JP3104350B2 - Transcription device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a musical notation apparatus for creating note data based on event data indicating the start and end of sounding of each melody sound constituting an input melody.

[0002]

2. Description of the Related Art If a melody reproduced as an audio signal or the like can be converted into a musical note, not only can the content of the melody be recorded, but also the content of the melody can be reproduced as a musical score on a display or an output sheet. Therefore, such a music transcription device is useful as an auxiliary device for music composition activities and the like.

When a plurality of musical tones are reproduced at the same time, it is difficult to convert them into notes. However, when a single tone is reproduced like a melody, an audio signal or the like from which the melody is reproduced is used. It is possible to convert the melody into notes while extracting the pitch from the melody.

[0004] In performing note conversion, it is necessary to align the reproduction timing of each melody sound with a unit of note length that can be expressed in a musical score. Such processing is
Generally called quantize.

For example, when the minimum note unit on a musical score is a sixteenth note, it is necessary to quantize the reproduction start and end timings of each melody sound to a reference time unit defined by the sixteenth note. .

As a conventional quantization method, a method in which pronunciation information (event) corresponding to a melody detected from an audio signal or the like is simply recorded in a time unit corresponding to, for example, a sixteenth note can be considered.

[0007]

However, even when a single sound is reproduced like a melody, a case where two melody sounds overlap for a short time, such as legato reproduction, may occur. In such a case, the note-off event and the note-on event of the two melody sounds can occur within a time shorter than a time unit corresponding to, for example, a sixteenth note. Can't properly record consecutive events,
There is a problem that two melody sounds cannot be properly quantized into two notes.

An object of the present invention is to appropriately process two melodies and convert them into notes even when input melody sounds overlap.

[0009]

The present invention is premised on having the following note data generating means. Such a note data generating means is based on the invention disclosed by the applicant of the present application in a patent application of Japanese Patent Application No. 2-123789.

[0010] That is, the note data generating means preliminarily sets R
Event data indicating the occurrence of a sound start or sound end event of a musical tone which is stored in the AM or the like and sequentially input thereto, pitch data indicating the pitch of the musical tone sequentially input;
And a minimum measurement of a valid sounding section and a mute section other than a valid sounding section based on time data indicating the time between occurrence timings of the sequentially input event data in units of a predetermined minimum measuring unit time. It is determined sequentially on a time axis with unit time as a unit.

Next, the note data generating means includes note data representing a note having a pitch corresponding to the above-mentioned effective sounding section and having a pitch based on the pitch data, and a note length corresponding to the silence section. And rest data representing the rest of the rest.

Under the above-mentioned premise, the note data generating means according to the present invention has the following configuration. That is, first, the note data generation means determines a sound start position and a sound end position on a time axis in units of the minimum measurement unit time based on each set of event data, pitch data and time data which are sequentially input. Determined sequentially.

Next, the note data generating means determines whether or not there is another sounding start position between each sounding start position sequentially determined as described above and the corresponding sounding end position. When both the first time length from the position to the corresponding sound generation end position and the second time length from the other sound generation start position to the corresponding sound generation end position are longer than 1/2 of the predetermined minimum note length, Determines sequentially from each tone generation start position to another tone generation start position as a tone generation section corresponding to each tone generation start position.

On the other hand, the note data generating means is provided with the first
If either one of the time length of the second sound and the second time length is shorter than 1/2 of the predetermined minimum note length, the sound generation from each sounding start position to the sounding end position corresponding to another sounding start position is performed. The sound generation section is sequentially determined as the sounding section corresponding to the start position, and the pitch data corresponding to the time length shorter than 1/2 of the predetermined minimum note length is rewritten with the content of the pitch data corresponding to the other time length.

The note data generating means determines the sounding section as an effective sounding section when the time length of the sounding section sequentially determined in this way is longer than 1/2 of a predetermined minimum note length.

After the above operation, the note data generating means:
The note data is generated as follows. That is, first, the note data generation means determines that the sounding start position of the valid sounding section determined as described above is the minimum note including the sounding start position on the time axis in units of a predetermined minimum note length. If it is included in the first half of the length range, it generates note data with the start position of the minimum note length range as the note start position of the note interval, and if it is included in the latter half, it generates the next minimum note length range Note data is generated with the start position of the range of note length as the sounding start position of the note section.

Next, the note data generating means determines that the end position of the valid sounding section determined as described above is the minimum note including the end position on the time axis in units of a predetermined minimum note length. If it is included in the first half of the length range, generate note data with the start position of the minimum note length range as the end position of the note section, and if it is included in the second half, the next minimum note following the minimum note length range Note data is generated with the start position of the length range being the end position of the note section.

[0018]

The note data generating means firstly generates a sound-producing section effective for converting into a note and a silence section other than the valid sound-producing section based on data indicating the performance contents of, for example, a melody sound stored in a RAM or the like in advance. They are sequentially determined by a predetermined operation algorithm.

First, a sounding start position and a sounding ending position are sequentially determined on a time axis in units of the minimum measurement unit time. Then, when there is another sounding start position between each sounding start position and the corresponding sounding end position, the first time length from each sounding start position to the corresponding sounding end position and another sounding start position If the second time length from the sounding start position to the corresponding sounding ending position is longer than 所 定 of the predetermined minimum note length, each sounding start position from each sounding start position to another sounding start position is used. Are sequentially determined as sounding sections corresponding to. As a result, even when the two melody sounds overlap for a short time as in legato performance, the two melody sounds can be appropriately separated into two sounding sections.

If one of the first time length and the second time length is shorter than 1/2 of the predetermined minimum note length, the sound generation start position is changed to another sound generation start position. Up to the corresponding sounding end position is sequentially determined as a sounding section corresponding to each sounding start position, and pitch data corresponding to a time length shorter than 1/2 of a predetermined minimum note length corresponds to the other time length. Rewritten according to the contents of the pitch data. As described above, when two melody sounds are overlapped and one melody sound has a short sounding time, unnatural note data can be generated by absorbing the section to the longer melody sound. It can be prevented from being generated.

Next, the note data generating means, if the time length of the sequentially determined sounding section is longer than a half of the predetermined minimum note length, converts the sounding section into an effective sounding section. To be determined. By determining an effective sounding section in this way, it is possible to remove in advance a case where the time length of the sounding section is extremely shorter than a predetermined minimum note length and cannot be converted into a note.

Further, the note data generating means determines the valid sounding section and the mute section for the time being as described above, and then determines the sound duration corresponding to each valid sounding section.
Further, note data representing a note of a pitch based on the pitch data and rest data representing a rest of a pitch corresponding to each silence section are sequentially generated according to a predetermined operation algorithm. In this case, the correspondence from each effective sounding section or mute section to each note data or rest data is time-wise so that each note or rest is configured in units of a predetermined minimum note length. This is performed while the correction is being made. As a result, the melody sound played along the time axis having the minimum measurement unit time as a unit can be appropriately associated with a note or a rest.

[0023]

An embodiment of the present invention will be described below with reference to the drawings. Diagram 1 of an embodiment of music transcription apparatus is a block diagram showing a configuration of an embodiment of a music transcription apparatus according to the present invention.

When the user presses a start switch for instructing the start of melody data generation (not shown) in the switch section 11, the following processing is executed.

First, the frequency band of an audio signal input by a user from an audio reproducing device (not shown) is limited by the anti-aliasing filter 1 to a band of サ ン プ リ ン グ of the sampling frequency, and then A / D converted. The signal is converted into a digital signal by the device 2.

Next, a DSP (Digital Signal Processor) 3
And a CPU (Central Processing Unit) 6 detects a pitch (pitch) of the digitized audio signal in real time, and based on the pitch, performs a note-on or a note-on, which indicates the start of each melody sound constituting the melody. Event information such as note-off indicating the end of sound generation and a note number indicating the pitch of the melody sound are generated, and are stored in the RAM 7 as melody data together with time data indicating the input timing of each event. Here, the filter coefficient ROM 4 stores filter coefficients for digital filtering operation executed in the DSP 3, and the work area RAM 5 stores various variables used in the digital filtering operation.

Finally, when the user presses a stop switch for instructing the end of melody data generation (not shown) in the switch section 11, the melody data generation processing described above is completed.

Thereafter, when the user presses a start switch for instructing the generation of note data (not shown) in the switch section 11, the CPU 6 executes the above-mentioned RA.
A process of converting the melody data stored in M7 into note data that can be expressed as a musical score is executed, and the generated note data is stored in the RAM 7.

The musical note data stored in the RAM 7 is displayed as a musical score on a display device (not shown). In response to the generation of the melody sound event data by the CPU 6, the tone generation circuit 8 generates a melody tone signal corresponding to the event data. The generated tone signal is emitted as a tone from the speaker 10 via the amplifier 9.

The operation of this embodiment having the above-described configuration will be sequentially described below. Principle of Pitch Detection in the Present Embodiment In the present embodiment, the DSP 3 executes N types of bandpass filtering processes on a digitized audio signal by passing only components near a predetermined frequency in a time-division manner. I do. Further, the DSP 3 executes signal processing for extracting an envelope signal from each of the N types of output signals obtained as a result of the above-mentioned N types of band-pass filtering processing in a time-division manner, and converts the resulting N types of envelope signals into the CPU 6. Output to Here, each transfer function of the N kinds of bandpass filtering processes is set to have a characteristic such that only components near predetermined frequencies corresponding to N pitches are passed.

As a result, when the CPU 6 detects one of the N types of envelope signals whose amplitude level exceeds a predetermined threshold value, the CPU 6 determines the pitch corresponding to the detected envelope signal as the pitch of the audio signal. The high or note number can be determined. The filtering process by the DSP 3 will be described below.

In the present embodiment, the DSP 3 performs the above-described N kinds of band-pass filtering processing, a high-pass filtering processing of one fixed characteristic, and N kinds of resonance-added resonances each having a peak at a predetermined frequency. The low-pass filtering processing is realized as signal processing sequentially executed as time-division cascade processing. The DSP 3 performs the signal processing for extracting the envelope signal as N types of low-pass filtering processes having cutoff frequencies sufficiently lower than the cutoff frequencies in the lowpass filtering processes corresponding to the envelope signals. Realize.

It should be noted that the filtering algorithm described below is based on Japanese Patent Application No.
In the following description, only the outline processing will be described, since it is in accordance with the embodiment disclosed in the above-mentioned patent application.

First, when the user presses a start switch for instructing the start of generation of melody data (not shown) in the switch section 11, a start instruction of a pitch detection process is output from the CPU 6 to the DSP 3, and As a result, the DSP 3 starts the processing shown in the operation flowchart of FIG. 2 for the filtering processing. Note that this operation flowchart is realized as an operation in which the DSP 3 executes a microprogram stored in a ROM (not shown).

First, all contents such as the work area RAM 5 are initialized (step S201). Next, in step S202, it is waited for the A / D converter 2 to complete the A / D conversion process for one sample,
When the D conversion processing ends, in step S203,
One sample of the audio signal converted into digital data is taken into the work area RAM 5.

Subsequently, in step S204, a high-pass filtering process having a fixed transfer characteristic is executed based on the A / D converted audio signal. Each filter coefficient used for this filtering process is read from the filter coefficient ROM 4. Then, an output obtained as a result of the filtering process is stored in the work area RAM 5.

The above-described high-pass filtering is a process common to each band, and thus is executed only once for one sample of the audio signal. Next, an IIR low-pass filtering process for implementing a band-pass filtering process and a low-pass filtering process for extracting an envelope are successively executed for each pitch.

Here, a register t for repetition control for performing time division processing for N pitches is provided in the DSP 3, and the value of the register t is initialized to 1 in step S205. Then, in steps S206 and S207, I for realizing the band-pass filtering process for one pitch.
Each time the IR low-pass filtering process and the low-pass filtering process for envelope extraction are completed, it is determined in step S209 whether or not the content of the register t has reached N. If not, the content of the register t is determined in step S210 Is incremented, and the processing of steps S206 and S207 is repeated.

In this case, each time the IIR low-pass filtering process for realizing the band-pass filtering process and the low-pass filtering process for extracting the envelope corresponding to each pitch are performed, the filter coefficient ROM 4 A filter coefficient group corresponding to the transfer function of the filtering process is read out, and each filtering operation is executed using the filter coefficient group.

As a repetition of the above-described steps S206 to S209, an IIR low-pass filtering section for realizing band-pass filtering processing and a low-pass filtering processing for envelope extraction are executed for N pitches. An envelope signal of N pitches corresponding to the sample of the audio signal is obtained, and is transferred to the CPU 6 together with each pitch number (1 to N) in step S211.

The above steps S203 to S211 are repeated every time the A / D converter 2 determines that the A / D conversion has been completed in the above-mentioned step S202. Therefore, an envelope signal for N pitches is obtained every sampling cycle of A / D conversion. Writing process of melody data Next, CPU 6 is, the melody data corresponding to the melody represented by the audio signal input, DS
An operation of generating the envelope signal based on the above-described envelope signal output from P3 and writing the generated envelope signal to the RAM 7 will be described.

First, FIG. 3 shows that the melody data is stored in the RAM 7.
FIG. 6 is a diagram showing a data format when data is written to a. First, as shown in FIG. 3A, a 2-byte melody composed of LEN [i] and CD [i] is stored in the RAM 7 for each event (note on or note off of the melody sound). Data is sequentially stored (i = 1,
2, ...). Here, LEN [i] (i = 1, 2,...)
..) represents the elapsed time from the previous event to the current event. CD [i] indicates a command, and usually includes a flag indicating a note-on or note-off event and a note number of the event.
At the end of the melody reproduction and at the time of time-out described later, it is composed of codes indicating respective states.

FIG. 3B is a diagram showing the contents of CD [i] when a note-on or note-off event occurs. As shown in the figure, the lower five bits store a note number indicating one of the 32 types of keys of the keyboard 1. In the sixth bit, “0” is stored unconditionally. The seventh bit stores the number (type) of the melody input device. Normally, a melody is input via an audio signal, and a melody is detected by detecting a pitch (pitch) from the signal, so that this value is “1”. 8th bit "0" or "1"
Are flags indicating note-on or note-off, respectively.

FIG. 3 (c) shows the RA at the end of the melody reproduction.
Represents the contents of CD [i] (END command) written to M7, and "FF" (hexadecimal value) is written. FIG. 3D shows a CD [i] written to the RAM 7 when a time-over occurs between the writing of the previous melody data and the writing of the next melody data.
(TIME OVER command).
E "(hexadecimal value) is written. The time-over command will be described later.

The operation of the CPU 6 when the melody data having the above-described data format is written into the RAM 7 will be described with reference to the operation flowchart of FIG. This flow is executed by the CPU 6 in a ROM (not shown).
It is realized as an operation of executing a control program stored in the storage device.

First, in step S401, as initialization processing, the contents of the storage area of the melody data in the RAM 7 and the duration of the detected melody sound (the time from note-on to note-off) are counted in the CPU 6. The counter LENGTH, which is a register provided in, and the pointer i indicating the storage position of the melody data in the RAM 7 are cleared to the value “0”.

Next, the repetition processing of steps S402 to S404 is a repetition processing for setting a timing for writing the melody sound with reference to a predetermined unit time.

That is, in step S402, the DSP
3, the note-on (O) of the melody sound having any pitch based on the envelope signal of N pitches output from N.
It is detected whether N) or note-off (OFF) has occurred. Here, for example, note-on or note-off is detected by the following algorithm.

First, of the envelope signals for N pitches output from the DSP 3, two envelope signals are extracted in the order of increasing amplitude value (absolute value). next,
A note number equal to the pitch corresponding to the envelope signal whose amplitude value satisfies a predetermined condition among the envelope signals is stored as the note number at which the sound of the melody sound is started. Then, the occurrence of note-on or note-off is detected by comparing the stored note number with the note number of the current note-on. That is, if the newly stored note number is not currently note-on, occurrence of a new note-on is detected. If the currently stored note number does not include the currently note-on note number, the occurrence of a new note-off is detected.

After the above note-on or note-off detection processing, in step S403, it is determined whether or not a STOP key (not shown) for stopping the recording of the melody data in the switch section 11 has been pressed by the user. Is determined. If this determination is NO, step S404
Is executed. The processing in steps S418 and S419 when this determination is YES will be described later.

In step S404, a hardware timer (not shown) determines whether the minimum measurement unit time has elapsed. If the determination is NO, the process returns to step S402. Now, the minimum measurement unit time is, for example, 1 / one of a whole note.
This is a time corresponding to 96, and the next control processing is not performed until the above-described minimum measurement unit time has elapsed since the last control processing related to the writing of the melody sound. That is, the control process for writing the melody sound is always performed in units of the minimum measurement unit time. Note that the minimum measurement unit time is a relative time corresponding to, for example, 1/96 of a whole note, as described above. If the user adjusts the tempo in a hard timer (not shown), the time length may be different. come.

If the determination in step S404 is YES, in the next step S405, the aforementioned step S4
02, a new note-on (ON) or a new note-off (OF)
It is determined whether or not F) has occurred.

If there is no new note-on or note-off and the determination in step S405 is NO, step S405
413 (described later), the content of the counter LENGTH is incremented in step S414,
Thereafter, the process returns to step S402 described above. Here, the content of the counter LENGTH is cleared in step S411 described later every time a new note-on or note-off event occurs and the information is written to the RAM 7. Therefore, the content of the counter LENGTH is
This represents the elapsed time in units of the minimum measurement unit time since the last event occurred.

Then, a new note-on or note-off event is detected, and the determination in step S405 is YES.
In step S406, the value of the counter LENGTH representing the elapsed time from the previous event is set to the time data LEN [of the first byte of the latest melody data in the melody data storage area indicated by the pointer i. i] (see FIG. 3A)
7 is written.

Next, the following data are written in the RAM 7 as a command CD [i] (see FIG. 3A) to be stored in combination with the time data LEN [i].
That is, first, in step S407, it is determined whether the currently detected event is note-on or note-off, and according to the determination result, step S407 is performed.
In 408 or S409, 8 of the command CD [i]
Either flag OFF = “1” indicating note-off or flag ON = “0” indicating note-on is written in the bit. Subsequently, in step S410, the command CD
In the lower 5 bits of [i], the note number of the currently detected event (note on or note off) is written.

Thereafter, the content of the counter LENGTH is cleared to "0" in step S411, the value of the pointer i is incremented in step S412, and the process returns to step S402.

Here, since the time data LEN [i] is 1-byte data, it can express only a value up to a unit of 255 hours. Therefore, during the above-mentioned melody data write operation, the counter L is set in step S413.
If it is determined that the value of ENGTH has reached “255”, a time-over command is written in the melody data storage area to temporarily stop the time measurement for the event.

That is, first, in step S415, the maximum value “255” in the time unit is set to the time data LEN of the first byte of the latest melody data in the melody data storage area pointed to by the pointer i.
It is written to the RAM 7 as [i].

Next, in step S416, the command C to be stored in combination with the time data LEN [i]
A code “FE” indicating time over is set as D [i].
(See FIG. 3D) is written.

Thereafter, the content of the counter LENGTH is cleared to "0" in step S417, the value of the pointer i is incremented in step S412, and the process returns to step S402.

By repeating the above processing, the melody data is sequentially written into the RAM 7 every time the melody sound is turned on or off. When the user presses a STOP key for stopping recording of melody data (not shown), the determination in step S403 becomes YES.

In this case, after the value of the counter LENGTH is written to the RAM 7 as time data LEN [i] of the first byte of the last melody data in the melody data storage area pointed by the pointer i in step S418. In the subsequent step S419, a code "FF" (see FIG. 3C) indicating the end of the write operation is written as the command CD [i] to be stored in combination with the time data LEN [i]. Note algorithm then melody data stored in the melody data storage area of RAM7 by the above-described operation, the algorithm of the processing operation of converting the musical note data can be represented as a score will be described.

As described above, the note-on or note-off of the melody sound is converted into melody data with an accuracy of, for example, a minimum measurement unit time of 1/96 of a whole note, and the RAM
7 is stored. Thereby, the reproduction of the melody sound can be recorded in real time with sufficient time accuracy. In order to generate note data that can be expressed as a musical score based on such melody data, a note data generation process is performed.

In this process, the time data LEN [i] and the command CD [i] recorded in the RAM 7 with an accuracy of, for example, a minimum measurement unit time of 1/96 of a whole note (see FIG. 3A) Is converted to note data having an accuracy of, for example, 1/16 of a whole note, that is, a 1 / 16th note time unit (hereinafter, referred to as a "minimum note unit"), and the note data in the RAM 7 is stored in the RAM 7. Is stored again in the storage area.

One set of note data is composed of note length data indicating a note length in 16th note time units, and a code indicating a note number or a rest of each note. Such note data can be easily converted into a score on a five-line staff.

Here, among the melody data, there is a case where the time length of the time data LEN [i] does not reach the minimum note length, and a case where the sounding timing of each melody sound slightly overlaps. Also, the start and end positions of each melody sound are not synchronized with each other in time units of sixteenth notes.

Therefore, in the process of generating note data, a note conversion algorithm for properly converting each time data LEN [i] to note length data is required. Therefore, first, the note conversion algorithm will be described with reference to the explanatory diagrams of FIGS.

FIG. 5 is a diagram showing how to explain the respective note conversion algorithms shown in FIGS. 6 to 10. FIG.
(a) shows the time unit of the note data. The interval between long ticks is the minimum note length, and the short ticks represent 1/2 of each minimum note length.

FIG. 5B shows the sound length (length of a line segment) of a melody sound that can be generated based on the detected melody data, and its start position (hereinafter referred to as “before correction start position”). ) And an end position (hereinafter, referred to as “pre-correction end position”). These timings are R
Time data LEN [i] sequentially read from AM7
(I = 1, 2,...).

FIG. 5 (c) shows the note length (length of a line segment) of a melody sound which can be generated based on the note data obtained by converting the melody data having the generation timing of FIG. 5 (b). , The start position thereof (hereinafter referred to as “post-correction start position”) and the end position thereof (hereinafter referred to as “post-correction end position”). These timings are determined by the note length data generated by the note conversion algorithm.

FIG. 6 is an explanatory diagram showing the first musical note conversion algorithm. That is, as shown in FIG. 6A, if the pre-correction start position is in the first half of the minimum note length range that includes the correction start position, the post-correction start position is set to A in FIG. ′, It is the left end position of the minimum note length range including the before-correction start position.

Further, the start position before correction is set to B or C in FIG.
As shown in FIG. 6, when it is in the latter half of the minimum note length range in which it is included, the corrected start position is, as shown in B ′ or C ′ in FIG. It is the right end position of the minimum note length range including the start position.

On the other hand, if the pre-correction end position corresponds to A or C in FIG.
As shown in FIG. 6, when it is in the latter half range of the minimum note length range in which it is included, the post-correction end position is, as shown in A ′ or C ′ in FIG. This is the right end position of the minimum note length range including the end position.

As shown in FIG. 6B, the end position before correction is the first half of the minimum note length range including the end position.
In the case of the range of / 2, the post-correction end position is the left end position of the minimum note length range including the pre-correction end position as shown in B 'of FIG.

Therefore, the melody data having the timing of A in FIG. 6 is converted into note data having a timing of A 'in FIG. 6 and having a note length of 2 (two 16th notes corresponding to eighth notes). The melody data having the timings B and C in FIG. 6 has a note length of 1 (corresponding to the sixteenth note) having the timings B 'and C' in FIG.
Is converted to each note data.

Further, rest note data having a note length of 1 is inserted between the note data having the timings B 'and C' in FIG. FIG. 7 is an explanatory diagram showing the second musical note conversion algorithm.

When a melody sound is reproduced by the legato playing technique, the sound of the next melody sound is often started immediately before the end of the sound of one melody sound, that is, immediately before the key is released. In this case, the timing of the two melody sounds is
As shown in FIG. 7A, B or C, they overlap for a very short time.

In such a case, as shown in the example of FIG. 7, when each note length of timing such as A, B, C determined by each melody data is longer than 最小 of the minimum note length, A The pre-correction end position of the melody data having the timing is shifted to the same position as the pre-correction start position of the melody data having the timing B. Similarly, B
Is shifted to the same position as the pre-correction start position of the melody data having the timing of C. A, B and C
, The pre-correction start position of each melody data having the respective timings is maintained as its own pre-correction start position.

After such replacement is performed, the above-described first note conversion algorithm of FIG. 6 is executed. As a result, each melody data having each timing of A, B and C is converted into each note data having each timing of A ', B' and C '.

FIG. 8 is an explanatory diagram showing a third musical note conversion algorithm. As in the case of the example of FIG. 7, when the timings of the two melody sounds overlap as in A, B or C, D in FIG. If there is a section shorter than 1/2,
After the section is absorbed (merged) into the other section such as A or D, the above-described first note conversion algorithm in FIG. 6 is executed. As a result, in FIG.
The melody data having the timing of
(Corresponding to the eighth note), the note number is converted to note data having the same A 'timing as the note number of the melody data having the timing of A. Similarly, the melody data having the timings C and D is converted to note data having a note length of 2 and the same D 'timing as the note number of the melody data having the note number D timing.

FIG. 9 is an explanatory diagram showing a fourth musical note conversion algorithm. As shown in FIG. 9A, when the note length determined by the melody data is shorter than 最小 of the minimum note length and does not overlap with melody data having another timing, the melody data is not converted to note data. Ignored.

FIG. 10 is an explanatory diagram showing a fifth note conversion algorithm. For example, as shown in FIG. 10B, when the note length determined by the melody data is exactly equal to 最小 of the minimum note length, the start position before correction is within the range of 前 of the first half of the minimum note length range in which it is included. , The melody data having the timing of B is converted to note data having a note length of 1 (minimum note length) having a timing of B '. On the other hand, as shown in FIG. 10A, a case where the note length also determined by the melody data is exactly equal to 最小 of the minimum note length, and the uncorrected start position corresponds to the latter half of the minimum note length range in which it is included. If it is in the range of 1/2, the melody data is ignored without being converted to note data, similarly to the above-described fourth note conversion algorithm in FIG. According note algorithm generation process described above for note data, the melody data stored in the melody data storage area of the RAM 7, the specific operation of the note data generating process to be stored in the RAM 7 and converts the note data that can be represented as a score This will be described below.

FIGS. 11 to 14 are operation flowcharts of note data generation processing, and FIG. 15 is an operation flowchart of note length provisional determination processing executed in the note data generation processing. These flows are realized as an operation in which the CPU 6 executes a control program stored in a ROM (not shown) or the like.

First, in step S1101, a work area in the RAM 7 and a note data storage area are cleared as initialization processing. Next, step S1102
, The content of the pointer i is initialized to “1”,
Further, the flag OFF_FLAG is initialized to “0”. These flags and pointers are provided as registers in the CPU 6 of FIG.

Subsequently, in step S1103, a register N_LEN for storing the length of each melody sound constituting the melody, a register R_LEN for storing a pause time of the melody sound, that is, a time in which the melody sound is not sounded, Register D_P for storing the current position on the time axis, register N_ for storing the start position of the melody sound
P, a register R_ for storing the pause start position of the melody sound
Each content of P is initialized to "0".

After the above operation, in step S1104, the time data LEN [i] and the command CD [i], which are the melody data at the storage position indicated by the pointer i, are read from the melody data storage area of the RAM 7 (step S1104). See FIG. 3 (a)).

Next, in step S1105, the register D
By adding the time data LEN [i] to _P, the current position on the time axis is moved. In step S1106, the type of the detected event is determined by determining the content of the read CD [i] (see FIGS. 3B, 3C, and 3D).

In this determination, if the event indicates the end of the melody (END, see FIG. 3C), note data indicating a rest up to the last bar is stored in the note data in the RAM 7 in step S1109. The data is written to the area, and the process ends.

If it is determined in step S1106 that the event indicates note-off (OFF, see FIG. 3B) or time-out (TIME OVER, see FIG. 3C), the melody sound is output in step S1107. , The elapsed time LEN [i] from the previous event is added to the register R_LEN indicating the pause time.

Then, in step S1108, the pointer i is incremented, and in step S1104
And the next melody data is read. The processes in steps S1104 to S1108 described above are repeatedly performed until the melody data to be read indicates a note-on (ON) event. When the note-on event is read, after step S1106, S111
0 is executed. In this way, the pre-correction start position (see FIG. 5) corresponding to one piece of melody data is determined.

In step S1110, the note number included in the note-on command CD [i] (FIG. 3B)
) Is stored in a first pitch register PITCH1 provided in the CPU 6 and a latest pitch register END_PITCH also provided in the CPU 6.

The following steps S1111 and S1112 will be described later, but initially OFF_FLAG is "1" and the determination in step S1111 is YES. Therefore, in step S1112, the time data LEN [i] = LEN stored at the head of the melody data storage area.
[0] is a register R_L indicating the pause time of the melody sound
Added to EN. The content of the register R_LEN corresponds to a rest from the start of writing of the melody to the start of reproduction of the first melody sound. This rest is actually stored as note data in the note data storage area in the RAM 7 if it is determined that the rest has a valid rest length by the determination processing in step S1144.

Thereafter, in step S1113, a value obtained by adding +1 to the content of the pointer i is stored in the pointer j. Also, a register N_L for storing the length of the melody sound
The contents of EN are cleared.

Next, in step S1114, the uncorrected start position (see step S1105) corresponding to the current melody data stored in the register D_P is stored in the register N_P storing the start position of the melody sound.

Subsequently, at step S1116, R
From the melody data storage area of AM7, the time data L which is the melody data at the storage position indicated by the pointer j
EN [j] and command CD [j] are read. This data is the next data of the melody data (LEN [i] and CD [i]) read in step S1105 (see step S1113).

In step S1117, the time data LEN [j] is added to the register D_P, so that the current position on the time axis is moved by the time of the read LEN [j]. .

Further, in step S1118, the time data LEN [j] is also added to the contents of the register N_LEN for storing the length of the melody sound. Next, in step S1119, it is determined whether or not the event indicated by the command CD [j] is an END command (see FIG. 3C) indicating the end of the melody.

If this determination is YES, provisional determination processing of the note length in step S1143 described later is executed. If the determination in step S1119 is NO, step S112
At 0, it is determined whether or not the event indicated by the command CD [j] is note-on.

If this determination is not note-on, in step S1121, the note number of the note-off event indicated by the command CD [j] is stored in the pitch register END_PITCH.
It is determined whether or not it is the same as the note number of the note-on event immediately before the event read out in 1104 (see S1110).

The melody sound having the same note number as the note number of the note-on event one event before is note-off, and if the judgment in step S1121 is YES, the flag OFF is set in step S1122.
_FLAG is set to “1”. Further, the content of the pointer i is set to a value obtained by adding +1 to the content of the pointer j.

FIG. 16 shows an operation example in which this state occurs. ON indicates a note-on event, and OFF indicates a note-off event. As understood from the figure,
This state is a state in which the melody sounds do not overlap, as represented by the state in which the first note conversion algorithm in FIG. 6 is applied. In this way, the pre-correction start position (N_P value), the pre-correction end position (D_P value), and the melody sound length (N_P) in a state where the melody sounds do not overlap.
LEN) is determined. Then, such a state is a state where the flag OFF_FLG is “1”.

As described above, when one set of the pre-correction start position, the pre-correction end position, and the melody sound length is determined, the above-mentioned provisional determination processing of the note length in step S1143 will be described later. First note conversion algorithm (FIG. 6)
) Is performed.

On the other hand, the event indicated by command CD [j] is note-on, and the determination in step S1120 is YE
When S is reached, the process proceeds to step S1123. FIG. 17 shows an operation example in which this state occurs. This state corresponds to the second state in FIG.
As shown in the state where the note conversion algorithm of FIG. 8 or the third note conversion algorithm of FIG.
This is a state in which another second melody sound is turned on before the first melody sound in which note-on is detected in 1106 is turned off, that is, a state in which melody sounds overlap.

In such a case, step S in FIG.
After the execution of step 1123, first, step S in FIG.
By the processing of 1124 to step S1138, the sound length of the first melody sound and the sound length of the second melody sound are detected, and subsequent steps S1139 to S114 in FIG.
2, the second note conversion algorithm of FIG. 7 or the third
Are executed in the first half of the note conversion algorithm.

First, in step S1123, OFF_F
The LAG is cleared and the contents of pointer i are replaced with pointer j
Is equal to the contents of As a result, step S in FIG.
After the processing of step S1140 in FIGS. 1124 to 14 is performed, the first half of the processing of the second note conversion algorithm in FIG. 7 or the third note conversion algorithm in FIG.
The process is continued from the event indicated by the pointer i equal to the value of the pointer j.

Next, in step S1124, the contents of the register N_LEN for storing the duration of the melody tone are stored in the first duration register N_LEN1 provided in the CPU 6 for storing the duration of the first melody tone. Is done.
As a result, the content of the first tone length register N_LEN1 initially indicates the time from the pre-correction start position of the first melody sound to the pre-correction start position of the second melody sound, as shown in FIG. It is set to the content of the indicated time data LEN [j]. At the same time, the contents of the second tone length register N_LEN2 provided in the CPU 6 for storing the tone length of the second melody tone are cleared.

Subsequently, in step S1125, note-on command CD corresponding to the second melody sound
The note number included in [j] is stored in the second pitch register PITCH2 provided in the CPU 6. At the same time, the content of the pointer k is set to a value obtained by adding +1 to the content of the pointer j.

Next, in step S1126, the time data LEN, which is the melody data at the storage position indicated by the pointer k, is read from the melody data storage area of the RAM 7.
[K] and the command CD [k] are read.

Then, steps S1127 and S112
8, the contents of the first tone length register N_LEN1 and the second tone length register N_LEN2 include the time data L
EN [k] is added.

At step S1129, the command CD
The event indicated by [k] is END indicating the end of the melody.
It is determined whether the command is a command (see FIG. 3C).
If this determination is YES, the process returns to step S1116 from step S1115 in FIG. 11, and after step S1118 in step S1116 to step S1118 in FIG. Be executed.

If the determination in step S1129 is NO, in step S1130, the note number included in the command CD [k] is the first melody tone whose pitch is the first melody tone.
It is determined whether or not the content is equal to the content of the pitch register PITCH1 and the command CD [k] is a note-off event. That is, here, it is determined whether or not the command CD [k] indicates the note-off of the first melody sound.

If the determination in step S1130 is NO, the flow advances to step S1137, where the note number included in the command CD [k] is equal to the content of the second pitch register PITCH2 which is the pitch of the second melody tone. Further, it is determined whether or not the command CD [k] is a note-off event.

If the determination in step S1137 is YES, before the first melody sound is turned off,
This is a case where the second melody sound ends sounding. In such a case, the second melody sound is ignored, and the process returns from step S1115 to S1116 in FIG. 11 to return to the process of searching for the note-off of the first melody sound.

If the determination in step S1137 is NO, the command CD [k] has nothing to do with either the first or second melody sound, so that step S1
After the content of the pointer k is further incremented by 1 at 138, the processing from step S1126 is repeated.

Then, the determination in step S1130 is YE
At the point of time S, as shown in FIG. 17, the length of the first melody tone is stored in the first tone length register N_LEN1. In this case, a series of processes described below are executed to detect the note-off point of the second melody sound.

First, in step S1131, the content of the pointer n provided in the CPU 6 is set to a value obtained by adding +1 to the content of the pointer k. Next, in step S1132, R
From the melody data storage area of AM7, the time data L which is the melody data at the storage position indicated by the pointer n
EN [n] and command CD [n] are read.

In step S1133, the time data LEN [n] is added to the contents of the second tone length register N_LEN2. Subsequently, step S113
In step 4, it is determined whether or not the event indicated by the command CD [n] is an END command (see FIG. 3C) indicating the end of the melody.

If this determination is NO, step S1135
And the note number contained in the command CD [n] is
Second pitch register P which is the pitch of the second melody tone
Command CD [n] equal to the content of ITCH2
Is a note-off event. That is, here, it is determined whether or not the command CD [n] indicates the note-off of the second melody sound.

If the determination in step S1135 is NO, in step S1136 the content of pointer n is further increased by +
After that, the processing from step S1132 is repeated.

Then, the determination in step S1135 is YE
At the time point when S is reached, as shown in the example of FIG.
This means that the tone length of the second melody tone has been stored in the tone length register N_LEN2.

If the determination in step S1134 is NO, and the event indicated by the command CD [n] is determined to be an END command indicating the end of the melody, that point in time is also the end of the second melody sound. It is time.

By the above series of processing, the tone length of each of the first and second melody sounds is set to the first tone register N_
After being obtained in LEN1 and the second tone register N_LEN2, respectively, the subsequent steps S1139 to S11 in FIG.
At 42, the first half processing of the second note conversion algorithm of FIG. 7 or the third note conversion algorithm of FIG. 8 is executed.

First, in step S1139, the tone length of the first melody tone stored in the first tone register N_LEN1 is set to 1 / the value of the variable comp_len in the RAM 7.
It is determined whether it is smaller than two. This variable stores a minimum note length indicated as a time length in units of the minimum measurement unit time. Now, the minimum measurement unit time serving as a reference for writing the melody sound is, for example, one whole note.
Has a time length of / 96. On the other hand, the minimum note length serving as a reference for generating note data has, for example, a time length of 1/16 of a whole note. Therefore, for example, a value (= 6) is stored as the minimum note length in the variable comp_len. In step S1139, it is determined whether the value of the first tone register N_LEN1 is smaller than the value 3.

If it is determined in step S1139 that the length of the first melody tone stored in the first tone register N_LEN1 is equal to or longer than １ ／ of the minimum note length, then in step S1140, 2nd tone register N
It is determined whether or not the length of the second melody tone stored in _LEN2 is smaller than の of the value of the variable comp_len in the RAM 7, that is, １ ／ of the minimum note length.

In step S1140, if it is determined that the length of the second melody tone stored in the second tone register N_LEN2 is also equal to or longer than 1/2 of the minimum note length,
At this point, a register N_ that stores the duration of the melody tone
As shown in FIG. 17, LEN stores the time length from the note-on time of the first melody sound to the note-on time of the second melody sound. That is, the pre-correction end position of the first melody sound is shifted to the same position as the pre-correction start position of the second melody sound, and the second note conversion algorithm described above (see FIG. 7). Has been executed.

Thereafter, the process proceeds to step S1143, and as will be described later, the provisional determination processing of the note length is performed on the contents of the register N_LEN for storing the note length of the melody sound, so that the first note described above is executed. The second half of the second note conversion algorithm, which is the same process as the note conversion algorithm (see FIG. 6), is executed.

On the other hand, if it is determined in step S1139 that the length of the first melody tone stored in the first tone register N_LEN1 is smaller than half the minimum note length, the flow advances to step S1141. Here, the first pitch register PITCH1 is added to the second pitch register PITCH2.
In the following step S1142, the second pitch register P is also stored in the latest pitch register END_PITCH.
The contents of ITCH2 are stored. After the above processing, FIG.
1, the process of steps S1115 to S1121 is repeated while the value of the pointer j is updated. Then, the command CD [j] indicating the note-off of the second melody sound is read out, and step S is executed.
When the determination at 1121 is YES, the time length from the note-on of the first melody sound to the note-off of the second melody sound is stored in the register N_LEN for storing the length of the melody sound, and , First pitch register P
The pitch of the second melody tone is stored in ITCH1. The content of the first pitch register PITCH1 is used as a note number of the note data in a note data writing process described later. In this manner, the section of the first melody sound is absorbed (merged) into the section of the second melody sound.

Next, in step S1139, after it is determined that the length of the first melody tone stored in the first tone register N_LEN1 is not less than half the minimum note length, in step S1140 Second tone register N_L
If it is determined that the length of the second melody tone stored in EN2 is smaller than 1/2 of the minimum note length, the process proceeds to step S1142, where the latest pitch register EN
The contents of the second pitch register PITCH2 are stored in D_PITCH. After the above processing, step S1 in FIG.
Returning to 115, the processing of steps S1115 to S1121 is repeated while the value of the pointer j is updated. Then, the command CD [j] indicating the note-off of the second melody tone is read out, and when the determination in step S1121 is YES, the first CD-NEN register stores the length of the melody tone. The time length from the note-on of the melody sound to the note-off of the second melody sound is stored, and the first pitch register PITCH1 stores the first time.
The pitch of the melody sound is stored.
In this way, the section of the second melody sound is absorbed (merged) into the section of the first melody sound.

As described above, the first half of the above-described third note-forming algorithm (see FIG. 8) is executed. Thereafter, the process proceeds to step S1143, and as described later, a register N_ for storing the length of the melody sound is provided.
The provisional determination processing of the note length is performed on the content of LEN, so that the second half of the third note conversion algorithm, which is the same processing as the above-described first note conversion algorithm (see FIG. 6). Is executed.

As described above, the registers N_P and N_P
When a set of data including the pre-correction start position and the melody sound length is determined as the contents of _LEN, a temporary note length determination process is executed in step S1143. Here, as described above, the first note conversion algorithm (see FIG. 6) and the second and
The second half of the processing of the third note conversion algorithm (see FIGS. 7 and 8), the processing of the fourth note conversion algorithm of FIG. 9, and the processing of the fifth note conversion algorithm of FIG. 10 are executed.

FIG. 15 is a flowchart showing the operation of the tentative note length determination process in step S1143 of FIG.
This operation will be described with reference to the operation example of FIG.
The numbers in parentheses in the following description are values corresponding to FIG. 18 and indicate the time length in units of the minimum measurement unit time of the melody data (corresponding to 1/96 of a whole note). ing.

First, a variable comp_len is defined when a program corresponding to the flow of FIG. 15 is started. As described above, this variable stores the minimum note length indicated as a time length in units of the minimum measurement unit time, for example, a value (= 6).

The variables start_point and mel_len are delivered from the program corresponding to the flow of FIG. The variable start_point is a variable in which the minimum note length is stored as the pre-correction start position stored in the register N_P.
The remainder value obtained by dividing by the value of comp_len is stored. In the example of FIG. 18, a value (= 4) is stored in the variable start_point.

The variable mel_len contains the register N_LEN
Is stored as it is.
In the example of FIG. 18, a value (= 6) is stored in the variable mel_len. After the above operation, processing corresponding to the flow in FIG. 15 is started.

First, in step S1501, the register no
te_length is cleared. In this register, note length data of each note finally stored in the note data storage area of the RAM 7 is stored.

Next, in step S1502, the register bu
The value (= 4) of the variable start_point is written to f1. Subsequently, in step S1503, the variable comp is stored in the register buf2.
The variable start_po from the minimum note length (= 6) stored in _len
The value (= 2) obtained by subtracting the value of int (= 4) is stored. This value corresponds to the length of the head of the melody sound included in the range of the minimum note length including the pre-correction start position.

In step S1504, the register bu
In f3, the length (= 6) of the melody sound stored in the variable mel_len is stored. After each of the storing processes described above, first, in step S1505, the duration (= 6) of the melody tone written in the register buf3 is set to a value that is a half of the minimum note duration (= 6) stored in the variable comp_len. It is determined whether (= 3) is larger, equal, or smaller.

As shown in FIG. 18, buf3> comp_len / 2
If it is determined in step S1508 that the register bu
The value (= 4) of the variable start_point stored in f1 is
1/2 value (= 3) of the minimum note length (= 6) stored in mp_len
It is determined whether it is greater than.

As in the example of FIG. 18, if this determination is YES, step S1509 is not executed and step S15 is executed.
10 is executed. This means that the pre-correction start position is located in the latter half of the first range of the minimum note length that includes the correction position. Will not be counted.

On the other hand, as shown by the broken line in FIG. 18, for example, when the pre-correction start position is a value (= 2) and the determination in step S1508 is NO, the pre-correction start position is the minimum note It means that it is located in the first half of the first range of the length. In this case, in step 1309,
The contents of the register note_length are incremented, and note lengths in the first range of the minimum note length are counted.

Next, in step S1510, the register bu
The length (= 6) of the melody sound stored in f3 is stored in register buf2.
The length (==) of the first part of the melody sound included in the first range of the minimum note length including the uncorrected start position stored in
It is determined whether the value is equal to or greater than the value of 2).

As in the example shown in FIG. 18, if this determination is YES, it means that the melody tone is present after the second range of the minimum note length. In this case, the value of each variable is updated in order to determine the validity / invalidity of the note within the second range of the minimum note length.

That is, in step S1511,
From the melody sound length (= 6) stored in register buf3,
The note length (= 2) of the first part of the melody sound included in the first range of the minimum note length including the uncorrected start position stored in the register buf2 is subtracted, and a new register buf3 is added.
(= 4).

Next, in step S1512, the register bu
The minimum note length (= 6) stored in the variable comp_len is stored in f2. Also, in step S1513, the register buf1
Is cleared.

After the above operation, in step S1505,
Again, the note length (= 4) of the melody note in the second range of the minimum note length stored in the register buf3 is １ ／ (= 6) of the minimum note length (= 6) stored in the variable comp_len. For 3), it is determined whether it is large, equal, or small.

As shown in the example of FIG. 18, buf3> comp_le again
If it is determined to be n / 2, it means that the melody sound exists also in the second half of the second range of the minimum note length and thereafter. In this case, in step S1508, the register buf1
Value (= 0) is the minimum note length (=
It is determined whether or not the value is greater than 1/2 (= 3) of (6).

As shown in the example of FIG.
After the execution of step 1513, the determination in step S1508 is always YES, and in step S1509, the register
The content of note_length (= 0) is incremented (= 1), and note lengths in the second range of the minimum note length are counted.

Next, in step S1510, the register bu
It is determined whether or not the melody note length (= 4) in the second range of the minimum note length stored in f3 is equal to or longer than the minimum note length (= 6) stored in the register buf2. .

As in the example of FIG. 18, if this determination is NO, it means that the melody tone does not exist after the third range of the minimum note length. In this case, the provisional determination process of the note length in step S1123 of FIG. 12 is terminated. Thus, in the example of FIG. 18, the note length finally obtained in the register note_length is (= 1).

On the other hand, if the melody sound exists after the third range of the minimum note length, step S151.
The processes after 1 are executed again, and the same operation is repeated.

During the above-described repetitive operation, step S
If it is determined in 1505 that buf3 = comp_len / 2,
This means that the pre-correction end position is located exactly in the middle of the currently noted minimum note length range. In this case, after the determination in step S1506 is NO (since the value of register buf1 is always set to 0 in step S1513), in step S1507, register no
After the content of te_length is incremented and the note length of the minimum note length range including the end position before correction is counted, the tentative determination process of the note length in step S1143 in FIG. 14 ends.

Further, during the above-described repetitive operation, step S
If it is determined at 1505 that buf3 <comp_len / 2,
This means that the pre-correction end position exists in the first half of the range of the minimum note length that is currently focused on. In this case, note lengths in the range of the minimum note length including the pre-correction end position are not counted, and are not counted, and are directly stored in step S11 in FIG.
The provisional determination process of the note length of 43 is completed.

The processing corresponding to the above-described first note conversion algorithm (see FIG. 6) is realized by the provisional determination processing of the note length in FIG. 15 described above. As described above, the second and third note conversion algorithms (FIGS. 7 and 8)
The same applies to the processing in the latter half of the above (see).

Next, in the provisional determination process of the note length in FIG. 15, when it is determined that buf3 <comp_len / 2 at the time when the determination process in step S1505 is first performed,
The length of the melody sound stored in register buf3 is
This is the case where the value is shorter than 1/2 the minimum note length stored in mp_len. In this case, the provisional determination process of the note length in step S1143 in FIG. 14 is terminated while the value of the register note_length is 0.

That is, this case is a case where the above-described fourth note conversion algorithm (FIG. 9) is to be applied. In this case, after the processing in step S1143 in FIG. 14, the processing in steps S1144 to S1146 is skipped, and the current melody data is ignored.

Further, in the provisional determination process of the note length in FIG. 15, when it is determined that buf3 = comp_len / 2 at the time when the determination process of step S1505 is first performed,
The length of the melody sound stored in register buf3 is
This is the case when it is exactly equal to half of the minimum note length stored in mp_len. In this case, the above-described fifth note conversion algorithm (see FIG. 10) is to be applied.

In this case, first, in step S1506, it is determined whether or not the value of the variable start_point stored in the register buf1 is larger than half the minimum note length stored in the variable comp_len.

If this determination is NO, it means that the pre-correction start position is located in the first half of the range of the minimum note length that includes it. In this case, step S1
In step 507, after the content of the register note_length is incremented to have the value 1, the value of the register note_length is increased to 1 in step S1143 of FIG.
The provisional determination processing of the note length of is ended.

On the other hand, the determination in step S1506 is YES
If it is, the value of the register note_length is set to 0, and the tentative determination processing of the note length in step S1143 in FIG. 14 ends. This means that the pre-correction start position is located in the latter half of the minimum note length range that includes it, in which case note lengths in the minimum note length range are not counted. Become. Then, in this case, after the processing of step S1143 in FIG.
The process of S1146 is skipped, and the current melody data is ignored.

As described above, step S1 in FIG.
As a result of the execution of the provisional determination process of the note length of 143, if a valid value greater than 0 is obtained as the value of the note length note_length, a process of writing note data and the like after step S1144 described later is executed. Is done.

On the other hand, step S114 in FIG.
As a result of the provisional determination processing of the note length of No. 3, the note length
If 0 is returned as the value of note_length, the processing of steps S1144 to S1146 is skipped, and the current melody data is ignored. As a result, the latter half of the processing of the fourth note-forming algorithm (FIG. 9) and the fifth note-writing algorithm (FIG. 10) is realized.

As described above, by the provisional determination processing of the note length in step S1143 in FIG. 14 corresponding to the first to fifth note conversion algorithms in FIGS.
When the effective note length data is obtained, the rest data is written to the note data storage area of the RAM 7 by the processing of steps S1144 and S1145 described later in FIG. 14, and then step S1146 is executed.

Here, as for the melody data to be currently converted to a note, the note number stored in the first pitch register PITCH1 and the note length data obtained in the register note_length are stored in the RAM as note data.
7 is written to the note data storage area.

Finally, the operation of writing rest data will be described. As exemplified in FIG. 6, FIG. 9 or FIG.
Rest data needs to be inserted at a timing when there is no note data. The rest data is created based on the contents of the register R_P storing the melody sound pause start position and the contents of the register R_LEN storing the pause time during which the melody sound is not sounded.

The setting of the register R_P for storing the pause start position of the melody sound and the initial setting of the contents of the register R_LEN for storing the pause time of the melody sound are executed after the processing of step S1146 in FIG. 14 is executed. It is set by the processing of steps S1147 to S1153 in FIG.

First, in step S1147, the register bu
It is determined whether the value of f3 is smaller than half the minimum note length stored in the variable comp_len. This register bu
f3 holds the time length from the beginning of the minimum note length range including the pre-correction end position to the pre-correction end position after the tentative note length determination process (FIG. 15) in step S1143 is performed. Have been.

Therefore, the determination in step S1147 is YE
If it is determined that the value of the register buf3 is smaller than 1/2 of the minimum note length stored in the variable comp_len, the pre-correction end position is set to the first half of the range of the minimum note length that includes it. Means to be located at In this case, the first note conversion algorithm shown in FIG. 6 is executed in the tentative note length determination process (FIG. 15) in step S1143, so that the post-correction end corresponding to the pre-correction end position is performed. The position is corrected to the left end position of the minimum note length range including the pre-correction end position, as exemplified by B 'in FIG.

As a result, the melody sound is not generated for the time length from the beginning of the minimum note length range including the end position before correction held in the register buf3 to the end position before correction.

Therefore, the determination in step S1147 is YE
If it becomes S, in step S1150, the register R_LEN storing the pause time of the melody sound is
The time held in the register buf3 is added.

In step S1151, the register R_P for storing the melody sound pause start position contains:
A value obtained by subtracting the content of the register buf3 from the content of the register D_P is stored. Now, the end position before correction is stored in the register D_P in step S1117 (for example, see FIG. 16). Further, the register buf3 holds the time length from the head of the range of the minimum note length including the end position before correction to the end position before correction. Therefore, the register R_P stores the post-correction end position which is the left end position of the minimum note length range including the pre-correction end position. That is, the rest starts from this position.

Subsequently, in step S1152, the flag O
It is determined whether the value of FF_FLAG is 1 or not. If this determination is NO, the above-described determination in step S1120 is YES, that is, before the melody sound for which note-on is detected in step S1106 is note-off, as shown in FIG. This is a state in which another melody sound is note-on, that is, a case where melody sounds overlap.

In such a case, step S1153
After the current position on the time axis stored in the register D_P is returned by the time data LEN [j] (see FIG. 17), the process returns to step S1104 in FIG.
The next melody data LEN [i] and CD [i] are read from the melody data storage area of AM7. Here, since the value of the pointer i is equal to the value of the pointer j in step S1123 described above, step S11
The melody data read back to 04 is the same as the data determined as the note-on event in step S1120 (see FIG. 17).

On the other hand, the determination in step S1152 is YES
In the case of, when the determination in step S1120 described above is NO and the determination in step S1121 is YES,
That is, as shown in FIG. 16, the melody sounds do not overlap.

In such a case, the process immediately returns to step S1104 in FIG. 11, and the next melody data LEN [i] and CD
[I] is read.

On the other hand, if the determination in step S1147 is NO and it is determined that the value of the register buf3 is equal to or more than の of the minimum note length stored in the variable comp_len, the pre-correction end position is determined. Means that it is located in the latter half of the range of the minimum note length that includes it. And
In this case, the first note conversion algorithm shown in FIG. 6 is executed in the tentative note length determination process (FIG. 15) in step S1143, so that the post-correction end position corresponding to the pre-correction end position becomes As shown in FIG. 6A 'or C', the correction is made to the right end position of the minimum note length range including the end position before correction.

In such a case, the correction operation does not generate any extra time during which the melody sound is not generated. Therefore, the determination in step S1147 is NO
In step S1148, the contents of the register D_P, which is the current final position of the melody sound, is stored in the register R_P for storing the pause start position of the melody sound (step S1148).

In step S1149, a value 0 is set in the register R_LEN for storing the melody sound pause time.
Is substituted. Thereafter, after the above-described processing in step S1152 or S1153 is executed, the process returns to step S1104 in FIG. 11, and the next melody data LEN [i] and CD [i] are stored in the melody data storage area of the RAM 7.
Is read.

Steps S1147 to S11 described above
By the process of 53, the setting of the register R_P for storing the pause start position of the melody sound and the initial setting of the contents of the register R_LEN for storing the pause time of the melody sound are performed.

Next, updating of the register R_LEN for storing the pause time of the melody sound is performed in steps S1107 and S1112 in FIG. 11 described above. Step S11
Step 07 is executed via step S1106 when the melody data is sequentially read from the melody data storage area of the RAM 7, and when the melody data indicating the note-off or time-over event is read. That is, each time data LEN [i] until the next note-on occurs is accumulated in the register R_LEN as a time indicating the pause time of the melody sound.

Step S1112 is identical to step S11.
This is executed when it is determined in step 1111 that the value of the flag OFF_FLAG is 1. In this case, in the previous note conversion process, when the above-described determination in step S1120 is NO and the determination in step S1121 is YES, that is, when the melody sounds do not overlap as shown in FIG. It is. In this case, there is no melody sound from the previous melody sound pre-correction end position (note-off position) to the current melody sound pre-correction start position (note-on position). Therefore, the time data LEN [i] representing this time is stored in the register R_LEN as the time indicating the pause time of the melody sound.
Will be accumulated.

By the operation described above, the contents of the register R_P for storing the pause start position of the melody sound and the register R_LEN for storing the pause time of the melody sound are determined.

Then, based on the contents of each register determined in this way, the rest length determination processing of step S1144 in FIG. 14 is executed, and rest data is created. The rest length determination processing in step S1144 is exactly the same as the note length provisional determination processing in FIG. 15 executed as the above-described step S1143 processing.

That is, in FIG. 15, the variables start_point and mel are read from the program corresponding to the flow of FIG.
_Len is passed. The variable start_point stores the remainder value obtained by dividing the pause start position of the melody sound stored in the register R_P by the value of the variable comp_len storing the minimum note length.

The variable mel_len contains the register R_LEN
The pause time of the melody sound stored in is stored as it is. After the above-described operation, processing corresponding to the flow of FIG. 15 is executed, and as a result, rest length data in units of the minimum note length is obtained in the register note_length.

As a result of the rest length determination processing in step S1144, when effective rest length data is obtained in the register note_length, the rest length data is stored in the RAM 7 as a part of the note data in step S1145. Written to the area.

[0186]

According to the present invention, the reproduction data of a melody sound recorded in real time along a time axis having a predetermined minimum measurement unit time as a unit is sequentially read out and converted into a musical note, thereby obtaining a melody sound. It is possible to appropriately associate notes or rests while correcting each sounding section.

In particular, according to the present invention, when two melody sounds overlap with each other for a short time like the legato playing method, the sounding times of the two melody sounds are both 1 / the predetermined minimum note length.
If the length is longer than 2, two melody sounds are generated in a short time, such as legato performance, by sequentially determining the sound generation section corresponding to each sound generation start position from each sound generation start position to another sound generation start position. Even in the case of overlapping, two melody sounds can be appropriately separated into two sounding sections.

On the other hand, if any one of the tone generation times is shorter than 1/2 of the predetermined minimum note length, the interval from each tone generation start position to the tone generation end position corresponding to another tone generation start position is set as each tone generation start position. The corresponding sounding sections are sequentially determined, and the pitch data corresponding to a time length shorter than 1/2 of the predetermined minimum note length is rewritten with the content of the pitch data corresponding to the other time length, thereby to change the section. Longer melody sounds can be absorbed. As a result,
Unnatural note data can be prevented from being generated.

Since the note data obtained by the above-described note conversion process represents notes or rests in units of a predetermined minimum note length, it is easy to express those notes or rests on a staff notation. Becomes possible.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an embodiment of a music transcription device according to the present invention.

FIG. 2 is an operation flowchart of a filtering process performed by a DSP.

FIG. 3 is a diagram showing a format of melody data written to a RAM 7;

FIG. 4 is an operation flowchart at the time of writing melody data.

FIG. 5 is an explanatory diagram (part 1) of a musical note conversion algorithm.

FIG. 6 is an explanatory diagram (part 2) of a musical note conversion algorithm.

FIG. 7 is an explanatory diagram (part 3) of a musical note conversion algorithm.

FIG. 8 is an explanatory diagram (No. 4) of the note conversion algorithm.

FIG. 9 is an explanatory diagram (No. 5) of the note conversion algorithm.

FIG. 10 is an explanatory diagram (part 6) of the musical note conversion algorithm.

FIG. 11 is an operation flowchart (part 1) of a note data generation process.

FIG. 12 is an operation flowchart (part 2) of a note data generation process.

FIG. 13 is an operation flowchart (part 3) of a note data generation process.

FIG. 14 is an operation flowchart (part 4) of a note data generation process.

FIG. 15 is an operation flowchart of a musical note length provisional determination process.

FIG. 16 is a diagram (part 1) illustrating an operation example of note data generation processing;

FIG. 17 is a diagram (part 2) illustrating an operation example of note data generation processing;

FIG. 18 is a diagram (part 3) illustrating an operation example of note data generation processing;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Anti-aliasing filter 2 A / D converter 3 DSP 4 ROM for filter coefficient 5 Work area RAM 6 CPU 7 RAM 8 Musical sound generation circuit 9 Amplifier 10 Speaker 11 Switch part

Continuation of the front page (56) References JP-A-5-108062 (JP, A) JP-A-4-347892 (JP, A) JP-A-58-123589 (JP, A) JP-A-58-7695 (JP, A) JP-A-62-30300 (JP, A) JP-A-1-219636 (JP, A) JP-A-1-219625 (JP, A) JP-A-1-321484 (JP, A) 60-169695 (JP, U) JP-B 4-1357 (JP, B2) JP-B 64-10079 (JP, B2) JP-B 1-48557 (JP, B2) (58) Fields surveyed (Int. Cl. ^7, DB name) G10G 3/04 G10H 1/00 G10H 1/00 101 G10H 1/00 102

Claims (1)

(57) [Claims] 1. Event data indicating the occurrence of an event to start or end sounding of a musical tone that is sequentially input, pitch data indicating the pitch of the musical tone that is sequentially input, and generation of the event data that is sequentially input Based on the time data indicated by the elapsed time in units of a predetermined minimum measurement unit time between timings, a time axis in which a valid sound generation section and a mute section other than the valid sound generation section are set in the minimum measurement unit time After sequentially determining above, note data representing a note of a pitch based on the pitch data with a pitch corresponding to the valid sounding section, and a rest of a note length corresponding to the silence section, And a note data generating means for sequentially generating rest data representing the note data, wherein the note data generating means comprises: the sequentially input event data, pitch data, and time data. Based on each set, the sound start position and the sound end position are sequentially determined on a time axis in units of the minimum measurement unit time, and another sound is generated between each sound start position and the corresponding sound end position. A first position from each sounding start position to a corresponding sounding end position when there is a start position;
If both the time length of the sound generation and the second time length from the other sound generation start position to the corresponding sound generation end position are longer than 所 定 of the predetermined minimum note length, the respective sound generation start positions are replaced with the other sound generation start positions. Up to the sounding start position is sequentially determined as a sounding section corresponding to each sounding start position, and one of the first time length and the second time length is set to be smaller than １ ／ of the predetermined minimum note length. In the case of a short sounding start time, a portion from each sounding start position to a sounding end position corresponding to the other sounding start position is sequentially determined as a sounding section corresponding to each sounding start position, and 1 / の of the predetermined minimum note length is determined. The pitch data corresponding to the time length shorter than 2 is rewritten by the content of the pitch data corresponding to the other time length, and the time length of the sequentially determined sounding section is longer than 1/2 of the predetermined minimum note length. If the pronunciation interval It is determined as a valid sounding section, and the sounding start position of the valid sounding section is included in the first half of the minimum note length range including the sounding start position on the time axis in units of the predetermined minimum note length. If this is the case, generate note data with the start position of the range of the minimum note length as the start position of the note interval, and if included in the latter half, set the start position of the range of the minimum note length next to the range of the minimum note length. Generate note data as a sounding start position of a note section, and set an end position of the effective sounding section to a range of a minimum note length including the end position on a time axis in units of the predetermined minimum note length. If it is included in the first half, it generates note data with the start position of the minimum note length range as the end position of the note section, and if it is included in the second half, it generates the next minimum note length range The start position is the end position of the note section Generating a note data that, music transcription apparatus characterized by.