveform data acquisition is performed is performed. Performs phase control processing to determine whether to start from the short-period sampling point, and the phase control processing is the waveform data of the end part of the sound element used immediately before the sound element currently being processed and the current processing. This is done by considering the consistency with the waveform data of each short period sampling point of the sound element.

The second aspect of the sound synthesizing method of the present invention is to use a plurality of sound elements for which waveform data for each preset sampling cycle, which is a reference, is used, and to determine the pitch of the sound elements. In a sound synthesis method for generating sound data of an arbitrary pitch by converting a sampling point having a short sampling period of 1 / n (n> 1) of the reference sampling period with respect to a predetermined sound segment. (Short cycle sampling point) is generated, each waveform data corresponding to each short cycle sampling point is obtained, and it is set as short cycle sampling sound element data, and the input pitch is input to the short cycle sampling sound element. Waveform data for each short period sampling point corresponding to the waveform data acquisition period specified by the information Acquired by, and to output the acquired waveform data as waveform data of each of the reference sampling period set in advance.

In the second invention of such a sound synthesizing method, the waveform data corresponding to each short cycle sampling point is linearly interpolated between the waveform data obtained for each of the reference sampling cycles. I try to ask for it.

In the second invention of the sound synthesizing method, when acquiring the waveform data for each short cycle sampling point corresponding to the waveform data acquisition cycle designated by the input pitch information, which waveform data acquisition is performed Performs phase control processing to determine whether to start from the short-period sampling point, and the phase control processing is the waveform data of the end part of the sound element used immediately before the sound element currently being processed and the current processing. The sound element should be considered in consideration of its consistency with the waveform data at each short-period sampling point.

The third aspect of the sound synthesizing method of the present invention is to use a plurality of sound elements for which waveform data for each preset sampling cycle is obtained, and to set the pitch of the sound elements. In the sound synthesizing method for generating sound data of an arbitrary pitch by converting the reference sampling cycle of the sound element to be subjected to the pitch conversion processing into 1 / n according to the designation of the input pitch information. As (n> 1), sampling points with a short sampling period (short period sampling points)
The waveform data of the short cycle sampling points corresponding to the waveform data acquisition cycle specified by the input pitch information is obtained, and the obtained waveform data is output as the waveform data for each of the preset reference sampling periods. I am trying to do it.

In the third invention of such a sound synthesizing method, the waveform data at the short period sampling points corresponding to the waveform waveform data acquisition period designated by the pitch information has a reference sampling period of 1 / n. And the value of n and the short period sampling point corresponding to the waveform waveform data acquisition period are in the section of the reference sampling period, the slope of the sound element waveform is acquired. Is prepared, and the slope obtained by the table is used to obtain the waveform data of the short-period sampling point.

Further, in the third invention of the sound synthesizing method, when acquiring the waveform data for each short period sampling point corresponding to the waveform data acquisition period designated by the input pitch information, which waveform data acquisition is performed is performed. Performs phase control processing to determine whether to start from the short-period sampling point, and the phase control processing is the waveform data of the end part of the sound element used immediately before the sound element currently being processed and the current processing. In consideration of the consistency with the waveform data of the short-period sampling points designated by the pitch information in the sound element.

The first aspect of the sound synthesizing device of the present invention is to use a plurality of sound elements for which waveform data for each preset sampling cycle, which is a reference, are used, and to set the pitch of the sound elements. In the sound synthesizing device for generating sound data of arbitrary pitch by sampling each sound segment with a short sampling period of 1 / n (n> 1) of the reference sampling period. As a result, a short cycle sampling sound element storage unit that holds waveform data corresponding to each short cycle sampling point (short cycle sampling point) as a short cycle sampling sound element, and the short cycle sampling sound element storage From the individual short-period sampling sound pieces held in the section, Acquire the short-period sampling sound segment to be processed, and for the acquired short-period sampling sound segment, obtain the waveform data for each short-period sampling point corresponding to the waveform data acquisition period specified by the input pitch information. And an output sound element generator that outputs the acquired waveform data as waveform data for each of the preset reference sampling periods.

In the first invention of such a sound synthesizer, the output sound segment generator acquires waveform data for each short period sampling point corresponding to the waveform data acquisition period designated by the input pitch information. At this time, a phase control processing unit that performs phase control processing for determining from which short cycle sampling point to start the waveform data acquisition is provided, and the phase control processing performed by the phase control processing unit is the sound currently being processed. This is performed in consideration of the consistency between the waveform data at the end of the sound segment used immediately before the segment and the waveform data at each short cycle sampling point of the sound segment currently being processed.

A second aspect of the sound synthesizer of the present invention is to use a plurality of sound elements for which waveform data has been obtained for each preset sampling cycle, and the pitch of the sound elements is used. In a sound synthesizer that generates sound data of an arbitrary pitch by converting a sampling point with a short sampling period of 1 / n (n> 1) of the reference sampling period for a predetermined sound segment. (Short-cycle sampling point) is generated, each waveform data corresponding to each short-cycle sampling point is obtained, and the short-cycle sampling sound segment generator that uses it as short-cycle sampling sound segment data and its short Corresponds to the waveform data acquisition period specified by the input pitch information for the periodic sampling sound element That acquires the waveform data of each short cycle sampling point, and an output sound segment generator outputting the acquired waveform data as waveform data of each reference sampling period the preset.

In the second aspect of the sound synthesizing apparatus, the waveform data corresponding to each of the short period sampling points is linearly interpolated between the waveform data obtained for each of the reference sampling periods. I try to ask by doing.

Further, in the second invention of the sound synthesizer, the output sound segment generator generates the waveform data for each short cycle sampling point corresponding to the waveform data acquisition cycle designated by the input pitch information. When acquiring from the periodic sampling sound segment data generating means, a phase control processing unit for performing a phase control process for determining from which short cycle sampling point the waveform data acquisition is started is provided. The phase control processing to be performed considers the consistency between the waveform data at the end of the sound element used immediately before the sound element currently being processed and the waveform data at each short cycle sampling point of the sound element currently being processed. Then do.

The third aspect of the sound synthesizer of the present invention is to use a plurality of sound elements for which waveform data has been obtained for each preset sampling cycle, and the pitch of the sound elements is used. In a sound synthesizer for generating sound data of an arbitrary pitch by converting the reference sampling cycle of the sound element to be subjected to the pitch conversion processing into 1 / n according to the designation of the input pitch information. As (n> 1), sampling points with a short sampling period (short period sampling points)
The waveform data of the short cycle sampling points corresponding to the waveform data acquisition cycle specified by the input pitch information is obtained, and the obtained waveform data is output as the waveform data for each of the preset reference sampling periods. It has an output sound element generator for

In the third invention of the sound synthesizing apparatus, the output sound segment generating means performs the waveform data generation processing of the short cycle sampling points corresponding to the waveform waveform data acquisition cycle designated by the input pitch information. , The value of n when the reference sampling period is 1 / n, and the short period sampling point corresponding to the waveform data acquisition period, depending on in which section of the reference sampling period, A table capable of acquiring the inclination of the sound element waveform is prepared, and the inclination data obtained from the table is used to obtain the waveform data of the short cycle sampling point.

In the third invention of the sound synthesizer, the output sound segment generator acquires waveform data for each short cycle sampling point corresponding to a waveform data acquisition cycle designated by input pitch information. At this time, a phase control processing unit that performs a phase control processing for determining from which short cycle sampling point to start acquiring the waveform data is provided, and the phase control processing performed by this phase control processing unit is the sound currently being processed. The waveform data at the end of the sound segment used immediately before the segment and the waveform data at the short period sampling point specified by the pitch information in the sound segment currently being processed should be considered in consideration of consistency. ing.

As described above, according to the first invention of the sound synthesizing method and the sound synthesizing apparatus, the waveform data (amplitude value) recorded for each individual sound segment at a cycle shorter than the reference sampling cycle. Is stored as a short-cycle sampling sound segment, and when obtaining sound data of an arbitrary pitch, the waveform data corresponding to the sampling point (short-cycle sampling point) of that short cycle is set for each predetermined cycle. By acquiring and outputting the acquired waveform data at the original sampling cycle (reference sampling cycle), sound data with an arbitrary pitch can be obtained. As a result, it is possible to generate sound data having a required pitch by performing an extremely simple calculation as compared with the conventional pitch conversion using a digital filter.

In the first invention of the sound synthesizing method and the sound synthesizing apparatus, when the waveform data for each short cycle sampling point corresponding to the waveform data acquisition cycle designated by the input pitch information is acquired, Perform phase control processing that considers the consistency between the waveform data at the end of the sound element used immediately before the sound element being processed and the waveform data at each short cycle sampling point of the sound element currently being processed. Therefore, when several sound elements are connected and used, the connection parts of each sound element can be connected smoothly, noise is prevented, and high quality synthetic sound is generated. be able to.

Further, in the second invention of the sound synthesizing method and the sound synthesizing apparatus of the present invention, since it is sufficient to generate the short-period sampling sound segment only for the necessary sound segment and hold it. It is not necessary to hold all short-cycle sampling sound elements corresponding to all sound elements, and memory (ROM) for that purpose.
Can be made smaller.

In the second invention of the sound synthesizing method and the sound synthesizing apparatus, the waveform data (amplitude value) corresponding to each short cycle sampling point is obtained for each of the reference sampling cycles. Since it can be obtained by linearly interpolating between the waveform data, it is not necessary to perform complicated calculation to generate the short-period sampling sound element,
It does not significantly affect the calculation of the entire sound synthesis process.

In the second invention of the sound synthesizing method and the sound synthesizing apparatus, when the waveform data for each short period sampling point corresponding to the waveform data obtaining period designated by the input pitch information is acquired, Perform phase control processing that considers the consistency between the waveform data at the end of the sound element used immediately before the sound element being processed and the waveform data at each short cycle sampling point of the sound element currently being processed. Therefore, when several sound elements are connected and used, the connection parts of each sound element can be connected smoothly, noise is prevented, and high quality synthetic sound is generated. be able to.

The sound synthesizing method and the sound synthesizing apparatus according to the third aspect of the present invention respectively include sampling points (short period sampling) having a short sampling period in which a reference sampling period is arbitrarily divided based on designated pitch information. The waveform data of the short-period sampling points corresponding to the waveform waveform data acquisition period specified by the pitch information is generated, and the obtained waveform data is waveform data for each of the preset reference sampling periods. Trying to play as. In this way, it is possible to freely set how many times the reference sampling period is divided, and thus the degree of freedom in setting the sampling period is high, and more versatile pitch data can be generated.

In the third invention of the sound synthesizing method and the sound synthesizing apparatus, the waveform data at the short cycle sampling points corresponding to the waveform data acquisition cycle designated by the pitch information has the reference sampling cycle. The slope of the sound element waveform is acquired depending on the value of n when 1 / n is set and in which section of the reference sampling cycle the short cycle sampling point corresponding to the waveform data acquisition cycle exists. Since the table that can be used is prepared and the waveform data of the short-period sampling point is obtained by using the slope obtained by the table, complicated calculation is required to generate the short-period sampling sound segment. There is no need to perform it, and it does not significantly affect the calculation of the entire sound synthesis process.

In the third invention of the sound synthesizing method and the sound synthesizing apparatus, when the waveform data for each short cycle sampling point corresponding to the waveform data acquisition cycle specified by the input pitch information is acquired, Perform phase control processing that considers the consistency between the waveform data at the end of the sound element used immediately before the sound element being processed and the waveform data at each short cycle sampling point of the sound element currently being processed. Therefore, when several sound elements are connected and used, the connection parts of each sound element can be connected smoothly, noise is prevented, and high quality synthetic sound is generated. be able to.

[0035]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below.

FIG. 1 shows the waveform of one sound element. The horizontal axis represents time and the vertical axis represents amplitude. The sampling points of this sound element are A1, A2, A3 ,.
It is indicated by A16. In addition, the time between these sampling points, that is, the sampling period is T.

It is considered that, in such a certain sound element, the pitch of the sound element is variously changed without changing the tone color. In each of the following embodiments, a case where the pitch is set to n times is described, but here, an example in which n> 1 is set, that is, the sound waveform is set along the time axis without losing the waveform information. The case of shortening by the following will be described. In the following description, the original sound segment as shown in FIG. 1 is referred to as a reference sound segment, and its sampling period T is the reference sampling period T, and sampling points A1, A2, A3 for each of the reference sampling periods T. , ... are referred to as reference sampling points.

[First Embodiment] First, a first embodiment of the sound synthesizing method and the sound synthesizing apparatus of the present invention will be described. In the first embodiment, for example, in a reference sound segment as shown in FIG. 1, a sound segment recorded at a sampling period shorter than the reference sampling period T of the reference sound segment is held. is there.

For example, as shown in FIG. 2, the sampling frequency is multiplied by 6, that is, the reference sampling period is T /
The actual sound is recorded as 6, and in the case of FIG. 2, the reference sampling cycle T is divided into 6 equal parts, so the sampling points obtained by this (this is called the short cycle sampling points a1, a2, The waveform data (hereinafter referred to as the amplitude value) corresponding to a3, ...

The short period sampling points a1 and a
The sampling cycle of 2, a3, ... Is represented by t. In addition, such short period sampling points a1,
A sound element having an amplitude value corresponding to a2, a3, ... Is called a short-period sampling sound element.

As described above, the sampling period t shorter than the normal sampling period (reference sampling period T)
By holding the amplitude value sampled by the short-period sampling sound segment as the short-period sampling sound segment, for example, when the pitch of the reference sampling segment is varied, the short-period sampling sound segments a1, a2, a3 , ... are thinned out by a predetermined number to obtain the amplitude value at each remaining short cycle sampling point, and the amplitude value is output at the reference sampling cycle T, so that the sound data at various pitches is obtained. Can be generated.

For example, a short-period sampling sound element whose amplitude value as waveform data has been acquired at a sampling period t as shown in FIG. 3A (assumed to be the same short-period sampling sound element as in FIG. 2). , The short period sampling points a1, a2, a3, ... Are seven times as long, that is, the sampling points a1, a2, a3 ,.
It is assumed that every 7 points are sampled (amplitude value acquisition as waveform data) after the sampling point a1.

In this case, the short-period sampling points to be sampled are a1, a8, a15, a22, ... And when they are output at the reference sampling period T, FIG.
As shown in (b), the sound waveform has a pitch of 7/6 of the pitch shown in (a) of the figure without losing the waveform information of the reference sound segment.

By the way, in the sound synthesis by the parameter editing method, it is general that a desired sound waveform is generated by connecting several sound elements, so that the connection part between the sound element parts is smooth. It is well known that if not connected, noise or the like is generated at the connecting portion.

In order to deal with this, in the present invention, the end of the sound element used immediately before the short-period sampled sound element being processed (connected to the beginning of the sound element currently being processed). (A part), and information such as whether the amplitude value is decreasing and reaching that amplitude value or rising and reaching that amplitude value, etc. is received as termination phase information, and based on that termination phase information , A process for controlling which short-period sampling point of the short-period sampling sound segment currently being processed to connect to the most smoothly is performed (this is called a phase control process).

That is, the short cycle sampling point currently being processed is the short cycle sampling point at which the connection between the end portion of the immediately preceding sound element and the short cycle sampling sound element currently being processed is the most natural connection. Find from the sound element and use it as the beginning of the short-period sampled sound element. However, the starting end is shown in FIG.
In, when the position is advanced too far on the time axis (to the right in the drawing), the waveform shape of the short-period sampling sound element is lost, so an appropriate limit is set.

In this way, short-period sampling that maximizes the consistency of connection between the amplitude value at the end of the immediately preceding sound segment and the amplitude value at the beginning of the short-cycle sampling sound segment currently being processed. When the point is determined as the starting end portion, the pitch converting process as described in FIG. 3 is performed from the starting end portion. For example, if it is determined that the short cycle sampling point a3 is optimum as the starting end portion, as described above, when sampling every seven short cycle sampling points, the short cycle sampling point a3 is set as the starting point, Thereafter, sampling (acquisition of an amplitude value) is performed every seven pieces.

By performing the phase control processing in consideration of the consistency of the connection between the sound elements as described above, the connection portions are smoothly connected and a large gap or the like is not generated, so that the noise is suppressed and the quality is high. Can reproduce synthetic sound.

FIG. 4 shows the configuration of the main part of the sound synthesizing device according to the first embodiment. The short-cycle sampling sound element storage unit 1, the output sound element generation unit 2, and the phase control processing unit 3 are shown. have.

Short cycle sampling sound segment storage unit 1
Is a storage means such as a ROM for storing sound segments sampled at a sampling period shorter than the reference sampling period T (short-period sampled sound segments). For example, according to the sampling period t as shown in FIG. A large number of sampled short-period sampling sound pieces are stored.

The output sound segment generator 2 acquires the short-period sampling sound segment to be processed at that time from the short-period sampling sound segment storage unit 1 and acquires the acquired short-period sampling sound segment. On the other hand, based on the input pitch information (information indicating what kind of pitch is used), the phase determination information determined by the phase control processing unit 3 (which will be described later) is used. From the short cycle sampling point corresponding to, the pitch conversion processing is performed on the short cycle sampling segment. Note that the pitch conversion processing here is the processing described with reference to FIG. 3 starting from the short cycle sampling point corresponding to the phase determination information.

The phase control processing section 3 reaches the amplitude value by making such a change with the amplitude of the end portion of the sound element used immediately before the short-period sampled sound element being processed. The end phase information indicating whether or not it is input is input, and which position of the short cycle sampling point of the short cycle sampling sound element to be processed acquired by the output sound element generation unit 2 is to be the start point based on the end phase information. To decide. In other words, it determines which short-period sampling point of the short-period sampling sound segment being processed has the best match with the amplitude value at the end of the immediately preceding sound segment, and uses that to determine the phase determination information. To the output sound segment generator 2.

In such a configuration, when the output sound segment generator 2 acquires the short-cycle sampling sound segment to be processed from the short-cycle sampling sound segment storage unit 1, the short-cycle sampling sound segment is generated based on the input pitch information. Pitch conversion processing is performed on the sampled sound element, but at this time, based on the phase determination information passed from the phase control processing unit 3, a short cycle determined to have the best consistency with the end portion of the immediately preceding sound element. The pitch conversion process is started with the sampling point as the starting point.

For example, suppose that the same short-period sampling sound segment as shown in FIG. 2 is acquired as the short-period sampling sound segment to be processed, and the pitch information is 7 times the sampling period t and the short-period sampling point is 7 times the period. It is assumed that it is specified that a1, a2, a3, ... Are sampled and the amplitude value is acquired every sampling period 7t. Further, it is assumed that the phase control processing unit 3 determines that the short-period sampling point a3 of the short-period sampling sound segment has good consistency with the end portion of the immediately preceding sound segment.

As a result, the output sound segment generator 2
Takes the short cycle sampling point a3 as the sampling start point, and first obtains the amplitude value of the short cycle sampling point a3 and then the amplitude value of the short cycle sampling point a10 with the short cycle sampling point a3 as the starting point. , And then obtain the amplitude value of the short-period sampling point a17,
Sampling is performed every seven points starting from the short cycle sampling point a3.

In this way, starting from the short cycle sampling point a3, every seven short cycle sampling points a1 thereafter.
When the amplitude values corresponding to 0, p17, ... Are acquired and reproduced at the reference sampling period T, in this case, the sound waveform reproduced by the output sound segment output from the output sound segment generation unit 2 is generated. Can be a waveform having a pitch of 7/6 of the reference sound segment without losing the information on the shape of the waveform of the reference sound segment.

At this time, the sampling start point is instructed by the phase control processing unit 3 so as to have the short cycle sampling point (in this example, the short cycle sampling point a3) which has the best matching with the immediately preceding sound element. Because,
It is possible to smooth the connection with the immediately preceding sound element, prevent the generation of noise due to gaps, etc., and obtain high-quality synthesized sound.

[Second Embodiment] In the first embodiment described above, a sound element in which waveform data is recorded at a sampling frequency higher than the normal sampling frequency (sampling period t shorter than the reference sampling period T). Although the pieces are generated in advance and stored in the ROM or the like, in the second embodiment of the sound synthesizing method and the sound synthesizing apparatus of the present invention, the processing is performed in the sound synthesizing apparatus. Then, an amplitude value corresponding to a short-period sampling point for each sampling period t shorter than the reference sampling period T is obtained from the reference sound segment necessary for the above, and the obtained waveform data is used as a short-period sampling sound segment. I have to.

FIG. 5 shows the structure of the main part of the sound synthesizer according to the second embodiment of the present invention. The point that it has an output sound segment generator 2 and a phase control processor 3 is the same as that shown in FIG. The sound synthesizing device according to the first embodiment is similar to the sound synthesizing device according to the first embodiment, but the sound synthesizing device according to the second embodiment has a normal sound segment (a reference having waveform data for each sampling period T as shown in FIG. 1). (Reference sound element storage unit 11 that stores a large number of sound elements), and if necessary, from among the reference sound element storage units 11 stored in the reference sound element storage unit 11, the short as shown in FIG. It has a short-period sampling sound segment generator 12 that generates a periodic sampling sound segment.

Short-cycle sampling sound segment generator 1
As described above, 2 generates a short-period sampling sound segment from the reference sound segment stored in the sound segment storage unit 11, and an example of the generation method will be briefly described.

The reference sound element storage unit 11 stores, for example, amplitude values corresponding to the reference sampling points A1, A2, A3, ... Of the reference sound element as shown in FIG. Such reference sampling points A1, A
Based on the amplitude value of each of A2, A3, ... By a linear interpolation method as shown in FIG.
The amplitude value corresponding to each short-period sampling point generated between 1, A2, ... Is calculated.

FIG. 6 shows a waveform of a part of a certain reference sound segment (between certain reference sampling points Ai and Aj). In the example of FIG. 6, the reference sampling is performed for the sake of simplicity. By dividing the period T into four equal parts, short-period sampling points ai1, ai2, ai3
To generate the respective short-period sampling points ai
Amplitude values corresponding to 1, ai2 and ai3 are obtained.

First, the amplitude value Vi2 corresponding to the reference sampling point Ai and the amplitude value Vj corresponding to the reference sampling point Aj are added and divided by 2, to obtain the amplitude value Vi2 of the short period sampling point ai2, and then, Its amplitude value V
i2 and the amplitude value Vi corresponding to the reference sampling point Ai are added and divided by two to obtain a short cycle sampling point a
The amplitude value Vi1 of i1 is obtained, and then the amplitude value Vi2 and the amplitude value Vj corresponding to the reference sampling point Aj are added and divided by 2, so that the amplitude value Vi3 of the short cycle sampling point ai3 is obtained. Then, the amplitude value corresponding to each of the short-period sampling points ai1, ai2, ai3 can be obtained by simply repeating a simple operation of dividing by two.

In the example of FIG. 6, an example of four equal parts has been described for simplification of description, but in reality, finer short cycle sampling points are often provided. In addition, when setting a finer short cycle sampling point,
If the reference sampling period T is equally divided by a power of 2 as in this case, the amplitude value corresponding to each short period sampling point is
As described in the above example, the simple operation of adding and dividing by 2 can be obtained by repeating the operation several times.

The reference sampling period T of the reference sound segment described here is equally divided into a plurality of new sampling points (short period sampling points), and the amplitude value corresponding to the short period sampling points is obtained. The method is shown in FIG.
However, the method is not limited to the method described above, and another method may be used.

In the sound synthesizing method and the sound synthesizing apparatus according to the second embodiment, the short-period sampling sound segment generator generates the short-period sampling sound segment from the reference sound segment as described above. In this way, the pitch conversion processing and phase control processing after the short-period sampling sound element is generated are
Since it can be performed in the same manner as the above-described first embodiment, the description thereof is omitted here.

In the second embodiment, the number of operations for performing the process of generating the short-period sampling sound element from the reference sound element for which the pitch conversion is performed is slightly increased (however, compared with the conventional digital filter method). However, it is possible to generate a short-period sampling sound segment only for the required reference sound segment and hold it, so that all reference sound segments can be stored. It is not necessary to hold all the corresponding short-cycle sampling sound pieces, and the capacity of the memory (ROM) for holding the short-cycle sound pieces can be reduced.

[Third Embodiment] Next, a third embodiment of the sound synthesizing method and the sound synthesizing apparatus of the present invention will be described.

FIG. 7 is a block diagram of a main part of a sound synthesizing apparatus according to the third embodiment of the present invention. The point that an output sound element generating unit 2 and a phase control processing unit 3 are included is shown in FIG.
5 is similar to the sound synthesizers of the first and second embodiments shown in FIG. 5, and is a reference sound element (sound element having waveform data for each sampling period T as shown in FIG. 1). The sound synthesizing device according to the third embodiment is similar to the sound synthesizing device according to the second embodiment in that it has a reference sound segment storing unit 11 for storing a large number of sound. Two
Is to calculate the amplitude value of the short cycle sampling point corresponding to the pitch information after receiving the pitch information, so the pitch conversion data necessary to calculate the data of the arbitrary short cycle sampling point. Has a pitch conversion data storage unit 21 for holding.

The pitch conversion data stored in the pitch conversion data storage unit 21 is for each individual reference sound segment necessary for calculating the amplitude value corresponding to the pitch information input to the output sound segment generation unit 2. Parameter table 21 for obtaining a slope parameter indicating the slope of the waveform of
a and the reference sampling period T of each reference sound segment are equally divided into n to generate short-period sampling points,
It is an amplitude value (this is referred to as processing start section amplitude information 21b) corresponding to each short cycle sampling point existing in the first appropriate section (referred to as a processing start section) of the reference sound segment.

The processing start section here is
In this case, for each reference sound segment, it is a predetermined section preset from the starting point, that is, an appropriate section in which the waveform shape information of the reference sound segment is not lost, and here, the processing is started. Let it be section S.

As the processing start section S, for example,
An appropriate section can be set in advance, such as one reference sampling period T, two reference sampling periods T, or three reference sampling periods T. Then, for each equal division information (information indicating how much the reference sampling period T is divided) in the processing start section S, the amplitude value corresponding to the short period sampling point generated by the equal division information is linearly calculated. It is obtained in advance by an appropriate method such as interpolation and used as the processing start section amplitude information 21b.

For example, if the equal division information is 5 divisions,
Of the short cycle sampling points generated by dividing the reference sampling cycle T into five, the amplitude value corresponding to the short cycle sampling point existing in the processing start section S is obtained, and the amplitude value is calculated as the processing start section amplitude information. 21b.
Further, when the equal division information is equal to six, the short-period sampling points existing in the processing start section S among the short-period sampling points generated by dividing the reference sampling period T into six equal to each other correspond to the short-period sampling points. The amplitude value is obtained in advance and is used as the processing start section amplitude information 21b.

The third embodiment differs from the first and second embodiments described above, and it is possible to freely set how much the reference sampling period T of the reference sound segment to be processed is divided. For example, the pitch information given to the output sound segment generator 2 is 7 of the reference sound segment.
It is possible to set an arbitrary sampling cycle such as sampling at a cycle of / 6 times or sampling at a cycle of 6/5.

It should be noted that sampling at a period of 7/6 times means that short period sampling points are generated by dividing the sampling period T of the reference sound segment into six equal parts, and the amplitude is set for every seven short period sampling points. The value is obtained, and sampling at a period of 6/5 times means that a short-period sampling point is generated by dividing the reference sampling period T of the reference sound segment into five equal parts, and the short-period sampling point is generated. That is, the amplitude value is acquired for every six items.

Such an arbitrary sampling cycle is given to the output sound segment generator 2 as pitch information.
Therefore, information on how much the reference sampling period T of the reference sound segment is divided (equal division information) and the short period sampling point corresponding to the waveform waveform data acquisition period at that time are
The pitch conversion data storage unit 21 holds a tilt parameter table 21a capable of acquiring the tilt of the sound element waveform corresponding to the short-period sampling point, depending on which section of the reference sampling period T exists, and the tilt is stored. The amplitude value of the short cycle sampling point is obtained using the tilt parameter obtained from the parameter table 21a.

FIG. 8 shows an example of the tilt parameter table 21a in a certain reference sound segment, and the section L
1, L2, L3, ... Are reference sampling points for each sampling period T of the reference sound segment (for example, reference sampling points A1, A2 shown in FIG. 1).
A3, ...) between the reference sampling points A1 and A2, the section L2 between the reference sampling points A2 and A3, and the section L3 between the reference sampling points A3 and A.
4 sections, such as each section L1, L2, L3.
... correspond to the reference samplings A1, A2, A3 ,.

Further, this inclination parameter table 21a
Is divided into two equal parts, three equal parts, four equal parts, ... Are equal parts information designated by the pitch information input to the output sound element generation unit 2, that is, the reference sampling period T of the reference sound element. It shows how much is divided.

For example, when it is specified that the pitch information is sampled at 6/5 times the sampling period T of the basic sound segment, a short period sampling point obtained by dividing the reference sampling period T into five equal parts is generated. Since the amplitude value is acquired for every 6 short cycle sampling points, the 5 equal parts are used as a key, and the waveform waveform data acquisition cycle is in this case for every 6 short cycle sampling points. Therefore, the sections L1, L2, L3, ...
To obtain the tilt parameter as. The third embodiment will be described below using a specific example.

FIG. 9 shows the waveform of the reference sound segment similar to that shown in FIG. 2, and the amplitude value is obtained for each of the reference sampling points A1, A2, A3, ... Of the reference sampling period T. .

For such a reference sampling segment,
Consider a case where the pitch information is such that sampling is performed at 6/5 times the sampling period T.

When such pitch information is input to the output sound element generation unit 2, the output sound element generation unit 2 acquires the reference sound element as the processing target acquired from the reference sound element storage unit 11. In contrast, the tilt parameter table 21a shown in FIG. 8 held in the pitch conversion data storage unit 21 is referred to. In this case, the equally divided information is 5 evenly divided. On the other hand, the phase control processing unit 3
Receives the phase information of the end portion of the sound element immediately before the reference sound element currently being processed as the end phase information, determines the processing start point in consideration of the consistency with the end phase information. In this case, the processing for determining the processing start point is
This is done as follows.

First, the equal division information of how much the division is made is obtained from the output sound segment generation unit 2, and based on the division information,
By referring to the processing start section amplitude information 21b stored in the pitch conversion data storage unit 21, it is determined which short cycle sampling point is the optimum processing start point.

That is, in this case, since the pitch information input to the output sound segment generator 2 is such that sampling is performed at a period of 6/5 times the reference sampling period T, the reference sampling period T is This means that the waveform data is acquired for every 6 pieces of the short-period sampling points obtained by dividing into 5 equal parts, so that the phase control processing unit 3 divides into 5 equal parts existing in the processing start section S. An amplitude value (processing start section amplitude information 21b) corresponding to a short cycle sampling point is obtained from the pitch conversion data storage unit 21, and processing is started from which short cycle sampling point based on the end phase information of the immediately preceding sound element. It is determined whether or not it is optimal, and it is passed to the output sound segment generator 2.

As a result, the output sound segment generator 2
Then, based on the phase determination information passed from the phase control processing unit 3, that is, information indicating which short-cycle sampling point is the best to start the processing, the waveform data acquisition processing is started from the short-cycle sampling point. To do.

For example, by the phase control processing unit 3,
If it is determined that it is optimal to use the second short-period sampling point a3 as the sampling start point, the output sound segment generator 2 performs sampling (waveform data acquisition) starting from the short-period sampling point a3. After the start, every 6 points from the short cycle sampling point a3, that is,
From the short-period sampling point a3 as a starting point, sampling is performed with short-period sampling points a9, a15, a21, ....

The amplitude value of each of the short cycle sampling points a9, a15, a21, ... Is obtained by acquiring the tilt parameter from the tilt parameter table 21a shown in FIG. 8 each time and using the tilt parameter. It is calculated by performing various calculations. The procedure will be described below. In this case, the processing start section S is assumed to be only the section L1 for convenience of description. Further, since the amplitude value corresponding to the short cycle sampling point existing in this section L1 is obtained in advance, in the case of this example, the amplitude value corresponding to the short cycle sampling point a3 can be obtained thereby.

The short cycle sampling point for which the amplitude value is to be obtained next is the short cycle sampling point a9,
Since the short cycle sampling point a9 exists outside the processing start section S, the short cycle sampling point a9
The amplitude value corresponding to 9 is obtained by referring to the tilt parameter table 21a shown in FIG.

First, in this case, the reference sampling period T
Is divided into 5 equal parts, and the short-period sampling point a9 exists in 3/5 of the section L2, as can be seen from FIG. L2
By referring to the tilt parameter table 21a shown in FIG. 8 using as a key, L2_5 is acquired as the tilt parameter.

When the inclination parameter is obtained in this way, the amplitude value f (a9) at the short-period sampling point a9 is f (a9) = f, where f (A2) is the amplitude value at the reference sampling point A2. It can be calculated by (A2) + (3/5) ・ L2_5 (1).

Next, the amplitude value corresponding to the short cycle sampling point a15 is obtained.
Since 5 exists in 4/5 of the section L3, the inclination parameter table 2 shown in FIG.
Referring to 1a, L3_5 is acquired as the inclination.

If the slope is obtained in this way, the amplitude value f (a15) at the short-period sampling point a15 is f (a15) = f (if the amplitude value at the reference sampling point A3 is f (A3). It can be obtained by A3) + (4/5) ・ L3_5 (2).

Further, the next short cycle sampling point a21
Find the amplitude value of. Since the short-period sampling point a21 coincides with the original reference sampling point A5, the amplitude value obtained in advance as the amplitude value of the reference sampling point A5 can be used as it is in this case.

In this way, the amplitude values corresponding to the short cycle sampling points a3, a9, a15, a21, ... Set by the output sound segment generator 2 can be obtained based on the pitch information.

The third embodiment is characterized in that it is possible to freely set how much the reference sampling period T is divided. This point is different from the first and second embodiments described above. different. That is, in the first and second embodiments, it is fixed how much the reference sampling period T is divided, but in the third embodiment, it can be set arbitrarily, and the short period sampling period T can be set. There is a high degree of freedom in setting, so that more versatile pitch data can be generated.

Further, since the matching in the connection portion with the immediately preceding sound element can be taken into consideration, the sound element parts can be smoothly connected to each other, thereby preventing noise due to a gap or the like. , You can get high quality synthetic sound.

The present invention is not limited to the embodiments described above, and various modifications can be made without departing from the gist of the present invention.

Further, according to the present invention, a processing program in which a processing procedure for realizing the above-described present invention is described is created, and the processing program is recorded in a recording medium such as a floppy disk, an optical disk or a hard disk. The present invention also includes a recording medium in which the processing program is recorded. Further, the processing program may be obtained from the network.

[0099]

As described above, according to the present invention, it is possible to generate sound data of various pitches from an original sound element with a light calculation load. Moreover, the phase control processing that considers the consistency between the waveform data at the end of the sound element used immediately before the sound element currently being processed and the waveform data at each short-cycle sampling point of the sound element currently being processed. When connecting and using several sound elements, you can smoothly connect the connection parts of each sound element, prevent the occurrence of noise, etc., and create a high-quality synthesized sound. Can be generated.

INDUSTRIAL APPLICABILITY The present invention can be applied to, for example, a process of generating a synthesized voice having an arbitrary pitch when synthesizing a text and a process of outputting a sound of an electronic musical instrument at an arbitrary pitch. Pitch conversion can be performed with a small amount of calculation, and high quality synthetic sound can be generated.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of a waveform of a reference sound segment used in an embodiment of the present invention.

FIG. 2 shows a reference sound segment as shown in FIG.
It is a figure which shows the example sampled by the frequency (short sampling period) higher than the reference sampling period T of the reference sound element.

FIG. 3 is a diagram illustrating a specific pitch conversion example according to the first embodiment of this invention.

FIG. 4 is a configuration diagram showing a main part of the sound synthesizing device according to the first embodiment of the present invention.

FIG. 5 is a configuration diagram showing a main part of a sound synthesizing device according to a second embodiment of the present invention.

FIG. 6 is a diagram showing a waveform of a part of a certain reference sound segment in order to explain an example of obtaining an amplitude value of a short-period sampling point in the second embodiment.

FIG. 7 is a configuration diagram schematically showing a main part of a sound synthesizing device according to a third embodiment of the present invention.

FIG. 8 is a diagram showing an example of a tilt parameter table 21a in a certain reference sound segment used in the third embodiment of the invention.

FIG. 9 is a diagram illustrating a specific pitch conversion example according to the third embodiment of this invention.

[Explanation of symbols]

1 Short-cycle sampling sound segment storage 2 Output sound segment generator 3 Phase control processor 11 Reference sound element storage 21 Pitch conversion data storage unit 21a Tilt parameter table 21b Processing start section amplitude information A1, A2, A3, ... Standard sampling points a1, a2, a3, ... Short-period sampling points L1, L2, ... Section T reference sampling period

Claims (16)

[Claims] 1. A plurality of sound pieces for which waveform data for each preset sampling cycle are obtained are used, and the pitch of the sound pieces is converted to obtain sound data of an arbitrary pitch. In the sound synthesis method to be generated, each sound segment is sampled at a short sampling period of 1 / n (n> 1) of the reference sampling period, so that the sampling points (short Waveform data corresponding to (periodic sampling point) is acquired and stored as a short-period sampling sound segment. The short-period sampling sound element is obtained, and the acquired short-period sampling sound element is acquired. On the other hand, the waveform data for each short cycle sampling point corresponding to the waveform data acquisition cycle specified by the input pitch information is acquired, and the acquired waveform data is used as the waveform data for each of the preset reference sampling cycles. A sound synthesis method characterized by outputting. 2. When acquiring waveform data for each short cycle sampling point corresponding to a waveform data acquisition cycle specified by the input pitch information, it is determined from which short cycle sampling point the waveform data acquisition starts. The phase control process is performed for each short-cycle sampling point of the sound element currently being processed and the waveform data at the end of the sound element being used immediately before the sound element currently being processed. The sound synthesizing method according to claim 1, wherein the sound synthesizing method is performed in consideration of consistency with waveform data. 3. A sound data having an arbitrary pitch is obtained by using a plurality of sound elements for which waveform data for each preset sampling cycle has been obtained and converting the pitch of the sound element. In the sound synthesizing method to be generated, sampling points (called short-period sampling points) having a short sampling period of 1 / n (n> 1) of the reference sampling period are generated for a predetermined sound segment, and each of them is generated. Each waveform data corresponding to the short cycle sampling point of is obtained, and it is set as the short cycle sampling sound element data, and the short cycle sampling sound element corresponds to the waveform data acquisition cycle specified by the input pitch information. Obtain waveform data for each short cycle sampling point,
A sound synthesizing method, characterized in that the acquired waveform data is output as waveform data for each of the preset reference sampling periods. 4. The waveform data corresponding to each short period sampling point is obtained by linearly interpolating between the waveform data obtained for each of the reference sampling periods. Sound synthesis method described. 5. When acquiring waveform data for each short cycle sampling point corresponding to a waveform data acquisition cycle designated by the input pitch information, it is determined which short cycle sampling point the waveform data acquisition starts from. The phase control process is performed for each short-cycle sampling point of the sound element currently being processed and the waveform data at the end of the sound element being used immediately before the sound element currently being processed. 5. The sound synthesizing method according to claim 3, wherein the sound synthesizing method is performed in consideration of consistency with waveform data. 6. A sound data having an arbitrary pitch can be obtained by using a plurality of sound pieces for which waveform data for each preset sampling cycle is obtained and converting the pitch of the sound pieces. In the sound synthesizing method to be generated, a sampling point having a short sampling period is set for the sound segment to be pitch-converted by setting the reference sampling period to 1 / n (n> 1) according to the designation of input pitch information. (Short period sampling point) is generated, waveform data of a short period sampling point corresponding to the waveform data acquisition period specified by the input pitch information is obtained, and the obtained waveform data is set to the preset reference sampling. A sound synthesis method characterized by outputting as waveform data for each cycle. 7. The waveform data at a short-period sampling point corresponding to the waveform data acquisition period designated by the pitch information has a value of n when the reference sampling period is 1 / n and the waveform data. A table that can acquire the slope of the sound element waveform according to which section of the reference sampling cycle the short-period sampling point corresponding to the acquisition cycle exists is prepared, and the table is obtained. 7. The sound synthesizing method according to claim 6, wherein the waveform data of the short cycle sampling point is obtained by using the slope. 8. When acquiring waveform data for each short cycle sampling point corresponding to a waveform data acquisition cycle designated by the input pitch information, it is determined from which short cycle sampling point the waveform data acquisition starts. Phase control processing for the sound element being processed, and the phase control processing is specified by the waveform data of the end portion of the sound element used immediately before the sound element currently being processed and the pitch information of the sound element currently being processed. 8. The sound synthesizing method according to claim 6, wherein the sound synthesizing method is performed in consideration of the consistency with the waveform data of the short cycle sampling points. 9. A sound data having an arbitrary pitch is obtained by using a plurality of sound elements for which waveform data for each preset sampling cycle is obtained and converting the pitch of the sound element. In the sound synthesizing device to be generated, each sound segment is sampled at a short sampling period of 1 / n (n> 1) of the reference sampling period, so that sampling points (short The short-period sampling sound segment storage unit that holds the waveform data corresponding to the (periodic sampling point) as a short-period sampling sound segment, and the individual short-period sampling sound units stored in this short-period sampling sound segment storage unit. Short-period sampling sound to be processed at that point Obtain the waveform data for each short-period sampling point that corresponds to the waveform data acquisition period specified by the input pitch information for the acquired short-period sampling sound segment. And an output sound element generator for outputting the waveform data as waveform data for each of the preset reference sampling periods. 10. When the output sound segment generator acquires the waveform data for each short cycle sampling point corresponding to the waveform data acquisition cycle specified by the input pitch information, which short cycle sampling is performed. A phase control processing unit that performs phase control processing for determining whether to start from a point is provided, and the phase control processing performed by the phase control processing unit is performed by the sound element used immediately before the sound element currently being processed. 10. The sound synthesizing apparatus according to claim 9, wherein the sound synthesis is performed in consideration of the consistency between the waveform data at the terminal end and the waveform data at each short-period sampling point of the sound segment currently being processed. 11. A sound data having an arbitrary pitch is obtained by using a plurality of sound elements for which waveform data is obtained for each preset sampling cycle and converting the pitch of the sound element. In a sound synthesizer for generation, sampling points (called short-period sampling points) with a short sampling period of 1 / n (n> 1) of the reference sampling period are generated for a predetermined sound segment, and each of them is generated. Of each waveform data corresponding to the short-period sampling point, and the short-cycle sampling sound segment generator that uses it as the short-period sampling sound segment data, and the input pitch information for the short-period sampling sound segment. Waveform data for each short-period sampling point corresponding to the specified waveform data acquisition period To get the data,
An output sound element generator that outputs the acquired waveform data as waveform data for each of the preset reference sampling periods, a sound synthesizing device. 12. The waveform data corresponding to each of the short period sampling points is obtained by linearly interpolating between the waveform data obtained for each of the reference sampling periods. 11. The sound synthesizer according to item 11. 13. The output sound element generation unit acquires waveform data for each short cycle sampling point corresponding to a waveform data acquisition cycle designated by the input pitch information from the short cycle sampling sound element data generation means. At this time, a phase control processing unit that performs a phase control processing for determining from which short cycle sampling point to start acquiring the waveform data is provided, and the phase control processing performed by this phase control processing unit is the sound currently being processed. 12. This is performed in consideration of the consistency between the waveform data at the end of the sound segment used immediately before the segment and the waveform data at each short cycle sampling point of the sound segment currently being processed.
Or the sound synthesizer according to item 12. 14. A sound data having an arbitrary pitch is obtained by using a plurality of sound pieces for which waveform data for each preset sampling cycle is obtained and converting the pitch of the sound pieces. In the sound synthesizing device for generation, for the sound element to be subjected to the pitch conversion processing, the reference sampling period is set to 1 / n (n> 1) according to the designation of the input pitch information, and the sampling point of the short sampling period (Short period sampling point) is generated, waveform data of a short period sampling point corresponding to the waveform data acquisition period specified by the input pitch information is obtained, and the obtained waveform data is set to the preset reference sampling. It has an output sound segment generator that outputs waveform data for each cycle. Sound synthesis device to. 15. In the waveform data generation process of the short cycle sampling point corresponding to the waveform data acquisition cycle designated by the input pitch information, the output sound segment generation means sets the reference sampling cycle to 1 / n. The slope of the sound element waveform can be acquired depending on the value of n and the short cycle sampling point corresponding to the waveform waveform data acquisition cycle in which section of the reference sampling cycle. 15. The sound synthesizing apparatus according to claim 14, wherein a table that can be created is prepared, and the waveform data at the short-period sampling point is obtained by using the inclination obtained by the table. 16. When acquiring the waveform data for each short cycle sampling point corresponding to the waveform data acquisition cycle designated by the input pitch information, the output sound segment generator determines which short cycle sampling A phase control processing unit that performs phase control processing for determining whether to start from a point is provided, and the phase control processing performed by this phase control processing unit is performed by the sound element used immediately before the sound element currently being processed. 16. The sound according to claim 14 or 15, wherein the matching is carried out in consideration of the consistency between the waveform data of the terminal end portion and the waveform data of the short-period sampling point designated by the pitch information in the sound segment currently being processed. Synthesizer.