JPH09134188A - Singing voice synthesizer and musical tone reproducing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent the degradation in the quality of a back chorus in spite of a change in playing speed and interval by applying a singing voice synthesizer capable of synthesizing the back chorus of PCM quality to a synthesizing KARAOKE machine. SOLUTION: The waveform data read out by a first reading out device 106a and second reading out device 106b are subjected to level conversion by level converters 108a and 108b according to the output values of first and second envelope formers 107a and 107b. Further, the outputs of these level converters 108a and 108b are added by an adder 109 and the singing voice synthesized output is obtd. at an output terminal 110. Further, the device is provided with an interval differential value generator 111 for controlling the pitch values of phonemes so as to gradually approximate the pitch value of the front phoneme to the rear phoneme at the time the front phoneme transfers to the rear phoneme when the pitches of the front and rear phonemes vary. As a result, the phonemes and the phonemes are extremely smoothly connected by controlling the reading out devices 106a and 106b.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a music reproducing apparatus used in a karaoke apparatus for performing a karaoke performance with a back chorus, and particularly, the quality of the back back chorus is deteriorated by changing the tempo (performance speed) and pitch (transposition) of the music. The present invention relates to a singing voice synthesizing device and a music reproducing device.

[0002]

2. Description of the Related Art In recent years, various karaoke systems have been proposed in accordance with the progress of digital technology, and in particular, a synthesizer system karaoke system using a musical tone synthesizer similar to an electronic musical instrument (hereinafter referred to as "synth karaoke system") has attracted attention. Is becoming popular.

Synthesized karaoke is a system in which a musical sound synthesizer is driven by performance instruction information generally called MIDI to generate back music for karaoke.
Compared with general karaoke (for example, laser disk karaoke), the feature is that the amount of data required for playing one song is very small.

This feature of the synthesizer system is utilized in synth karaoke (so-called communication karaoke) in which performance information is transferred through an ISDN or a telephone line as a transmission line, and in synth karaoke in which performance information is stored in a semiconductor memory or the like. Therefore, the growth of synth karaoke in this field is expected in the future.

On the other hand, synthesizer karaoke has a drawback that it is difficult to play voice and back chorus. This is because synth karaoke is configured using a musical tone synthesizer for electronic musical instruments, and the musical tone synthesizer is configured for playing musical instrument sounds (musical instruments other than voices such as piano and guitar). There is.

As a conventional music reproducing apparatus for karaoke using a synthesizer system to overcome such drawbacks, there is one disclosed in Japanese Patent Application Laid-Open No. 6-161479.

This conventional music reproducing apparatus will be described below with reference to the drawings described in the publication. In FIG. 1 of the publication, 1 is a tone data recording unit, 2 is a tone playing unit, 3 is an input unit, 4 is a character data recording unit, 5 is a voice synthesis unit, 6 is a synthesis amplification unit, 7 is an output unit, and 8 is It is a synchronization unit.

The conventional music reproducing device configured as described above will be described below. The musical sound data recording unit 1 is for recording the recording data of a musical composition, and is composed of, for example, a compact disk (CD), a laser disk (LD) or a magnetic recording medium and its drive device. The musical tone playing section 2 receives the musical tone data of the musical tone data recording section 1 and converts it into an analog musical tone signal. Here, if the musical tone data recorded in the recording unit 1 is a MIDI code, the musical tone playing unit 2 may use a synthesizer that generates a sound using an electronic oscillator. The input unit 3 also inputs the voice of the singer using a microphone or the like. Further, the character data recording unit 4 records a character code string, and like the musical tone data recording unit 1, is composed of a magnetic recording medium and a drive device, and can also function as the musical tone data recording unit 1. . The voice synthesizer 5 converts the character data string recorded in the character data recorder 4 into a received voice signal. The synthesis amplification section 6 synthesizes and amplifies the output of the musical tone performance section 2, the voice signal input from the input section 3 and the output of the voice synthesis section 5. The resulting output is sent to the output unit 7 such as a speaker. Further, the synchronizing section 8 gives an instruction to control the character data sent from the character data recording section 4 to the voice synthesizing section 5 in accordance with the performance of the musical tone performance section 2.

A MIDI code is recorded in the music data recording unit 1. FIG. 2 of the publication shows an example of character data recorded in the character data recording unit 4. The operation of this apparatus will be described with reference to FIGS. MIDI
When the chord starts to be played, the reading speed from the tone data recording section 1 set in the tone playing section 2, that is, the value of the tone playing speed is sent to the synchronizing section 8 and the synchronizing section memory 9
Is stored in Next, the data of the first measure of the MIDI code is read from the musical tone data recording section 1 and sent to the musical tone performance section 2 to be converted into a musical tone signal. MID for one bar at the same time
The reception of the I code is notified to the synchronization unit 8. When the synchronizing section 8 knows that the one-measure MIDI code has been sent to the musical tone playing section 2, it reads out one line of data from the character data recording section 4, counts the number of characters, and calculates a value designating a voice speed. To do. Here, the character data for one bar is one line up to the line feed. Then, the voice synthesizing unit 5 sends character code data which is pronounced from the designation of speed. The speed value is calculated as follows. Value of speed = (1 bar time) / (1
Number of characters in line).

For example, the song that is about to be played is 1
Assuming that the bar is set to be played at a speed of 5 seconds and the number of characters in one line of the back chorus is 10, the voice synthesizer 5 outputs the characters sent at the specified speed to the voice signal. To the composite amplification unit 6. As a result, the musical sound generated by the musical sound playing unit 2 and the voice generated by the voice synthesizing unit 5 are synchronized and sent from the synthesizing and amplifying unit 6 to the output unit 7. In this way, the voice sound data is recorded as a character code string, and the voice is generated from the character code string. Therefore, the voice sound data can be coded and recorded similarly to the MIDI data of the musical sound, and the data can be greatly Can be compressed.

However, in the music reproducing device disclosed herein, only the characters are stored in the character data recording unit 4 as shown in FIG. 2, and it is necessary to convert these characters into phonemes. 6 can be instructed, but the pitch cannot be instructed to the speech synthesizer 6, and the pronunciation timings of these phonemes cannot be finely controlled because they are evenly allocated within one measure. However, there was a problem that only a back chorus for special songs could be realized.

As a conventional synth karaoke music reproducing apparatus in which this point is improved, for example, Japanese Patent Application Laid-Open No. 7-140991.
There is a karaoke device disclosed in the publication.

The conventional music reproducing apparatus will be described below with reference to the drawings of the publication.

In FIG. 1 of the publication, 1 is a CPU, 2 is a RAM, 3 is a bus, 4 is a song data storage device, 5 is a panel I / F, 6 is a controller, 7 is a background image storage / reproduction unit, and 8 Is an image / lyrics display section, 9 is a video selector, 10
Is a monitor, 11 is a microphone, 12 is a mixer and an effector, 13 is an amplifier and a speaker, 14 is a sound source, 15
Is a program ROM and sequencer, and 16 is a digital audio decoder.

FIG. 2 of the publication is a structural diagram of music data,
FIG. 3 is an explanatory diagram showing the details of the performance data track,
4 is an explanatory diagram showing details of the audio data instruction track, FIG. 5 is a time-series structural diagram of each track, and FIG.
7 is an explanatory diagram of a portion for decoding audio data, and FIG.
Is a time chart of chorus voice control.

The operation of the conventional karaoke apparatus configured as described above will be described below.

FIG. 1 is a block diagram of an entire conventional karaoke apparatus. In this figure, 1 is a CPU (central processing unit) that controls and manages the operation of the entire system, and 2 is a CPU
A RAM (random access memory) used when controlling and managing the operation of the entire system by 1 is a data and address bus for integrating the entire system.

A storage device 4 stores a plurality of pieces of music data,
For example, an HDD (hard disk), 5 are panel I / Fs (interfaces), 6 are controllers such as a plurality of remote controllers that give instructions to the system via the panel I / Fs 5, 7 is a background image storage / playback unit, and 8 is a background image. An image / lyrics display section for displaying still images and lyrics, and 9 is a video selector for selecting and synthesizing the background video (moving image) from the background video storage / playback section 7 and the image / lyrics display section 8. Is a monitor that displays the image selected and synthesized by the video selector.

Reference numeral 11 is a microphone for inputting the singer's singing voice, 12 is a mixer for mixing the singing voice and the musical sound of the performance song, and an effector for providing various acoustic effects. 13
Is a speaker that amplifies and outputs the synthesized singing voice and performance music. Reference numeral 14 is a sound source for producing a musical tone of a musical composition, and 15 is a sequencer for controlling the sound source 14 and the mixer and effector 12. This sequencer 1
Reference numeral 5 has a program ROM that stores programs used by the CPU 1. A voice decoder 16 decodes encoded digital voice data (for example, PCM or ADPCM data).

The operation will be described below. When the musical composition is designated by the operation of the controller 6, the CPU 1 refers to the musical composition list stored in the storage device 4 and transfers the musical composition data of the corresponding musical composition and the encoded voice data for the back chorus to the RAM 2. Then, the control is transferred to the sequencer 15. The sequencer 515 simultaneously executes a plurality of events including music performance in parallel based on a plurality of event data included in the music data.

That is, in the sequencer 15, the data relating to the tone color of the musical composition and the tone color of the pseudo voice are sent to the sound source 14, and the encoded voice data is sent to the digital voice decoder 16.
In addition, the background video number is in the background video recording / playback unit 7,
The lyrics numbers are supplied to the image / lyrics display section 8, respectively. As a result, the background image is displayed on the entire surface of the monitor 10 and a part of the background image is displayed in a superimposed state, while the speaker 13 is in a playing state in which the playing song and the back chorus are output.

As shown in FIG. 2, the music data includes a header portion,
It consists of a data sequence part and a voice data part. In the header portion, as the information peculiar to the song, a song number, a song name, a composer name, a singer name, background image selection information, and lyric font information are written.

In the data sequence section, a plurality of tracks describing a plurality of types of events that are executed simultaneously in parallel are set. A plurality of voice numbers selected by the voice data instruction data of the data sequence portion are written in the voice data portion.

Among the tracks included in the data sequence section, data for uttering the musical tone of the musical piece to be played from the sound source 14 is described in time series on the performance data track. The pseudo voice data track contains a pseudo chorus voice from the sound source 14.
The data for uttering (for example, "Wow", "Woo") is described in time series. In the voice data instruction track, voice data numbers, pitches, and pitch data for instructing the types of true chorus voices (for example, "Hakodate" and "Nagasaki") to be decoded by the decoder 16 are described. In the lyrics data track, data indicating the type of lyrics to be displayed on the monitor 10 along with the performance is described in time series. In the effect control data track, control data for controlling the mixer and the effector 12 are described in time series.

FIG. 3 is an explanatory diagram showing details of the track played. In the performance data track, note event information, tone color change event information, and pitch bend event information are described. A CH number, a note number (pitch), a velocity (volume), and a note length for designating one CH (channel) to be uttered by the sound source 14 are written in the note event. CH number and tone color data are written in the tone color change event. The CH number and pitch bend information are written in the pitch bend event. The note event of the performance data track is for the musical tone pronunciation of the performance song, but the pseudo voice data track also
It has a note event with a similar structure but with different CH numbers.

FIG. 4 is an explanatory diagram showing details of the audio data instruction track. In this voice data instruction track,
Each information of the voice instruction event is described. That is, it is information about the voice data number, the pitch, and the volume. The voice data number is the number of the encoded true chorus voice data to be decoded by the decoder 16, and is the voice number of the voice data of FIG. As shown in FIG. 5, the plurality of tracks have a structure in which the type of event and the waiting time Δt until the next event is uttered are arranged in time series.

FIG. 6 is an explanatory diagram of a portion for decoding the voice data according to the information of the voice instruction event. The sequencer 15 reads the encoded digital voice data of the corresponding voice data number from the RAM 2 using the voice data number of the voice instruction event and inputs it to the digital voice decoder 16. This digital voice data is true chorus voice data in the conventional example. As an example, when the digital audio data is adaptive delta (AD) PCM data compressed to reduce the data capacity, the decoder 16 is a PCM decoder having a function of performing bit number conversion and frequency conversion to expand. .

A processor 17 for controlling the pitch and the volume is arranged in the subsequent stage of the decoder 16, and after adjusting the pitch and the volume of the analog waveform decoded here based on the pitch and the volume information included in the voice instruction event. , Mixer and effector 12. The playback speed of the intrinsic chorus voice is 1 using this path.
When it is 00% or when the pitch change amount is within a predetermined range. When the reproduction speed of the musical composition is changed to a value other than 100% by the instruction from the controller 6, the processor 17 sets the volume for the output of the decoder 16 to 0 and stops the reproduction of the intrinsic chorus voice. The same applies when the pitch change amount exceeds a predetermined amount of 200 to 300 cents.

The reason why the intrinsic chorus voice is not used when changing the pitch is as follows. That is, the pitch / volume controller 1
In No. 7, if the pitch for the intrinsic chorus voice is changed to a certain extent or more, the quality of the voice deteriorates and it becomes practically unusable. A guideline for "somewhat" is ± 2 to 3 in semitone units.
Sound (200 to 300 cents). Therefore, the tempo
The pseudo chorus voice is used not only when the (playback speed) is changed but also when the pitch (pitch) is changed by a predetermined amount or more.

For the above reasons, the playing speed of the performance music is 10
When it becomes a value other than 0% or when the pitch change amount exceeds a predetermined amount of 200 to 300 cents, the sequencer 51
5 is a sound source 51 according to the data of the pseudo voice data track.
Produce a pseudo-chorus from 4. The pseudo-chorus voice is a voice channel in which the voices “wow” and “woo” are set in the sound source 514 in advance in the same manner as the musical tone, and the voice channel is selected by the pseudo-voice data to be sounded.

FIG. 7 shows how the sequencer 15 controls when the tempo (reproduction speed) changes around 100%. In this example, the events of the audio data track change in order at times t1, t5, and t8. On the other hand, the tempo changes from 1005 to 70% at time t4 at time t4.
70% to 120% at 6 and 120% to 1 at time t7
This is an example of change to 00%.

Since the tempo at time t1 is 100%, at this time, the decoding of the true audio data "Hakodate" is started according to the event of the audio data track, and the volume of the audio data is set to the predetermined amount described in the audio event. Set. At this time, nothing is processed with respect to the pseudo voice data track, and the voice event corresponding to the first "ha" is read and the pseudo voice (for example, "wow") is sounded from the sound source 14. However, in reality, the volume of the pseudo sound is set to 0 so that the speaker 13 cannot output the sound. At the same time, the volume value (the above-mentioned predetermined value) regarding the intrinsic voice data is stored for restoration.

Since the tempo is 100% at time t2, the true chorus voice is continuously output. At this time, the pseudo voice track enters the sounding event of "ko", but the volume of the pseudo voice remains 0. Tempo is 100 even at time t3
%, So the output of the true chorus voice is continued. At this time, the pseudo voice track enters a sounding event of "da", but the volume on the pseudo voice side remains 0.

At time t4, the tempo changes from 100% to 7
Since it decreases to 0%, the true voice is faded out, and instead, the pseudo voice generation channel is faded in to a predetermined value stored in advance. At this time, since there is a possibility that the true voice is being reproduced, a gentle crossfade is performed to avoid interruption.

At time t5, the audio data side enters the next event ("Shinji Kotosa"), but the tempo is 100%.
Since the sound has not been returned to, the pseudo sound (for example, "wa") corresponding to "shi" is uttered from the sound source 14 instead of producing the true sound with the volume set to 0. This pseudo voice is controlled using the CH number, note number (pitch), velocity (volume), and note length described in the event.

At time t6, the tempo changes from 70% to 120%, but since it is not 100%, the pseudo voice corresponding to "n" (for example, "wa") is continuously produced. Since the tempo changes from 70% to 100% at time t7, one condition for returning to true voice is satisfied, but at this stage, since the previous voice data event is being decoded, until the next voice data event starts, Continue to pronounce the pseudo voice.

At time t8, the tempo is 100%, and a new voice data event starts, so that the real voice is faded in at this timing. At this time, since it is the start of the intrinsic voice, there is no problem even if it suddenly fades in. At this time, although there is a sounding portion corresponding to "d" of the pseudo sound, the sound volume of the sounding channel is set to 0 to prohibit sounding. By performing the above-described control by the sequencer 15, it is possible to provide a karaoke performance piece with a chorus part without any time lag even if the reproduction speed changes.

[0038]

However, in the above-mentioned conventional configuration, when the playing speed or the pitch is changed, the back chorus formed by the PCM and the PCM compressed waveform is changed to the pseudo back chorus formed by the tone synthesizer. Since the sound quality is changed, the sound quality is changed, and especially when the playing speed and the pitch are not within the specified values, a pseudo back chorus is generated and the quality of the back chorus is deteriorated. In particular, while PCM and a back chorus formed by PCM compression play lyrics, "Ah",
The pseudo back chorus composed of "wow" is different.

Further, the data amount of the PCM and the back chorus in the PCM compression format is much larger than the performance information generally called MIDI which drives the tone synthesizer.
Because it is (100 times or more), it takes a lot of time to transmit these for each song via a communication line in online karaoke,
A large amount of data storage device is required for a stand-alone synth karaoke. In order to avoid this, there is a problem that the pseudo back chorus must be used even if the playing speed or the pitch is within the specified value.

The present invention solves the above-mentioned conventional problems, and provides a singing voice synthesizer capable of synthesizing a back chorus of PCM quality, and by applying the singing voice synthesizer to a synth karaoke, the performance speed and pitch are changed. Also provides a music reproducing device in which the quality of the back chorus does not deteriorate.

[0041]

In order to solve these problems, the singing voice synthesizing apparatus of the present invention first provides performance information consisting of an input phoneme, pitch, volume and sounding timing.
And a second output sequence for alternately outputting the pitch, and the first and second output sequences of the assigner to instruct the pitch and sounding timing to output the first and second outputs respectively. A pitch difference generator that outputs a pitch change value, and a pitch and a phoneme and a sounding timing are instructed by the first and second output sequences of the assigner, respectively, and the pitch difference generator generates the pitch by the first and second outputs, respectively. First and second readers for which change values are instructed, and first and second envelope formers for generating waveform envelope data in which volume and sound generation timing are instructed by the first and second output sequences of the assigner, respectively. And a first and second waveform memory in which voice waveforms of a plurality of phonemes and pitches are stored in advance and data is read by the first and second readers, respectively. And first and second level converters for level converting the outputs of the first and second readers with the outputs of the first and second envelope formers, respectively, and the first and second level converters. And an adder for adding the outputs of the above.

Further, the music reproducing apparatus of the present invention comprises a performance data memory storing performance information of musical instruments and singing,
A sequencer that outputs performance instruction information while controlling the timing according to the performance information read from the performance data memory and transposition input and performance speed input, a tone synthesizer that synthesizes musical sounds according to the performance instructions of the sequencer, and the sequencer performance A singing voice synthesizer that synthesizes a singing voice according to instructions,
An adder for adding the output of the musical sound synthesizer and the output of the singing voice synthesizer is provided.

[0043]

BEST MODE FOR CARRYING OUT THE INVENTION The invention according to claim 1 of the present invention is an assigner for alternately outputting performance information consisting of an input phoneme, pitch, volume and sounding timing to first and second output sequences. And first and second readers whose pitch, phoneme, and sounding timing are instructed by the first and second output series of the assigner, respectively, and the volume of sound by the first and second output series of the assigner, respectively. And the first and second envelope formers that generate the waveform envelope data instructed by the sounding timing, and the voice waveforms of a plurality of phonemes and pitches are stored in advance and the data is read by the first and second readers, respectively. First is read
And a second waveform memory, first and second level converters for level-converting the outputs of the first and second readers with the outputs of the first and second envelope formers, respectively. An adder for adding the outputs of the first and second level converters,
The voice waveform of the pitch is read out by the reading device alternately for each phoneme, and smoothly output while crossfading.
Except for the cross-fade part, the original sound PCM is unchanged, so that a high quality singing voice can be synthesized.

Further, when there is a pitch difference between the front phoneme and the back phoneme, the pitch of the front phoneme is gradually asymptotically approached to the back phoneme at the time of crossfade, so that the pitch change becomes smooth and natural. You can synthesize different singing voices.

Further, since the above-mentioned singing voice synthesizer and tone synthesizer are incorporated in karaoke to generate a back chorus, even when the pitch to be played is changed, only the read position of the voice waveform is changed, so that the quality is maintained regardless of the playing speed and the pitch. A high back chorus is obtained.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. (Embodiment 1) FIG. 1 is a block diagram showing a configuration of a singing voice synthesizing apparatus according to Embodiment 1 of the present invention. In FIG. 1, 100, 101, 102 and 103 are input terminals,
104 is an assigner, 105a and 105b are waveform memories,
106a and 106b are readers, and 107a and 107b.
Is an envelope former, 108a and 108b are level converters,
109 is an adder, 110 is an output terminal, and 111 is a pitch difference generator.

FIG. 2 shows the reader 1 according to this embodiment.
It is a figure showing operation and composition of 06a and 106b. In FIG. 2B, 701, 702, 703, and 708 are input terminals, 707 is an output terminal, 709 and 706 are adders, 704 is an address generator, and 705 is a read start address generator.

FIG. 3 is an explanatory diagram for explaining the operation of the pitch difference generator 111 according to the present embodiment, FIG. 4 is an explanatory diagram for the data storage state of the waveform memories 105a and 105b according to the present embodiment, and FIG. 5 is for the present embodiment. Of loop reading of the waveform memories 105a and 105b in the form of FIG.
FIG. 6 is an explanatory view of the crossfade processing of singing voice synthesis in this embodiment, and FIG. 7 is an assigner 1 in this embodiment.
It is operation | movement explanatory drawing of 04.

The operation of the singing voice synthesizing apparatus according to the present embodiment configured as described above will be described.

Performance information consisting of phonemes, pitches, volumes and sounding timings of singing voices to be synthesized is given to the assigners 104 from the input terminals 100, 101, 102 and 103, respectively. The assigner 104 alternately outputs the input phoneme, pitch, volume, and performance information consisting of sounding timing to the first and second output sequences.

Further, the first and second of the assigner 104 are
A first reader 106a and a second reader 106b are respectively connected to the output series of the assigner 1
From 04, the pitch, phoneme, and sounding timing are instructed.

The readers 106a and 106b operate as follows. When the phoneme and the pitch are instructed and the start of the pronunciation is instructed, the readers 106a and 106b sequentially read the waveform data from the head address where the instructed phoneme and the pitch are stored in the waveform memory.

In general, the waveform of a voice is the number 1 of the sound generation start part.
00 msec is the transition part, and thereafter, there is a property that a relatively similar waveform repeatedly appears. This repeated part is also number 1
It is known that it can be expressed in 00 msec. That is, assuming that the waveform of the original voice is as shown in FIG. 5 (a), the sound generation start portion + repetition portion is stored in the waveform memory, and when reading out, the repetition portion is looped as shown in FIG. 5 (b). By reading out, the waveform memory amount can be set to about 500 msec to 1 sec per phoneme.

In this way, the waveform memories 105a and 105b
Has a configuration as shown in FIG. A phoneme is designated from the input terminal 703, and a pitch is designated from 708. The read start address is output from the read start address generator 705 according to the combination of the phoneme and the pitch that has been played.

At the same time, the tone generation timing is given from the input terminal 701 and the pitch difference value is given from 702. Adder 709
Outputs the pitch difference value and the specified value of the pitch, and outputs this as ΔA. ΔA is given to the address generator 704, and the address generator 704 basically integrates ΔA. However, every time the integrated value exceeds the waveform data length, a value corresponding to the length of the repeating portion is subtracted from the integrated value to generate an address value as shown in FIG. This address value is added to the output of the read start address generator 705 and the adder 706, and the address given to the waveform memory is output to the output terminal 707.

Further, the pitch difference generator 111 is instructed of the pitch and tone generation timing from both the first and second output series of the assigner 104. The pitch difference generator 111 temporally changes the pitch of the preceding phoneme to gradually approach the pitch of the following phoneme when transitioning from the preceding phoneme to the following phoneme as shown in FIG. The second correction values are respectively the first and second outputs, and these outputs are respectively the first reader 106a and the second reader 106b.
It is given as the pitch change value input of.

Further, the first and second of the assigner 104 are
The first output sequence for generating the waveform envelope data
Envelope former 107a and second envelope former 10
7b is connected, and the volume and sound generation timing are instructed from the first and second output series of the assigner 104, respectively.

The first waveform memory 1 read by the first reader 106a and the second reader 106b
Reference numerals 05a and 105b store waveform data of all phonemes and pitches that are normally soundable or necessary for performance in Japanese, English and other languages as shown in FIG. Reader 106b sequentially reads the waveform data of the specified pitch and phoneme.

The waveform data read by the first reader 106a and the second reader 106b are the first and second envelope formers 107a and 107b, respectively.
The level is converted by the level converters 108a and 108b in accordance with the output value of.

Further, the level converters 108a and 108b.
Is added by the adder 109 and a singing voice synthesis output is obtained at the output terminal 110.

As an example of the above operation, for example, FIG.
The case of synthesizing the singing voice shown in the score will be described.

In the case of this example, the input of the assigner 104 and the outputs of the first and second output series are as shown in FIG.

The output of the assigner 104 is given to the first reader 106a and the second reader 106b,
The first reader 106a displays the phoneme from the time t1.
And the phoneme (pitch pitch) is read from time t3.
Similarly, the second reader 106b starts phonologically from time t2.
(Pitch L) is read out, and the phoneme (Pitch S) is read out from time t4.

The output of the assigner 104 is applied to the first envelope former 107a and the second envelope former 107b to generate the waveform envelopes shown as Ea and Eb in FIG. 6, respectively. That is, a waveform envelope is generated so as to crossfade both phoneme waveforms at the alternate portions of the phonemes and the phonemes that come and go.

The waveform envelope E formed in this way
a and Eb, the output Sa of the reader 106a, and the output Sb of the reader 106b are the level converters 108a and 108b, respectively.
108b.

The envelope formers 107a and 107b calculate Ea × Sa and Ea × Sb, respectively, and the envelope former 1
The outputs of 07a and 107b are added by the adder 109, the singing voice synthesis process is completed, and an output waveform is obtained at the output terminal 110.

At the same time, the pitch difference generator 111 operates as follows. Upon receiving the output of the assigner 104, the pitch difference generator 111 generates the first and second correction outputs as shown in FIG. For example, when a phoneme with a phoneme (pitch ra) is started to be produced at time t2, a phoneme with a phoneme a (pitch do) is being pronounced immediately before, and there is a pitch difference of 900 cents between them. To reduce the pitch difference gradually, the pitch difference generator 111
Gives the first correction output to the first reader 106a as shown in FIG.

As described above, the readers 106a, 1a
Since 06b reflects the value obtained by adding the specified pitch value and the correction value to the reading speed of the waveform data, the phoneme changes from the pitch do to the pitch la gradually. In this way, the pitch of the immediately preceding phoneme is smoothly approached to the pitch of the next phoneme, and by the synergistic action with the crossfade processing described above, the phonemes and the phonemes are connected very smoothly, and the quality is good. Singing voice can be synthesized.

As described above, according to the present embodiment, the waveform memories 105a and 105b in which the waveform data of a plurality of pitches and phonemes are stored in advance are provided in the two waveform readers 106.
a and 106b are alternately read, and envelope generators 107a and 107b are provided to generate envelopes so as to cross-fade the read outputs, whereby the outputs of the waveform readers 106a and 106b are level-converted and added, and then forward and backward. When the pitches of the phonemes are different, a pitch difference value generator 111 is provided to control the pitch value of the previous phoneme to gradually approach the pitch value of the back phoneme when transitioning from the previous phoneme to the back phoneme. This allows the reader 10
By controlling 6a and 106b, the phonemes can be connected to each other very smoothly, and a high quality singing voice can be synthesized.

In the present embodiment, the waveform memory 10
Although 5a and 105b, readers 106a and 106b, envelope formers 107a and 107b, and level converters 108a and 108b are used, it goes without saying that these can be configured by a single hardware as time division multiplexing processing.

Further, although the level conversion is multiplication, for example, bit shift or the like may be used. (Second Embodiment) Next, a second embodiment of the present invention will be described with reference to the drawings.

FIG. 8 is a block diagram showing the structure of the music reproducing apparatus according to the embodiment of the present invention. In FIG.
200 is a performance data memory, 201 is a sequencer, 20
Reference numeral 2 is a singing voice synthesizer, which is the same as that described in the first embodiment. Reference numeral 203 is a musical sound synthesizer, 204 is an adder, 205 is an output terminal, 206 is a transposition input terminal, 2
Reference numeral 07 is a performance speed input terminal.

The operation of the music reproducing apparatus of the present embodiment configured as above will be described below.

Performance information is stored in the performance data memory 200. The storage format of the performance information is as described in the conventional example, and will not be described repeatedly here. The sequencer 201 reads the performance information from the performance data memory 200, and in the case of the performance of a singing voice, provides the singing voice synthesizer 202 with the performance information including the phoneme of the singing voice to be synthesized, the pitch, the volume, and the sounding timing. If there is, the musical tone synthesizer 203 is given pronunciation information consisting of the pitch, the volume, and the pronunciation timing.

Singing voice synthesizer 202 and tone synthesizer 203
Is added by the adder 204, and the singing voice and the musical instrument playing sound are simultaneously obtained at the output terminal 205.

Reference numeral 206 denotes a transposition instruction value input terminal. When, for example, a value of 1 is input, the sequencer 201 raises the pitch instruction by a semitone and gives it to the singing voice synthesizer 202 and the musical tone synthesizer 203. Reference numeral 207 is an input terminal for a playing speed instruction value, and the sequencer makes the playing speed faster or slower depending on the value given here.

In this embodiment, since the singing voice synthesizer 202 having the same structure as that of the first embodiment is used, when a transposition instruction is given from the input terminal 206 and the sequencer 201 changes the pitch instruction value, the waveform is changed. Memories 105a, 105b
The address to read is changed.

However, since this does not change the quality of the waveform data to be read, the singing voice synthesizer does not deteriorate regardless of transposing / not transposing.

Similarly, even if the performance speed instruction value changes, the reading units 106a and 106b and the envelope forming units 107a and 1
Since the timing information given to 07b and the like only changes and does not change the quality of the waveform data to be read as in the case of transposing, the singing voice synthesizing device does not deteriorate regardless of transposing.

As described above, the performance data memory 200 storing the performance information, the sequencer 201, the singing voice synthesizer 202 according to the first embodiment, and the musical sound synthesizer 2
By providing 03, it is possible to play music without degrading the quality of the singing voice even if the playing speed or the pitch are changed, and even if a key console or the like is operated during music playback such as karaoke, the quality of the music can be improved. Does not deteriorate.

[0081]

As described above, according to the present invention, the performance information consisting of the phoneme, the pitch, the volume, and the sounding timing is first set.
And a second output sequence for alternately outputting the pitch, and the first and second output sequences of the assigner to instruct the pitch and sounding timing to output the first and second outputs respectively. A pitch difference generator that outputs a pitch change value and a pitch and a phoneme and a pronunciation timing are respectively instructed by the first and second output sequences of the assigner, and the pitch is generated by the first and second outputs of the pitch difference generator, respectively. First and second readers for which change values are instructed, and first and second envelope formers for generating waveform envelope data in which volume and sound generation timing are instructed by the first and second output sequences of the assigner, respectively. And a first and second waveform memory in which voice waveforms of a plurality of phonemes and pitches are stored in advance and data is read by the first and second readers, respectively. And first and second level converters for level converting the outputs of the first and second readers with the outputs of the first and second envelope formers, respectively, and the first and second level converters. By adding an adder for adding the outputs of the above, the phoneme waveform of the designated phoneme / pitch is alternately read out by the reader for each phoneme alternately and smoothly output while crossfading. Except for the crossfade part, the original PCM is the same as it is, so a high quality singing voice can be synthesized, and if there is a pitch difference between the front phoneme and the back phoneme, the pitch of the front phoneme is gradually changed to the back pitch during the crossfade. Since the pitch is made asymptotic, the pitch change becomes smooth and a natural singing voice can be synthesized.

Also, performance data memory storing performance information of musical instruments and singing, performance information read from the performance data memory, and performance instruction information is output while controlling timing according to transposition input and performance speed input. A sequencer, a tone synthesizer that synthesizes musical tones according to the performance instructions of the sequencer, a singing voice synthesizer that synthesizes singing voices according to the performance instructions of the sequencer, and an adder that adds the output of the tone synthesizer and the output of the singing voice synthesizer. By having
Even when the pitch or speed of the performance is changed, high-quality music with a singing voice can be reproduced.By applying this to a music playback device for synth karaoke, the quality of the back chorus can be improved even if the performance speed or pitch changes. It is possible to provide a synth karaoke that does not deteriorate and has an excellent effect in practical use.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the configuration of a singing voice synthesizing device according to a first embodiment of the present invention.

FIG. 2 is an explanatory diagram showing an operation and a configuration of a reader according to the same embodiment.

FIG. 3 is an explanatory diagram illustrating an operation of the pitch difference generator according to the same embodiment.

FIG. 4 is an explanatory diagram of a data storage state of the waveform memory in the same embodiment.

FIG. 5 is an explanatory diagram of loop reading of the waveform memory according to the same embodiment.

FIG. 6 is an explanatory diagram of a crossfade process for singing voice synthesis in the same embodiment.

FIG. 7 is an operation explanatory view of the assigner according to the same embodiment.

FIG. 8 is a block diagram showing a configuration of a music reproducing device according to a second embodiment of the present invention.

[Description of Reference Signs] 104 Assigner 105a, 105b Waveform memory 106a, 106b Reader 107a, 107b Envelope generator 108a, 108b Level converter 109 Adder 111 Pitch difference generator 200 Performance data memory 201 Sequencer 202 Singing voice synthesizer 203 Musical sound synthesis Device 204 Adder

Claims (4)

[Claims] 1. An assigner which alternately outputs performance information consisting of an input phoneme, pitch, volume and sounding timing to first and second output series, and first and second output series of the assigner. First and second readers for which a pitch, a phoneme, and a sounding timing are respectively instructed, and a volume and a sounding timing are instructed by the first and second output sequences of the assigner, respectively, and first waveform envelope data is generated. And a second envelope former, first and second waveform memories for preliminarily storing speech waveforms of a plurality of phonemes and pitches, from which data is read by the first and second readers, respectively. First and second level converters for level converting the outputs of the first and second readers with the outputs of the first and second envelope formers, respectively; A singing voice synthesizing apparatus comprising: an adder that adds outputs of the first and second level converters. 2. An assigner which alternately outputs performance information consisting of an input phoneme, pitch, volume and sounding timing to a first and a second output sequence, and a first and a second output sequence of the assigner. The pitch difference generator that outputs the first and second pitch change values to the first and second outputs, respectively, in response to the pitch and the sounding timing, and the first and second output sequences of the assigner, respectively. First and second readers which are instructed of pitch, phoneme and pronunciation timing, and instructed by the first and second outputs of the interval difference generator, respectively, and first and second of the assigner. Volumes and sounding timings are respectively instructed by two output sequences, first and second envelope formers for generating waveform envelope data, and speech waves having a plurality of phonemes and pitches in advance. A first and a second waveform memory for storing a shape and data being read by the first and second readers respectively; and outputs of the first and second readers respectively for the first and second readers. 1. A singing voice synthesizing device comprising: first and second level converters for level conversion with the output of the envelope former; and adder for adding outputs of the first and second level converters. 3. Performance data memory storing performance information of musical instruments and singing, performance information read from the performance data memory and performance instruction information while controlling timing according to transposition input and performance speed input. A sequencer for outputting, a musical tone synthesizer for synthesizing musical tones according to performance instructions of the sequencer, a singing voice synthesizer for synthesizing singing voices according to performance instructions of the sequencer, an output of the musical tone synthesizer and an output of the singing voice synthesizer are added. Music playback device equipped with an adder that 4. The music reproducing apparatus according to claim 3, wherein the singing voice synthesizing apparatus has the configuration according to claim 1 or 2.