JP5609281B2 - Method and apparatus for creating differential code data - Google Patents

Description

  The present invention relates to creation of code data such as a MIDI format obtained by encoding an acoustic signal, and more particularly to a technique for creating differential code data suitable for use in comparing two acoustic signals.

  Conventionally, when processing such as encoding an acoustic signal recording music or voice into a format such as MIDI or embedding a digital watermark in the acoustic signal, comparative evaluation with the original sound has been performed. In the case of an acoustic signal, it is difficult to present a plurality of images at the same time or to present them in a superimposed manner, and therefore, comparative evaluation with a time difference is performed. In the case of comparative evaluation with a time difference, humans memorize the sound paths heard immediately before, so the evaluation result changes depending on the order of presentation, and it is possible to evaluate at the same time and under the same environmental conditions was difficult.

  In view of this, it is conceivable to create a differential waveform between two waveform forms of an acoustic signal and evaluate the difference between the two in terms of waveform. In addition, the applicant has proposed a technique for creating and comparing integrated spectrum data and envelope data in order to quantify the comparison between a MIDI-encoded decoded signal and the original sound (see Patent Document 1).

JP 2002-372968 A

However, in the method of creating a differential waveform, if the phases change, a differential waveform with a very large amplitude is obtained although there is essentially no substantial difference. For example, even if two waveforms are originally the same waveform, if the phases are inverted by 180 degrees, the difference waveform has an amplitude twice that of the original signal. Conversely, if the signal components in the high-frequency region are significantly different, there may be no significant difference on the differential waveform, and the visual difference on the differential waveform and the auditory difference on which each is reproduced and compared. Inconsistencies may occur between the two. Moreover, since the technique described in Patent Document 1 is intended only for visual presentation, it is difficult to make an auditory evaluation.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a differential code data creation method and apparatus capable of obtaining differential code data that can visually and audibly evaluate two types of encoded acoustic signals. And

In order to solve the above-described problem, in the first aspect of the present invention, when creating one differential code data representing a difference between two independent code data obtained by encoding an acoustic signal, the code The data consists of event data arranged in chronological order, and each event data includes identification information of the event at the start or end of pronunciation , time of the event , pitch, channel number, and sound intensity attribute information. the provided, while applying different channel numbers for each event data of the two code data, the time is sorted in ascending order, into one time-series code data with two channel numbers, create a 2-channel code data, the start of sounding of the event data, two pronounced end of event data having the height of the event data and the same sound When a section composed of a pair of event data is defined as a sound generation section, the sound generation section of one channel in the two-channel code data is temporally different from the sound generation section of the other channel having the same pitch. When all or a part of the sounding part overlaps, a section where both sounding sections overlap is generated as a new sounding section with a new sound intensity as the difference between the sound strengths of both sounding sections, Among the overlapping sound generation sections, for a section where a note-on section exists only in one channel, a sound generation section having the sound intensity of the one channel is generated, and each of the updated sound generation sections is generated. The event data at the start of pronunciation and the event data at the end of pronunciation are sorted in time series and encoded in a predetermined format.

  According to the first aspect of the present invention, two code data are integrated into one two-channel code data, and a sounding section of the same note number where sounding times overlap in two channels is divided into an overlapping section and a non-overlapping section. Since the velocity of the overlapping section is updated to a differential value, the note-on event data and note-off event data that specify each sounding section are sorted in time series and encoded as differential code data. It is possible to obtain differential code data that can visually and audibly evaluate the two types of acoustic signals.

  Further, in the second aspect of the present invention, in the first aspect of the present invention, when calculating the difference between the overlapping pronunciation sections, if the sound intensity of the overlapping section is 0, the sound intensity is set to other than 0. In the range of 1 to 127, the value is changed to a small value.

  Further, in the third aspect of the present invention, in the first aspect of the present invention, when calculating the difference between the overlapping pronunciation sections, if the sound intensity of the overlapping section is 0, the overlapping section is detected from the encoding target. It is characterized by deleting.

  According to the second and third aspects of the present invention, 0 is not finally given as the sound intensity, and it becomes possible to distinguish from the section that was 0 before the difference calculation.

  Also, in the fourth aspect of the present invention, in any one of the first to third aspects of the present invention, the integration of the code data is an event of sounding start and sounding end of all one of the two code data. A plurality of offset values are uniformly added to the time of data, and based on the plurality of offset values, the steps up to the differential code data encoding step are performed a plurality of times to encode a plurality of differential code data. For each of the plurality of differential code data encoded by the encoding of the differential code data, the sounding start event data and the corresponding sound end 2 having the same pitch A sounding interval composed of a pair of event data is extracted, and the sound intensity and time interval of the sounding intervals of all the extracted sounding intervals (time and event start time of sounding end event data) are extracted. And of calculating the product sum of the difference) between the time of the event data, wherein the total value was made to sort the encoded difference data is minimal.

  According to the fourth aspect of the present invention, the offset value is changed and added to the time of the sounding start and sounding end event data, and a plurality of types of differential code data are created according to the change of the offset value, Since the differential code data that minimizes the sum of the products of the sound intensity and the time interval of the sounding section is selected, the differential code data that corrects the positional deviation in the time axis direction can be created in consideration of the phase. It becomes possible.

  Further, in the fifth aspect of the present invention, in any one of the first to fourth aspects of the present invention, the encoding of the differential code data is started as a sounding section obtained by the difference processing of the overlapping sounding sections. And when the sound generation end time is the same, a process for deleting event data constituting the sound generation section is further performed.

  According to the fifth aspect of the present invention, after the difference between the overlapping sound generation intervals, the event data constituting the sound generation interval is deleted when there is the same sound generation start time and sound end time. It is possible to prevent the occurrence of an event that causes an operation.

  Further, in a sixth aspect of the present invention, in any one of the first to fifth aspects of the present invention, the differential code data obtained by encoding the differential code data is displayed on a screen and / or the difference is displayed. The code data is acoustically reproduced by an acoustic reproduction means.

  According to the sixth aspect of the present invention, since the differential code data is displayed on the screen of the display means and / or the differential code data is acoustically reproduced by the acoustic reproduction means, the two kinds of acoustic signals are visually and audibly heard. It becomes possible to evaluate.

  According to the present invention, it is possible to obtain differential code data that can visually and audibly evaluate two types of encoded acoustic signals.

It is a functional block diagram which shows one Embodiment of the difference code data production apparatus which concerns on this invention. It is a flowchart which shows the outline | summary of the production method of the difference code data which concerns on this embodiment. It is a flowchart which shows the detail of a MIDI data integration process. It is a figure for demonstrating the outline | summary of an overlapping note-on area difference process. It is a flowchart which shows the detail of an overlap note-on area difference process. It is a flowchart which shows the detail of the overlap note-on area difference process with respect to the channel 0. FIG. It is a flowchart which shows the detail of a 0 duration event deletion process. It is a figure which shows the mode of the phase shift at the time of a difference process.

<Device configuration>
DESCRIPTION OF EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a functional block diagram showing an embodiment of a differential code data creation device according to the present invention. As shown in FIG. 1, the differential code data creation device according to the present embodiment includes a data input unit 11, an arithmetic processing unit 12, a display unit 13, an acoustic output unit 14, and a data output unit 15, and a hard disk or the like (not shown). It has a storage device, a keyboard for inputting instructions, an input device such as a mouse, and the like, and is realized by incorporating a dedicated program into a general-purpose computer to which an input / output device is connected.

  In FIG. 1, a data input unit 11 is for inputting code data such as MIDI data. The arithmetic processing unit 12 has a CPU and a memory, reads a dedicated program from the storage device into the memory and executes it, and realizes the main functions of the differential code data creation device of the present invention. The display unit 13 is realized by a display device such as a liquid crystal display, and displays the differential code data created by the arithmetic processing unit 12. The acoustic output unit 14 is realized by an amplifier and a speaker, and acoustically outputs the difference code data created by the arithmetic processing unit 12. The data output unit 15 is for outputting code data such as MIDI data in a data format.

<Process overview>
FIG. 2 is a flowchart showing an outline of a method for creating differential code data according to the present embodiment. The differential code data creation method according to the present embodiment is performed by the differential code data creation apparatus shown in FIG. 1 executing a program that records the detailed procedure of each step (each stage) shown in FIG. Is called.

  As the code data used in the present invention, any data may be used as long as each sound is in a format expressing the sound intensity, pitch and time information. Regarding time information, it is possible to specify the start time, end time, and pronunciation time by some method, whether it is specified by the start time and end time or the format specified by the start time and pronunciation time. It ’s fine. In this embodiment, as the code data, MIDI is used in which note-on event data is set at the start of sound generation, note-off event data is set at the end of sound generation, velocity is set as the sound intensity, and note number is set as the pitch. . Therefore, the following description will be made taking MIDI data as an example.

  As a method of creating differential code data, first, the arithmetic processing unit 12 which is a computer main body reads two MIDI data from the data input unit 11. Here, it is assumed that two MIDI data M1 and M2 are read. When the two MIDI data M1 and M2 are read, the arithmetic processing unit 12 integrates the two read MIDI data M1 and M2 into one time-series two-channel MIDI data (S1).

  At the time of integration, the arithmetic processing unit 12 sets Event1 [] and Event2 [] as the MIDI event arrays of the two read MIDI data M1 and M2. The total number of note events of the two MIDI event arrays is Ne1 and Ne2, and each event array has four values [time, channel, note-number, velocity] as attributes. An absolute value of a note event is given to time, and a value indicated by a relative time called delta time is converted into a cumulative value from the start time. The unit of the absolute time is a user-specified integer value like the delta time. The channel can take a value of 0 to 15 according to the MIDI standard, but Event1 [] is 0 and Event2 [] is 1. “note-number” and “velocity” take a value of 0 to 127. When the note is on, a value of 1 or more is given as velocity, and when the note is off, 0 is given as velocity.

  Then, an array EventW [] as an integration destination is prepared. Further, the pointers to Event1 [], Event2 [], and EventW [] are p1, p2, and pw, and p1 = p2 = pw = 0 is set in the initial state. When the total number of note events of EventW [] is Nw, Nw = Ne1 + Ne2. For phase alignment, offset1 [h] and offset2 [h] are added to the time of Event1 [] and Event2 [], and the initial value of h is set to zero. For example, when the two MIDI data are subjected to frequency analysis with a minimum frame interval of 64 (64/1536 sec) for two acoustic signals and converted to MIDI data, respectively, H = 12, and −512/1536. Suppose that phase shift correction is performed in a range of ˜ + 512/1536 sec, and offset1 [h] = {0,0,0,0,0,0,0,64,128,384,448,512}, offset2 [h] = {0, 64, 128, 384, 448, 512, 0, 0, 0, 0, 0, 0} are prepared. As will be described later, in the present invention, it is possible not to perform phase correction. In this case, H = 1, offset1 [h] = 0, and offset2 [h] = 0 are set. .

<MIDI data integration>
When the array is ready, the arithmetic processing unit 12 starts the MIDI data integration process. Details of the MIDI data integration process in S1 will be described with reference to the flowchart of FIG. First, the arithmetic processing unit 12 determines whether or not the pointer p2 is smaller than the total number of note events Ne2 (S101). As will be described later, p2 is incremented by 1 every time a note event in the array Event2 [] is processed. Therefore, if p2 is smaller than Ne2, an unprocessed note event remains in the array Event2 []. Means. That is, in S101, it is determined whether or not the processing of the note event in the array Event2 [] has been completed.

  If the pointer p2 is smaller than the total number of note events Ne2, Event2 []. time is Event1 []. It is determined whether it is smaller than time (S102). This determines whether the absolute time of the note event in Event2 [] specified by the pointer p2 is earlier than the absolute time of the note event in Event1 [] specified by the pointer p1. And Event2 []. time is Event1 []. If it is smaller than time, the process for Event2 [] is performed (S103), and Event2 []. time is Event1 []. If it is not less than time, the process for Event1 [] is performed (S105). In S101, if the pointer p2 is equal to or greater than the total number of note events Ne2, this means that all the processing of note events in the array Event2 [] has been completed. In this case as well, the process proceeds to S105, where Event1 Process for [].

  In the process for Event2 [] in S103, the arithmetic processing unit 12 executes a process according to the following [Equation 1].

[Formula 1]
EventW [pw]. time ← Event2 [p2]. time + offset2 [h]
EventW [pw]. channel ← 1
EventW [pw]. note-number ← Event2 [p2]. note-number
EventW [pw]. velocity ← Event2 [p2]. velocity
pw ← pw + 1
p2 ← p2 + 1

  Of the above six expressions of [Formula 1], in the first expression, the third expression, and the fourth expression, the time, note-number, and velocity of the event event of Event2 [p2] are set to the time of the note event of EventW [pw]. , Note-number, and velocity. In the second equation, the channel of the note event of EventW [pw] is set to 1. According to these four formulas, the note event existing in the MIDI data M2 is transferred as the channel 1 note event of the integrated MIDI data. Expressions 5 and 6 are pointer additions.

  In the process for Event1 [] in S105, the arithmetic processing unit 12 executes a process according to the following [Equation 2].

[Formula 2]
EventW [pw]. time ← Event1 [p1]. time + offset1 [h]
EventW [pw]. channel ← 0
EventW [pw]. note-number ← Event1 [p1]. note-number
EventW [pw]. velocity ← Event1 [p1]. velocity
pw ← pw + 1
p1 ← p1 + 1

  In the above [Equation 2], p1 is used instead of p2 in [Equation 1], and channel is set to 0. In the above [Expression 2], the note event existing in the MIDI data M1 is transferred as the note event of channel 0 of the integrated MIDI data by the first to fourth expressions. Expressions 5 and 6 are pointer additions.

  After the process for Event1 [] in S105, it is determined whether the pointer p1 is smaller than the total number of note events Ne1 (S106). As a result of the determination, if the pointer p1 is smaller than the total number of note events Ne1, the process returns to S101. On the other hand, if the pointer p1 is equal to or greater than the total number of note events Ne1, this means that the processing of all the note events in the array Event1 [] has been completed, and thus the process proceeds to S103 and the process for Event2 [] is performed.

  After the process for Event2 [] in S103, it is determined whether the pointer p2 is smaller than the total number of note events Ne2 (S104). As a result of the determination, if the pointer p2 is smaller than the total number of note events Ne2, the process returns to S103 and the process for Event2 [] is performed. Conversely, if the pointer p2 is equal to or greater than the total number of note events Ne2, this means that all the note event processing in the array Event2 [] has been completed, so the integration processing ends.

<Duplicate note-on section difference processing>
When the integration process ends, the arithmetic processing unit 12 executes an overlapping note-on section difference process (S2 in FIG. 2). The note-on section (sound generation section) is a section composed of a pair of note-on event data and two note-off event data having the same note number corresponding thereto. Further, the overlapping note-on section is a note-on section that overlaps in part or part with the note-on section of the other channel in time. As a result of the integration process of S1, EventW [] is composed of a 2-channel MIDI event, and a process of converting it into a single-channel MIDI event array EventS [] is performed. However, in the event W [], note-on intervals with the same note number may overlap at the same time. In the overlap note-on interval difference process, the intensity value (velocity) difference calculation is performed for the portion where the note-on interval overlaps, and the result is converted into a single channel MIDI event array EventS [].

  Here, an outline of the overlapping note-on section difference processing will be described with reference to FIG. FIG. 4A shows one 2-channel MIDI data after the integration of two MIDI data, and FIG. 4B shows one-channel MIDI data after overlapping note-on section difference processing. For convenience of explanation, FIG. 4 shows only one note number. In each rectangle, the left end is the note-on event time, the right end is the note-off event time, and the horizontal width indicates the note-on interval. Also, the vertical direction of each rectangle indicates the velocity. In the example of FIG. 4A, all the note-on intervals overlap all or part of the note-on interval of the other channel in time, so that all are overlap note-on intervals. Also, when a part of the note-on section 02 overlaps with the note-on section 12 of the other channel as in the note-on section 02, the overlapping section is referred to as an overlapping section, and the non-overlapping section is referred to as a non-overlapping section. I will decide.

  In the example of FIG. 4A, the note-on section 01 of channel 0 (indicated by ch: 0 in the figure) and the note-on section 11 of channel 1 (indicated by ch: 1 in the figure) overlap. Therefore, the difference calculation process is performed, and a new note-on section is obtained for channel 0 as shown in FIG. The velocity of the new note-on section is the difference between the velocity of the note-on section 01 and the velocity of the note-on section 11. The note-on interval 01 and the note-on interval 11 have the same note-on event time and note-off event time, and the note-on interval completely overlaps. The time of the on event and the time of the note off event are the same as the note on section 01 and the note on section 11.

  When a part overlaps like the note-on section 02 and the note-on section 12, a difference calculation process is performed only about the part which overlaps. Since the note-on section 02 and the note-on section 12 have the same velocity, no note-on section exists in the section where the note-on section 02 and the note-on section 12 overlap. In addition, since there is a portion that does not overlap with the note-on interval of channel 1 in the note-on interval 02, a new note-on interval is obtained in channel 0 for the portion that does not overlap. The velocity of the new note-on section is the same as the velocity of the note-on section 02. Further, the note-on time of the note-on section 12 is given as the note-on time of the note-on section 12, and the note-off time of the note-on section 02 is given as it is.

  In the overlapping note-on section difference process, the arithmetic processing unit 12 sets pw and ps as pointers to the arrays of EventW [] and EventS [], and sets pw = ps = 0 in the initial state. Also, all the Channel values of EventS [] are set to 0. Further, the arithmetic processing unit 12 uses a note-on table that stores a pointer to the MIDI event array at the time of note-on corresponding to the note numbers 0 to 127 as tableW0 [] corresponding to channel 0 and channel 1 of EventW []. , TableW1 [] and tableS [] corresponding to EventS []. Then, all initial values of tableW0 [], tableW1 [], and tableS [] are set to -1. Further, the total number of note events in the generated MIDI event array EventS [] is Ns.

  When the arrangement is ready, the arithmetic processing unit 12 starts overlapping note-on section difference processing. Details of the overlapping note-on section difference processing will be described with reference to the flowchart of FIG. First, after specifying the pointer pw, the arithmetic processing unit 12 executes processing according to the following [Equation 3] to determine the note number nn and the velocity to be processed as vel.

[Formula 3]
nn ← EventW [pw]. note-number
vel ← EventW [pw]. velocity

  And EventW [pw]. It is determined whether channel is 0 or 1 (S201). EventW [pw]. When channel is 1, processing for channel 1 is performed in S202 to S207, and EventW [pw]. When channel is 0, processing for channel 0 is performed in S208 to S213.

  EventW [pw]. When channel is 1, it is first determined whether or not tableW0 [nn] is 0 or more (S202). That is, it is determined whether or not there is an event for which note-on with the same note number exists in channel 0. If tableW0 [nn] is greater than or equal to 0, a note-on event with the same note number exists at least on the channel 0 side. A process of interrupting the note-on section by issuing an off event is performed (S203).

  Specifically, it is determined whether or not tableS [nn] is equal to or greater than 0. If tableS [nn] is equal to or greater than 0, the arithmetic processing unit 12 executes processing according to the following [Equation 4]. To do.

[Formula 4]
EventS [ps]. time ← EventW [pw]. time
EventS [ps]. channel ← 0
EventS [ps]. note-number ← nn
EventS [ps]. Velocity ← 0
tableS [nn] ← −1
ps ← ps + 1

  Of the above six formulas of [Formula 4], in the first and third formulas, eventW [pw] note event time and note-number are given as EventS [ps] note event time and note-number. ing. In the second equation, the channel of the note event of EventS [ps] is set to 0. Since note events in EventS [] are stored in one channel, all are set to channel 0. In the fourth equation, the velocity of the note event of EventS [ps] is set to 0. This indicates a note-off event. By these four expressions, note-off is forcibly set in EventS []. Thus, conceptually, the note-on interval of the same number of the other channel is divided into an overlap interval and a non-overlap interval by a note-on or note-off event of the same note number of the other channel. (At this stage, we don't know how the two channels overlap at that note number in time). The fifth equation is a process for indicating that the note-on interval for the note number nn has ended. The sixth equation is addition of pointers.

  When the processing according to [Formula 4] is executed, the arithmetic processing unit 12 determines whether or not vel is greater than 0 (S204). If tableS [nn] is smaller than 0 in S203, the process proceeds to S204 without executing the process according to [Equation 4]. If it is determined in S204 that vel is greater than 0, it is determined that the two channels overlap in time starting from the note event, and the arithmetic processing unit 12 performs processing according to the following [Equation 5]. Is executed (S205).

[Formula 5]
vel ← | vel-EventW [tableW0 [nn]]. velocity |
tableW1 [nn] ← pw

  Of the two formulas in [Formula 5], in the first formula, the velocity difference from the note event of the same note number of channel 0 is calculated, and the difference is set to vel. The difference calculation process of the overlapping section is performed by the first expression. In addition, when the difference vel as a result is 0, a process of vel ← 1 is performed. The second expression is a process for specifying the note event in which the processed note number enters the note-on section and is turned on.

  If it is determined in S204 that vel = 0, it is determined that the section that has been temporally overlapped in the two channels ends, and the arithmetic processing unit 12 performs processing according to the following [Equation 6]. Is executed (S206).

[Formula 6]
vel ← EventW [tableW0 [nn]]. velocity
tableW1 [nn] ← −1

  If it is determined in step S204 that vel = 0, the note event of channel 1 is already in the note-on state and is considered to be note-off in this event. Give the velocity of the note event of 0 to vel. The second equation is a process for indicating that the note-on interval for the note number nn in channel 1 has ended.

  On the other hand, if tableW0 [nn] is smaller than 0 in S202, it is necessary to perform duplication processing regardless of whether the channel 0 side is in the note-on state and the channel 1 side is in the note-on state or the note-off state with the same note number. Therefore, the process proceeds to tableW1 [nn] (S207). In S207, when vel> 0 and the note is on, the arithmetic processing unit 12 executes a process of setting tableW1 [nn] ← pw, and when vel = 0 and the note is off, tableW1 [nn] ← −1.

  The processing of S202 to S207 is performed by EventW [pw]. This is processing for channel 1 when it is determined that channel is 1. However, as described above, EventW [pw]. If it is determined that channel is 0, processing for channel 0 is performed in S208 to S213. Here, a flowchart of the processing for channel 0 is shown in FIG. The processes in S208 to S213 differ only in the channel to be processed, and are basically the same as the processes in S202 to S207.

  First, in S208, it is determined whether tableW1 [nn] is 0 or more (S202). That is, it is determined whether a note having the same note number is turned on in channel 1. If tableW1 [nn] is greater than or equal to 0, there is at least a note that is turned on with the same note number on the channel 1 side, so if there are overlapping parts in time, Processing is performed (S209).

  Specifically, it is determined whether or not tableS [nn] is equal to or greater than 0. If tableS [nn] is equal to or greater than 0, it is in the note-on interval. 4] is executed to forcibly end the note-on section.

  When the processing according to [Formula 4] is executed, the arithmetic processing unit 12 determines whether or not vel is greater than 0 (S210). If tableS [nn] is smaller than 0 in S209, the process proceeds to S210 without executing the process according to [Equation 4]. If it is determined in S210 that vel is greater than 0, note events with the same note number are overlapped and note-on occurs, so the arithmetic processing unit 12 executes processing according to the following [Equation 7]. (S211).

[Formula 7]
vel ← | vel-EventW [tableW1 [nn]]. velocity |
tableW0 [nn] ← pw

  The above [Formula 7] is obtained by replacing the tableW1 and the tableW0 in the above [Formula 5]. Of the two equations, in the first equation, the velocity difference from the note event of the same note number of channel 1 is calculated, and the difference is set as vel. The difference calculation process of the overlapping section is performed by the first expression. In addition, when the difference vel as a result is 0, the process of vel ← −1 is performed. In the subsequent processing, if vel = 0, it is treated as note-off, and otherwise, it is treated as note-on including a negative value such as vel = -1. The second equation is a process for specifying the note event of the note number that is turned on in channel 0.

  If it is determined in S210 that vel = 0, the arithmetic processing unit 12 executes processing according to the following [Equation 8] (S212).

[Formula 8]
vel ← EventW [tableW1 [nn]]. velocity
tableW0 [nn] ← −1

  The above [Equation 8] is obtained by replacing the tableW1 and the tableW0 in the above [Equation 6]. If it is determined in step S210 that vel = 0, the note event of channel 0 is considered to be a note-off event. Therefore, in the first equation of [Equation 8], the velocity of the note event of channel 1 is directly set to vel. give. The second equation is a process for indicating that the note-on interval for the note number nn in channel 0 has ended.

  On the other hand, if tableW0 [nn] is smaller than 0 in S208, the same note number is note-off on the channel 0 side, so there is no need to perform duplication processing, so processing of tableW0 [nn] (S213). In S213, when vel> 0, the arithmetic processing unit 12 executes a process of setting tableW0 [nn] ← pw, and when vel = 0, setting tableW0 [nn] ← −1.

  When any of the processes of S205, S206, and S207 is performed on channel 1 and when any of the processes of S211, S212, and S213 is performed on channel 0, A note on / off process is executed (S214). In S214, first, if tableS [nn] = − 1 and vel> 0, or tableS [nn]> 0 and vel = 0, the processing according to the following [Equation 7] is executed, and the output array EventS Generate note-on event or note-off event in []. Here, the condition tableS [nn] = − 1 and vel> 0 indicates that the note number nn is in the note-off state, and the note event currently being processed is a note-on event. The other condition, tableS [nn]> 0 and vel = 0, indicates that the note number nn is already in a note-on state and the note event currently being processed is a note-off event. Yes.

[Formula 9]
EventS [ps]. time ← EventW [pw]. time
EventS [ps]. channel ← 0
EventS [ps]. note-number ← nn
EventS [ps]. velocity ← vel
If vel> 0, tableS [nn] ← ps, if vel = 0, tableS [nn] ← −1.
ps ← ps + 1

  Of the six formulas in [Formula 9], the first formula, the second formula, the third formula, and the sixth formula are the same as [Formula 4]. The difference from [Equation 4] is that there is a case of both note-off or note-on depending on the value of vel. Therefore, in the fourth equation, the velocity currently being processed or the equation 5 ] To [Velocity] corrected by [Equation 8]. This is because there is a possibility of note-on or note-off, and the velocity value may be corrected. In the fifth equation, the value set in tableS [nn] differs depending on the value of vel. When vel = 0, -1 is set as in [Formula 4], but when vel> 0, the value of ps is set.

  When the note-on process in S214 is completed, it is determined whether or not all the note events stored in the event W [] on the input side have been processed (S215). Specifically, after setting pw ← pw + 1, if pw <Nw, the processing for all the note events is not completed, so the process returns to S201 to execute the processing for the next note event. In the case of pw ≧ Nw, since all the note events have been processed, the overlap note-on section processing is ended as Ns ← ps.

<0 duration event deletion process>
When the overlapping note-on section processing ends, the arithmetic processing unit 12 performs 0 duration event deletion processing (S3 in FIG. 2). When the processing for the overlapping note-on period is completed, there is no note event in the integrated single-channel MIDI event array EventS [] that has the same note number and whose note-on period overlaps in time. However, EventS [] may include a 0-duration note event in which note-on and note-off are at the same time. Therefore, next, a process of deleting a note event of 0 duration is performed.

  In the 0 duration event deletion process, the arithmetic processing unit 12 prepares a differential MIDI event array to be output after the deletion process as EventD [h,], sets the total number of note events as Nd [h], and manages EventS [], EventD [ The pointers to the respective arrays h and] are ps and pd, and ps = pd = 0 is set in the initial state. Also, all the Channel values of EventD [h,] are set to 0. Further, the computer (encoding device) defines a note-on table tableS [] that stores a pointer to a MIDI event array when a note is turned on, corresponding to a note number from 0 to 127. Then, all initial values of tableS [] are set to -1.

  When the arrangement is ready, the arithmetic processing unit 12 starts the 0 duration event deletion process. Details of the 0 duration event deletion processing will be described with reference to the flowchart of FIG. First, after specifying the pointer ps, the arithmetic processing unit 12 executes processing according to the following [Equation 10], and determines the note number nn and the velocity to be processed as vel.

[Formula 10]
nn ← EventS [ps]. note-number
vel <-EventS [ps]. velocity

  Then, it is determined whether or not vel is greater than 0 (S301). If it is determined that vel is greater than 0, processing according to the following [Equation 11] is executed (S302).

[Formula 11]
tableS [nn] ← ps

  On the other hand, if it is determined in S301 that vel = 0, first, EventS [tableS [nn]]. time and EventS [ps]. It is determined whether or not time matches, and only when both match, processing according to the following [Formula 12] is executed (S303).

[Formula 12]
EventS [tableS [nn]]. note-number ← -1
EventS [ps]. note-number ← -1
tableS [nn] ← −1

  EventS [tableS [nn]]. time and EventS [ps]. If the time matches, it means that the duration is 0, and the note event is deleted. In this embodiment, as shown in the first and second formulas of [Formula 12], a negative value (−1) that is invalid as a note number is set for the note event. In S303, EventS [tableS [nn]]. time and EventS [ps]. If it is determined that the times do not match, the process according to [Formula 12] is not executed.

  If it is determined in S301 that vel <0, it means that “−1” is set to vel in the above [Equation 7]. This is because the velocity difference in the overlapping section is “0”. ”Means it was. In this case, one of the following two processes is executed according to the prior setting. The first is a process of changing the value of vel to a small value in the range of 1 to 127 other than 0. In the present embodiment, vel ← 1. The “small value in the range of 1 to 127 other than 0” does not necessarily have to be “1”, and can be appropriately set according to the characteristics of the acoustic signal recorded in the MIDI data.

  The second is a process of deleting the overlapping section from the encoding target. Specifically, the corresponding note-off event is searched, and the processing according to the above [Equation 12] is executed and invalidated. By this invalidation, it is deleted from the encoding target in a later process.

  Regardless of whether the processing of S302 or S303 is performed, the energy value update processing is executed (S304). Specifically, the process according to the following [Formula 13] is executed.

[Formula 13]
Energy [h] ← Energy [h] + EventS [tableS [nn]]. velocity (EventS [ps] .time-EventS [tableS [nn]]. time)
ps ← ps + 1

  In the first equation of [Equation 13], the energy value is calculated by multiplying the velocity of the note-on interval by the duration, and adding it to the total energy value calculated so far, so that the total energy value Energy [h] Has been updated. When the total energy value Energy [h] is updated, the value of ps is incremented according to the second equation of the above [Equation 13]. Then, it is determined whether or not ps <Ns (S305). If ps <Ns, the process returns to S301. On the other hand, if ps ≧ Ns, it means that all note events have been processed, and the process proceeds to generation of an output event array. After all, in the loop of S301 to S305, the note event whose duration is 0 is deleted, and the process of calculating the total energy value is executed.

  When processing for all note events is executed, deletion of note events with duration 0 and calculation of the total energy value are performed, an output event array is generated (S306). First, ps = pd = 0 is set. Subsequently, EventS [ps]. It is determined whether or not note-number ≧ 0. EventS [ps]. If note-number ≧ 0, it means a note event that has not been invalidated, so the processing according to the following [Equation 14] is executed, and the value in the output event array EventD [h, pd] Is output.

[Formula 14]
EventD [h, pd]. time <-EventS [ps]. time
EventD [h, pd]. channel ← 0
EventD [h, pd]. note-number ← EventS [ps]. note-number
EventD [h, pd]. velocity ← EventS [ps]. velocity
pd ← pd + 1
ps ← ps + 1

  When the output to the output event array EventD [h, pd] for one note event is completed according to the above [Equation 14], it is determined whether or not ps <Ns is satisfied (S307). If ps <Ns, the process returns to S306. On the other hand, if ps ≧ Ns, it means that all note events have been processed for the phase h, and the processing is terminated.

  In the present invention, it is also possible to perform processing that does not consider phase shift. When the phase shift is not taken into consideration, the processing is completed at the end of FIG. 7, and the output event array EventD [] obtained by the processing of S1 to S3 of FIG. 2 is obtained as differential MIDI data. However, when the phase shift is not taken into account, there is a problem that when the phase shift (temporal position shift) between the MIDI data M1 and the MIDI data M2 occurs, the difference between the two becomes larger than the actual difference. Here, FIG. 8 shows the phase shift during the difference processing.

  FIGS. 8A, 8B, and 8C show MIDI data M1, MIDI data M2, and differential MIDI data, respectively. In both cases, the horizontal direction indicates the time axis. The vertical direction indicates the pitch (note number), but the vertical width of each rectangle indicates the sound intensity (velocity). A shaded portion indicates a note-on interval. As can be seen from a comparison between FIGS. 8A and 8B, the MIDI data M1 and M2 are the same MIDI data with the position shifted in the time axis direction.

  When the difference calculation between the MIDI data M1 shown in FIG. 8A and the MIDI data M2 shown in FIG. 8B is performed, the difference MIDI data as shown in FIG. 8C is obtained. A shaded portion in FIG. 8C is a portion where the velocity difference is not zero. In this way, even if the MIDI data is essentially the same, if there is a positional shift, a pseudo difference as shown in FIG. 8C is generated. In the present embodiment, in order to eliminate such a pseudo difference, the time is adjusted by moving one of the MIDI data M1 and M2 in the time axis direction so as to obtain the difference. In reality, it is impossible to perfectly match the time, so the difference calculation is performed while changing the time to move, and when the difference is minimized, the time position is considered to be correct. , Output the difference.

  When phase shift is taken into account, when the 0 duration event deletion process of FIG. 7 or the process of S3 of FIG. 2 is completed, whether N ≧ [h] ← pd, h ← h + 1 in FIG. Is determined (S4). That is, it is determined whether or not the processing of S1 to S3 has been completed for all phases. In the case of h <H, since the unprocessed phase remains, the processes of S1 to S3 are repeated. If h ≧ H, the processing has been completed for all phases, so differential MIDI data is selected (S5).

  Specifically, the difference for obtaining the MIDI event array EventD [hm,], where h is the minimum energy value Energy [h] of note events calculated at h = 0,..., H−1. It is adopted as MIDI data and is output as a MIDI code.

  The differential MIDI data can be output in various formats. It can be converted into display data and output on the screen to the display unit 13, or it can be converted into an acoustic signal and output to the acoustic output unit 14 as it is, or data on a storage medium or the like as it is. It is also possible to output data to the output unit 15. When the screen is output to the display unit 13, note-on and note-off are converted into note-on intervals to determine the left and right widths, the vertical width corresponding to the velocity is determined to generate a rectangle, and the corresponding note numbers are generated. Display data is generated and output to the display unit 13. That is, it is displayed in a format as shown in FIG. In the case of simultaneous output to the display unit 13 and the sound output unit 14, two types of sound signals can be visually and audibly evaluated.

  The present invention can be used in an industry in which some modification is made to an acoustic signal such as development of various encoding techniques and development of a digital watermark embedding technique, and the degree of signal degradation is evaluated. In addition, the development results are based on the technology that converts sound signals obtained by PCM into code codes such as MIDI codes, broadcast media (digital radio and television broadcasts such as terrestrial and BS), and communication media (CS Broadcast, Internet streaming broadcast, mobile phone service, mobile music distribution service, etc.) and audio media production industry for package media (CD, DVD, BlueRay, memory IC card, etc.).

DESCRIPTION OF SYMBOLS 11 Data input part 12 Arithmetic processing part 13 Display part 14 Sound output part 15 Data output part

Claims (16)

A method of creating one differential code data representing a difference between two independent code data obtained by encoding an acoustic signal,
The code data is composed of event data arranged in chronological order, and each event data includes identification of an event at the start or end of pronunciation , time of the event , pitch, channel number, and sound intensity. With attribute information,
While applying different channel numbers for each event data of the two code data, the time is sorted in ascending order, into one time-series code data with two channel numbers, 2 channel code A code data integration stage for creating data;
When a section composed of two event data pairs of event data at the start of sound generation and event data at the end of sound generation having the same pitch as the event data is defined as the sound generation section, the two-channel code data In the case where the sounding section of one channel overlaps with the sounding section of the other channel that has the same pitch, all or part of the sounding section overlaps in time, For a section in which a note-on section exists only in one of the overlapping sound generation sections, with a new sound generation section having a new sound intensity as a difference value of the sound intensity of the section as a new section. Is an overlapping pronunciation interval difference processing stage for generating a pronunciation interval having the sound intensity of the one channel ;
Sorting the event data of the pronunciation start and the event data of the pronunciation end of each of the pronunciation sections updated in the overlapping pronunciation section difference processing step in time series,
Differential code data encoding stage for encoding in a predetermined format;
A method for creating differential code data, comprising: In claim 1,
The difference pronunciation data characterized in that, in the overlapping pronunciation section difference processing step, when the sound intensity of the overlapping section is 0, the sound intensity is changed to a small value within a range of 1 to 127 other than 0. How to create In claim 1,
The method of creating differential code data, wherein, in the overlapping pronunciation section difference processing step, when the sound intensity of the overlapping section is 0, the overlapping section is deleted from the encoding target. In any one of Claims 1-3,
In the code data integration step, a plurality of offset values are uniformly added to the time of the event data of the start and end of sound generation of any one of the two code data,
Based on the plurality of offset values, a plurality of steps until the differential code data encoding step is performed to encode a plurality of types of differential code data,
For each of the plurality of differential code data encoded in the differential code data encoding step, a pair of event data for starting the sound and two event data for the end of the sound having the same pitch as the corresponding sound data. Is extracted, and the sound intensity of all of the extracted sounding intervals and the time interval of the sounding interval (the difference between the time of event data at the end of sounding and the time of event data at the start of sounding) A differential code data creation method comprising: after the differential code data encoding step, a differential code data selection step of calculating a sum value of products and selecting differential code data having a minimum total value. In any one of Claims 1-4,
The difference code data encoding step deletes event data constituting the sounding section when there is a sounding section obtained by the overlapping sounding section difference processing step having the same sounding start time and sounding end time. A method of creating differential code data, characterized by further performing a process of In any one of Claims 1-5,
The differential code data further comprising the step of displaying the differential code data obtained in the differential code data encoding step on a screen on a display means, and / or the step of acoustically reproducing the differential code data by an acoustic reproduction means. How to create In any one of Claims 1-6,
The code data is MIDI data, the start of sound identification is note-on, the end of sound generation is note-off, the time is delta time, the pitch is a note number, and the sound intensity is velocity. The method of generating differential code data, wherein the sounding section is expressed as a note-on section. An apparatus for creating one differential code data representing a difference between two independent code data obtained by encoding an acoustic signal,
The code data is composed of event data arranged in chronological order, and each event data includes identification of an event at the start or end of pronunciation , time of the event , pitch, channel number, and sound intensity. With attribute information,
While applying different channel numbers for each event data of the two code data, the time is sorted in ascending order, into one time-series code data with two channel numbers, 2 channel code Code data integration means for creating data;
When a section composed of two event data pairs of event data at the start of sound generation and event data at the end of sound generation having the same pitch as the event data is defined as the sound generation section, the two-channel code data In the case where the sounding section of one channel overlaps with the sounding section of the other channel that has the same pitch, all or part of the sounding section overlaps in time, For a section in which a note-on section exists only in one of the overlapping sound generation sections, with a new sound generation section having a new sound intensity as a difference value of the sound intensity of the section as a new section. Is an overlapping pronunciation interval difference processing means for generating a pronunciation interval having the sound intensity of the one channel ,
Sorting the event data of the pronunciation start and the event data of the pronunciation end of each of the pronunciation intervals updated by the overlapping pronunciation interval difference processing means in time series, and differential code data encoding means for encoding in a predetermined format;
An apparatus for creating differential code data, comprising: In claim 8,
The differential pronunciation section difference processing means changes the sound intensity to a small value within a range of 1 to 127 other than 0 when the sound intensity of the overlap section is 0. Creation device. In claim 8,
The apparatus for creating differential code data, wherein the overlapping pronunciation section difference processing means deletes the overlapping section from the encoding target when the sound intensity of the overlapping section is 0. In any one of Claims 8 to 10,
The code data integration means uniformly adds a plurality of offset values with respect to the time of the event data of the start and end of sound generation of either one of the two code data,
Based on the plurality of offset values, the differential code data encoding means is executed a plurality of times to encode a plurality of types of differential code data,
For each of a plurality of types of differential code data encoded by the differential code data encoding means, a pair of event data at the start of sound generation and two event data at the end of sound generation having the same pitch corresponding thereto. Is extracted, and the sound intensity of all of the extracted sounding intervals and the time interval of the sounding interval (the difference between the time of event data at the end of sounding and the time of event data at the start of sounding) An apparatus for creating differential code data, further comprising differential code data selection means for calculating a total value of products and selecting differential code data having a minimum total value. In any one of Claims 8-11,
The difference code data encoding means deletes event data constituting the sounding section when the sounding section obtained by the overlapping sounding section difference processing means has the same sounding start time and sounding end time. An apparatus for creating differential code data, characterized by further performing a process to perform the process. In any one of Claims 8-12,
An apparatus for creating differential code data, further comprising: display means for displaying differential code data obtained by the differential code data encoding means on a screen; and / or sound reproducing means for acoustically reproducing the differential code data. In any one of Claims 8 to 13,
The code data is MIDI data, the start of sound identification is note-on, the end of sound generation is note-off, the time is delta time, the pitch is a note number, and the sound intensity is velocity. The sound generation section is expressed as a note-on section.   A program for causing a computer to execute the differential code data creation method according to any one of claims 1 to 7.   15. A program for causing a computer to function as the differential code data creating apparatus according to claim 8.

Images