JPH0352061B2 - - Google Patents

Abstract

PURPOSE:To extract only a part that a student plays with confidence from an input tone series by comparing tone series patterns of converted tone series data and tone series data corresponding to a reference tone series with each other as to specific features. CONSTITUTION:Song information, etc., is converted into tone series data having a dimension of code length, the toner series patterns of the tone series data and tone series data corresponding to the reference tone series are compared with each other as to the specific features, and a phrase which is most similar to the reference tone series is extracted from the input tone series according to the comparison result. Namely, a similar pitch tone series extracting circuit 7 while shifting reference pitch tone series data PDline(ref) and input pitch tone series data PDline(full) in time base reference point compares tone series parts of the same length with the reference tone series after conversion into a tone series pattern of elements connected in time series in the same time zone, thereby extracting tone series parts according to similarity to the reference pitch tone series. Consequently, the algorithm of grading is approximated to the algorithm by a music tutor as much as possible.

Description

[Detailed description of the invention]

This invention mainly focuses on the performance of a performer or a singer (hereinafter referred to as a performer, etc.) from among a series of notes played or sung (hereinafter referred to as a performance, etc.), which the performer, etc. is confident is correct. The present invention relates to a method for extracting only a part of a harmonic tone string, and a performance result display device and a performance result scoring device using this method. As is well known, when teaching singing practice or piano or electronic organ practice, the teacher generally first performs a model performance, then has the students perform, etc., and then reviews the results of the student's performance etc. The teacher then provides musical criticism, and by repeating the above process, the teacher provides performance guidance. However, in the case of this type of performance instruction, the student's progress is greatly influenced by the qualifications and abilities of the teacher, and since the number of excellent teachers is limited, the number of students is also limited to a relatively small number. There is. In order to solve these conventional problems, various devices have been proposed that automatically score performances performed by students instead of teachers. By the way, when a student performs a model performance or plays according to the musical score, if he or she realizes that he or she has played incorrectly during the performance, the student usually returns to the point just before the part where he or she made the mistake. It is known from experience that if a player does not notice a mistake during the performance, he or she will continue playing until the end. On the other hand, when a performance is performed that includes such errors, the teacher may unconsciously recognize the portions of the student's incorrect performance that have already been replayed correctly by the student. The system extracts from the entire performance only the parts that the student is confident played correctly, and then compares these extracted parts with the model performance to provide the best advice to the student. be. In other words, it is not possible to improve the teaching effect in any way by pointing out parts where students have already noticed mistakes in their performance. This means working to effectively improve technology. However, in conventional automatic performance scoring devices of this type, the pitches or the arrangement order of notes for a model performance are stored in advance, and then input from a microphone or keyboard according to the student's performance. The input order of pitches and notes is successively compared with the memorized note array and pitch array, or similarly, each note constituting the note array is compared in the previously stored note array and pitch array. The pitches are sequentially read out based on the length, and it is determined whether or not the timing of each read pitch matches the timing of each input pitch within a predetermined tolerance range. This is used to determine whether the melody played is correct or incorrect, so if there is a part in the middle of a performance performed by a student where a student notices a performance error and performs it again, Even if all the parts of the performance were performed correctly, the scoring result would be extremely low, and there would be problems such as not being able to provide students with the optimal criticism needed to improve their performance skills. This may result in students losing their desire to practice. This invention was made in order to solve the above problem, and its main purpose is to make the algorithm for grading as much as possible by the music teacher (human) in this type of automatic grading device. It is about approximation. The purpose of the first invention related to this application is that, as mentioned above, if there is a part of the student's performance that has been replayed after noticing an error, such erroneous part can be automatically corrected. The purpose of the present invention is to provide a method for extracting from an input tone sequence only the parts played by the student with confidence that they are correct. In order to achieve the above object, the first invention provides performance or singing information in the form of 1 to 2 notes having dimensions of note length.
The sound sequence data is converted into sound sequence data with a dimension higher than that, and then the sound sequence data and the sound sequence data corresponding to the reference sound sequence are compared for predetermined characteristics between sound sequence patterns, and based on the comparison result, the sound sequence data that corresponds to the reference sound sequence is This system is characterized in that similar phrases are extracted from the input sound string. Next, the purpose of the second invention related to this application is to compare the similar phrases extracted by the above-mentioned first invention with the reference sound sequence so that the student can be sure that the phrase is correct. The performance I did,
Comparison is made with a pre-memorized model tone sequence, and from this, among the parts of each incorrect performance included in the student's performance, parts that the student has already recognized are deleted, and the part that the student has already recognized is deleted. The goal is to make the students aware of only the parts of the performance that were incorrect. In order to achieve the above object, the device of the second invention sequentially converts each constituent sound sequentially generated by playing or singing into one- or two-dimensional or more single note data having the dimension of note length. Single note data detection means for converting and detecting; Input note string data forming means for storing the detected single note data for each dimension and in the order of occurrence to form input note string data corresponding to performance or singing; The input sound sequence data of a dimension and the reference sound sequence data of the corresponding dimension are compared between the sound sequence parts that exist in the same time period while shifting the reference points of the time axes of both, and the sound sequence data is compared in descending order of similarity. Similar sound string part extracting means for extracting one or more selected sets of similar sound string parts from the input sound string data; and standard sound string data of a corresponding dimension from each of the extracted similar sound string parts. a data superimposition means for extracting matching portions of and superimposing them on each other to form the most similar sound string data corresponding to the most similar phrase; means for printing or displaying the formed most similar sound string data; It is characterized by comprising the following. Next, the purpose of the third invention related to this application is to compare the similar phrase obtained using the first invention with a model performance or an original performance performed by a student. By doing this, even if there is a replayed part in the middle of a student's performance, the scoring results will not be extremely low, and this will prevent the actual music teacher (human) The goal is to obtain a scoring result that is as close as possible to the actual score. In order to achieve the above object, in the third invention, as a component thereof, each constituent sound sequentially generated by playing or singing is sequentially converted into one or two or more dimensional single note data having a dimension of note length. an input note string data forming means that stores the detected single note data for each dimension and in the order of occurrence to form input note string data corresponding to performance or singing; Compare the input sound sequence data and the reference sound sequence data of the corresponding dimension by shifting the reference points of the time axes of both, and select the parts of the sound sequence that exist in the same time period, and select the ones with the highest degree of similarity. Similar sound string part extracting means for extracting one or more sets of similar sound string parts from the input sound string data; data superimposition means for extracting matching parts and superimposing them on each other to form the most similar sound sequence data corresponding to the most similar phrase; Find the ratio between the total number of sound data included in the most similar sound sequence data and the total number of each component sound data of the reference sound sequence data for the necessary dimensions, and score the input sound sequence using the value of these ratios as the scoring factor. The present invention is characterized by comprising a scoring calculation means. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. FIGS. 1 to 37 are drawings for explaining an embodiment (hereinafter referred to as the first embodiment) that also incorporates the first to third aspects of the invention. In the following, melody performance will be described as a representative example, but the same applies to other performance modes such as rhythm performance and chord performance. First, before proceeding to a detailed explanation of the operation of the apparatus shown in the first embodiment, the basic operation flow of this apparatus will be schematically explained with reference to the overall diagram shown in FIG. In FIG. 1, when a start command (not shown) is given, the reference tone pitch string generation circuit 1 outputs each pitch data PD (ref) (see FIG. 25B) constituting the reference melody (see FIG. 25A). ) are output in a time-sharing manner in synchronization with the sound timing of each note,
Upon receiving this data PD (ref), the tone forming circuit 2
is driven, and the standard melody is sounded from the speaker 3. Next, the student memorizes this pronounced standard melody in his/her head and plays the standard melody using the keyboard 4, relying on his/her memory. The key press detection circuit 5 outputs a key press timing signal Skon' (see FIG. 26B), which becomes a minute width "1" pulse every time any key is pressed on the keyboard 4.
is output, and at the same time, pitch data PD (in) corresponding to each pressed key on the keyboard 4 (second
6A) are output, and these signals are supplied to the pitch string forming circuit 6. The pitch string formation circuit 6 discriminates the length of each note based on the signal Skon' for each original pitch data (see FIG. 27A) that is sequentially supplied in a time-divisional manner, and distinguishes the length of each note as a musical element. Pitch data that is less than the minute note length that cannot be obtained is regarded as jitter and is removed. Then, the remaining pitch data is stored in the order of occurrence, and the input pitch string data PDline
(full) (see FIG. 27B). Then,
This data PDline (full) is supplied in parallel to a similar pitch sequence extraction circuit 7 and a matching pitch partial superimposition circuit 8. The similar tone high pitch string extraction circuit 7 uses the reference tone high pitch string data output from the reference tone high pitch string generating circuit 1.
PDline (ref) and the input pitch string data PDline (full) output from the input pitch string formation circuit 6 are made identical while shifting the reference points of their time axes, just as if trains were passing each other. After converting to a tone sequence pattern in which each constituent tone is connected in chronological order that exists in a time period, the tone sequence parts of the same length as the reference tone string are compared, and within these tone sequence parts, the reference tone high pitch tone string is A maximum of k sets of items with high similarity to are extracted in order. (See Figures 27 to 30) Then, these extracted maximum k sets of similar pitch string partial data are used as the input pitch string data.
Shift number data PDline sml-1 to -k (27th
(See Figures C to J) and output. At the same time, the similar pitch string extraction circuit 7 extracts matching pitch number data PDeq-1 to -k between each of the extracted similar pitch string partial data and the reference pitch string data PDline (ref). Output. The matching pitch partial superimposition circuit 8 uses the shift number data output from the similar pitch string extraction circuit 7.
PDline sml-1 to -k and input pitch string data output from the input pitch string forming circuit 6
Based on PDline (full), each similar pitch string partial data is played back, and each reproduced similar pitch pitch string partial data is compared with the reference pitch string data PDline (ref) to find a match between the two. The pitch data is superimposed one after another, thereby creating the most similar pitch sequence data PDsample.
form. (See FIG. 35) Next, the input note length string forming circuit 9 uses the signal output from the input pitch string forming circuit 6.
Skon'' (see Figure 26C), form note length data corresponding to each constituent note of the input pitch sequence data PDline (full), and store this data sequentially in the order of occurrence. Long string data
Forms LDline (full). (See FIG. 31B) The similar note length note string extraction circuit 10 uses the reference note length note string data output from the reference note length note string generation circuit 12.
LDline (ref) (see FIG. 25C) and the input note length note string data (see FIG. 31B) output from the input note length string forming circuit 9 are mutually combined in the same way as in the case of the pitch. The reference point of the time axis is shifted and the parts of the sound sequence that exist in the same time period are compared, and the parts with the highest degree of similarity are determined in descending order of similarity. (Refer to Figures 32 to 34) Then, each similar note length note string partial data is converted into input note length note string data in the same manner as in the case of the pitch mentioned above.
Shift count data LDline for LDline (full)
Convert to sml-1 to -k and output. At the same time, the similar note long note string extraction circuit 10 outputs each similar note long note string partial data and reference note long note string data.
Matching note length note number data LDeq-1 to -k representing the number of matching note length notes with LDline (ref) is output. The matching note length partial superimposition circuit 11 uses the shift count data LDline sml-1 to -k output from the similar note length note string extraction circuit 10 and the input note output from the input note length note string forming circuit LDline (full). Based on the long note string data LDline (full), each similar note long note string partial data is played back, and the reproduced similar note long note string partial data is sequentially converted into the reference note long note string data.
By comparing with LDline (ref), a matching note portion between the two is found, and by superimposing them on each other, most similar note long note string data LDsample is formed.
(See FIG. 36) Next, the most similar pitch tone string data PDsample output from the matching pitch partial superimposition circuit 8 and the most similar note length note string data LDsample output from the matching note length partial superimposition circuit 11. means display circuit 1
Based on these data, the results of the student's performance are displayed on the screen of the CRT display that constitutes the display circuit 13, as shown in FIG. Here, on the display screen, a staff notation, a pitch string, and a note length string are displayed in three rows above and below, and high-intensity flashing is displayed for incorrectly played parts in each row. It is done. Therefore, by looking at this display, the student can first check on the staff notation display where he or she unconsciously made a mistake in the melody performance, and then check the middle and lower pitch notes. By checking the display and note length note string display, it becomes possible to clearly confirm whether the incorrect performance is in pitch or note length. Next, the individual pitch matching discrimination circuit 14 uses the reference pitch string data PDline (ref) and the most similar pitch string data.
Based on the PDsample, it is determined whether each pitch constituting the reference pitch string data matches the reference melody, and the determination result is expressed as a 1-bit signal for each pitch individually. Pitch matching discrimination data
Output PDeq−bit. The matching pitch number detection circuit 15 determines the matching pitch number N for the reference pitch string data based on the data PDeq-bit output from the individual pitch matching discrimination circuit 14.
1, and outputs matching sound number data D(N1). The individual note length match discrimination circuit 16 determines the match between both notes for each note based on the reference note length note string data LDline (ref) and the most similar note length note string data LDsample, and determines the match for each note. Individual code length match discrimination data LDeq- that represents the judgment result as a 1-bit signal
Output bit. The matching note length number detection circuit 17 calculates the number N2 of matching notes between the reference note length note string data and the most similar note length note string data LDsample based on the individual note length matching discrimination data LDeq, and calculates this number of matching notes. Output the data D(N2). The similar pitch sequence number detection circuit 18 calculates the number N3 of similar pitch sequence partial data based on each matching pitch number data PDeq-1 to -k output from the similar pitch sequence extraction circuit 7. and outputs data D(N3) representing this number N3. The similar note length note string number detection circuit 19 calculates the number N4 of sets of similar note length note string partial data based on each matching note length note number data LDeq-1 to -k output from the similar note length note string extraction circuit 10. Find this number of pairs N4
Outputs data D (N4) representing . The reference tone pitch number detection circuit 20 calculates the number N5 of constituent notes of the reference tone treble string data based on the reference tone treble string data PDline (ref) output from the reference tone treble string data generation circuit 1, and calculates this number N5. Data D(N5) representing N5 is output. The reference note length number detection circuit 21 calculates the number N6 of constituent notes of the reference note length note string data based on the reference note length note string data LDline (ref) output from the reference note length note string generation circuit 12, and calculates this number N6. Outputs data D (N6) representing . The performance time/time difference detection circuit 22 uses the individual pitch coincidence discrimination data PDeq-bit output from the individual pitch coincidence discrimination circuit 14 and the most similar note length note string output from the coincidence note length partial superposition circuit 11. Data LDsample and the reference note length tone string generation circuit 12
Standard note length string data LDline (ref) output from
and the individual note length coincidence discrimination data LDeq-bit outputted from the individual note length coincidence discrimination circuit 16, the model performance time T and the performance error time ΔT are determined, and the data D(T), T representing these are calculated. (ΔT) is output. Next, the similarity calculation circuit 23 calculates the matching pitch number data D(N1) and the matching note length number data D(N1).
2), Pitch similar note string set number data D (N3), Note length similar note string set number data D (N4), Reference pitch number data D
(N5), standard note length data D (N6), model performance time data D (T), and performance time difference data D (ΔT)
Based on, calculate the similarity score data Dscore,
This is displayed on the score display 24. The calculation formula for determining the similarity score X based on each of the above data is as follows. X={[100×(N1/N5)×(N2/N6)×(T−|ΔT|)/T]−Y}÷5×5 Here, N1; Number of matching pitches N2; Number of matching note lengths N3; Number of note strings with similar pitches N4; Number of note strings with similar note lengths N5; Reference number of pitches N6; Reference number of note lengths T; Model performance time ΔT; Performance time difference Y; N3 or N4 In this way, this similarity In the degree score calculation method, the main scoring elements for calculating the similarity score are the number of matching pitches N1 and the number of matching note lengths N2. Even if there are errors that are replayed during the performance, these will not greatly affect the final score.
It is possible to obtain grading results that are extremely similar to the feelings of a music teacher. Next, the basic operation flow of the present apparatus described above will be explained in detail with reference to the drawings from FIG. 2 onwards. Since the operation of this device is controlled by various control signals output from the control circuit 25, first of all, the detailed configuration of the control circuit 25 is shown in FIGS.
This will be explained based on FIG. 23. As shown in FIG. 19, the control circuit 25 includes first to k-th similar pitch sequence detection circuits, which will be described later, or first to
Similar step designation signal Ssml-1 to - for alternatively designating any one of the k-th similar note long tone sequence detection circuits
Similar stage designation signal generation circuit 2 that outputs (n+1)
510, a shift signal generation circuit 2520 that outputs a shift signal Sshift for stepwise control of the shift register in the specified similar note pitch sequence detection circuit and similar note length note sequence detection circuit, and the specified similar note sequence detection circuit 2520; a latch signal generation circuit 2530 that outputs a latch signal Slatch for latching input pitch string data and input note length string data to the shift registers in the high note string detection circuit and the similar note length string detection circuit; Stage designation signal generation circuit 251
Each similar stage designation signal Ssml−1 to be output from 0
-(n+1) load signal generation circuit 25 that outputs a minute width "1" pulse in response to each rising edge
40, and the shift signal Sshift output from the shift signal generation circuit 2520 is applied to the similar pitch sequence detection circuit and the similar note length sequence detection circuit of the similar stage designated by the similar stage designation signal generation circuit 2510. a selector 2550 for switching and supplying to each shift register; and the latch signal generating circuit 253.
a selector 256 for switching and supplying the latch signal Slatch output from 0 to the similar pitch sequence detection circuit and similar note length sequence detection circuit of each similar stage designated by the similar stage designation signal generation circuit 2510;
0. The details of the similar stage designation signal generation circuit 2510 are explained in the second section.
Shown in Figure 0. As shown in the same figure, the similar stage designation signal generation circuit 2510 includes the input pitch string formation circuit
The decoder 2515, which will be described later, is set by the determination enable signal Sjudge output from the decoder 2515.
RS reset by the n+2 bit output of
The flip-flop 2511 and the latch signal Slatch output from the latch signal generation circuit 2530
D is configured to be forcibly reset by the AND gate 2513 and to receive the output in response to the rise of the output "1" of the AND gate 2513, which will be described later.
type flip-flop 2512, an AND gate 2513 whose opening and closing are controlled by the Q output of the RS flip-flop 2511, and the output of the D-type flip-flop 2512, and which passes the clock signal Sφ output from the shift signal generation circuit 2520. , a counter 2514 that counts the pulses output from this AND gate 2513 and is reset by the n+2 bit output of a decoder 2515, which will be described later, and a decoder 2 that decodes the counting output of this counter 2514.
515, and this decoder 25
Each of the 15 bit outputs is the load signal Sload.
-1 to -(n+1). Details of shift signal generation circuit 2520 are shown in FIG. As shown in the figure, shift signal generation circuit 2
520 is the similar stage designation signal generation circuit 2510
an OR gate 2521 for taking the logical sum of the outputs of each bit of the decoder 2515 in the circuit, and a monomulti 252 for outputting a minute width "1" pulse in response to the rise of the output "1" of this OR gate 2521.
2, a delay circuit 2523 for delaying the minute width "1" pulse output from this monomulti 2522 by a minute time dt, and this delay circuit 2523.
The RS flip-flop 2524 is set by the "1" pulse output from the latch signal generation circuit 2530 and reset by the latch signal Slatch output from the latch signal generation circuit 2530, and the RS flip-flop 2524 is opened and closed by the Q output of the RS flip-flop 2524. and an AND gate 2526 for controlling the opening and closing of the clock signal Sφ output from the clock generator 2525, and the output of the AND gate 2526 becomes the shift signal Sshift. Details of the latch signal generation circuit 2530 are shown in FIG. As shown in the figure, the latch signal generation circuit 25
30 counts the shift signal Sshift output from the shift signal generation circuit 2520, and
A shift count counter 253 is reset by the coincidence output Seq of the coincidence determination circuit 2533, which will be described later.
1 and a key press number counter that counts the sure key press signal Spush outputted from the input pitch sequence forming circuit 6 and is reset by the determination end signal Send output from the similar stage designation signal generation circuit 2510. 2532 and the shift number counter 2
531 and the key press number counter 2532
A coincidence determination circuit 253 that determines coincidence with the count value of
3, and this match determination circuit 25
This means that the coincidence signal of No. 33 is output as the latch signal Slatch. As a result, when each circuit explained above operates,
As shown in the time chart of FIG. 23, when a “1” pulse arrives at the judgment enable signal Sjudge, one “1” pulse is first output to the load signal Sload-1, and then a delay determined by the delay circuit 2523 Shift signal Sshift delayed by time dt
−1, the number of “1” pulses corresponding to the number of key presses is output, and the load signal Sload−1 is output in the same manner.
Similar pulses are output to Sshift-2 to -4 and shift signals Sshift-2 to -4, which are the final shift signals.
After a predetermined number of pulse trains are output to Sshift-4, when the time required for a certain degree of similarity calculation has elapsed, a "1" pulse is finally output to the determination end signal Send. In addition, in the time chart of Figure 23,
This shows the case where n=3 in FIG. 19. Next, referring to each control signal explained above,
The details of each circuit shown in FIG. 1 will be explained along the flow of data processing. First, the details of the input pitch series forming circuit 6 are shown in FIG. In the same figure, RS flip-flop 60
1 is a determination end signal Send which is set by a minute width "1" pulse output from a differentiation circuit 609, which will be described later, and which is output from the control circuit 25.
It is reset by . The AND gate 602 is controlled to open and close by the output of the RS flip-flop 601, and the key press timing signal output from the key press detection circuit 5.
Pass Skon′. The gate circuit 603 is an RS flip-flop 6
The input pitch data is controlled to open and close by the output of 01, and is thereby output from the key press detection circuit 5.
Pass PD (in). A coincidence determination circuit 604 determines whether input pitch data PD (in) stored in the first stage of a shift register 607 (described later) matches input pitch data PD (in) output from the gate circuit 603. If it is determined that the two match, the
“1” is output to the EQ output. The or gate 605 is the AND gate 602
and the output of the coincidence determination circuit 604, and the output of this OR gate 605 becomes the reliable key press signal Spush. The mono multi 606 outputs a minute width "1" pulse in response to the rising edge of "1" output from the OR gate 605, and the output of this mono multi 606 is an inversion of the pitch data acquisition signal.
The shift register 607 is the monomulti 606.
In response to the inverted data acquisition signal Skon'' output from the gate circuit 603, the input pitch data passed through the gate circuit 603 is input into the first stage, and each stage is sequentially shifted to the right in the figure. The mono multi 608 is configured to be repeatedly triggered in response to the rising edge of the "1" output of the mono multi 606, and each time it is triggered, it generates a relatively long "1" pulse (for example, 2.5 seconds). ), and this mono multi 6
The output of 08 becomes the playing signal Splay. The differentiating circuit 609 is configured to output a minute width "1" pulse in response to the rising edge of the output "1" of the monomulti 608, and the output of this differentiating circuit 609 is a judgment enable signal Sjudge.
becomes. The OR gate 610 takes the logical sum of the determination end signal Send and the initial clear signal Sic.
The output of the OR gate 610 clears all the stages of the shift register 607, and at the same time, the output of the OR gate 610 is output as a clear signal Sclr. As a result, when each circuit explained above operates,
As shown in the time chart of FIG. 26, the monomulti 606 is repeatedly triggered at the rising edge of the key press signal Spush output from the OR gate 605.
In addition, the shift register 607 takes in the input pitch data output from the gate circuit 603 in response to the rise of the data acquisition signal inverted Skon'' output from the monomulti 606, so the shift register 607 is configured as shown in FIGS. 26A and 27A. As shown, if there is erroneous pitch data due to jitter etc. in the input original pitch data, these input pitch data will not be taken into the shift register 607. As shown in FIG. 27B, each stage of the shift register 607 removes erroneous input pitch data, and only receives pitch data corresponding to note lengths that are long enough to be used as musical elements. Then, the pitch data taken into each stage is output in parallel, and this parallel output becomes the input pitch string data PDline (full).Next, the input note length string forming circuit 9 The details are shown in Fig. 3. In the figure, an OR gate 901 outputs the logical sum of the determination end signal Send and the initial clear signal Sic.
Outputs the logical sum of Sic and the determination enable signal Sjudge. The differentiating circuit 903 outputs a very small width "1" pulse in response to the rising edge of the data acquisition signal Skon'' output from the pitch string forming circuit 6. The RS flip-flop 905 is controlled to open and close, and passes the minute width "1" pulse output from the differentiating circuit 903.The RS flip-flop 905 is set at the rising edge of the data acquisition signal Skon'' and reset at the output of the OR gate 902. Tempo clock oscillator 906 outputs a tempo clock with a predetermined period (eg, 100 ms, 500 μs). In this example, the frequency is configured to be variably controllable. Counter 907 is enabled by the Q output of RS flip-flop 905 and counts the tempo clock TCL output from tempo clock oscillator 906. Furthermore, shift register 9
It is repeatedly reset by the minute width "1" output of AND gate 904 delayed through 08. The shift register 908 includes the AND gate 9
The counter 907 is shifted and controlled by the minute width "1" pulse output from
The count output is taken into the first stage as note length data, and this shift register 908 is cleared by the output of the OR gate 901. As a result, when each circuit explained above operates,
Each stage of the shift register 908 includes a 31st
As shown in FIG. B, the note length data corresponding to the pitch data stored in each stage of the shift register 607 in the input pitch string forming circuit 6 is sequentially taken in, and the parallel outputs of these stages are The data becomes input note length string data LDline (full). Next, the details of the similar pitch string extracting circuit 7 are shown in FIG. As shown in the figure, the similar pitch string extraction circuit 7 receives input pitch string data PDline (full), respectively.
and reference pitch string data PDline (ref) are configured to be supplied in parallel.
similar pitch sequence detection circuits 700-1 to 700-
It is composed of k. 1st to k-th similar pitch tone string detection circuits 700-1 to 700-1.
−k is the input pitch string data as described above.
PDline (full) and reference pitch sequence data PDline (ref)
If we judge the degree of similarity between objects that exist in the same time zone while shifting the reference points of their time axes, as if trains were passing each other, we can find the 1st, 2nd to kth similar high-pitched tones. As shown in FIGS. 27C to 27J, each of these detected similar pitch string parts has a first value representing the number of shifts for the input pitch string data.
~Kth similar pitch pitch string partial data PDline sml-1
It is output as ~-k. In addition, each similar pitch tone sequence detection circuit 700-1 to 700-7
From 00-k, matching pitch number data PDeq-1 indicates how many notes match the reference pitch string data in each detected similar pitch string partial data PDline sml-1 to -k. ~-k is output. Next, the details of the first similar pitch string detection circuit 700-1 are shown in FIG. In the figure, the shift register 701 loads the input pitch string data PDline (full) output from the input pitch string forming circuit 6 in response to the rise of the load signal Sload-1. Further, in response to the "1" pulse included in the shift signal Sshift-1, the shift control is performed to the left in the figure, and the contents of all stages are simultaneously cleared by the clear signal Sclr. The parallel pitch coincidence determination circuit 702 uses pitch string partial data PDline (1 to 7) output in parallel from the first to seventh stages of the shift register 701.
and the reference tone pitch string data PDline (ref) outputted from the reference tone pitch string generation circuit 1, a match is determined for each stage, and "1" and "0" signals corresponding to each determination result are generated. Output to terminals EQ1 to EQ7. The pitch match number detection circuit 703 detects the number of matches between the two based on each match output EQ1 to EQ7 of the parallel pitch match determination circuit 702, and outputs the corresponding pitch match number data. The counter 704 receives a determination enable signal
It is reset by Sjudge and the shift signal Sshift
The number of "1" pulses included in the data is counted, and the number of shifts is outputted. A match number comparison circuit 705 compares pitch match number data latched in a latch circuit 707 (described later) with pitch match number data output from the pitch match number detection circuit 703, and determines the pitch match number. Only when the pitch matching number data output from the detection circuit 703 is larger than the pitch matching number data latched by the latch circuit 707, "1" is output. The AND gate 706 is connected to the match number comparison circuit 7.
The opening/closing is controlled by the output of 05, thereby allowing the shift signal Sshift-1 to pass. The latch circuit 707 receives a determination enable signal.
It is reset by Sjudge, and in response to the "1" pulse output from the AND gate 706, it latches the pitch coincidence number data output from the pitch coincidence number detection circuit 703. The latch circuit 708 is similarly reset by the determination enable signal Sjudge, and latches the shift number data output from the counter 704 every time a "1" pulse is output from the AND gate 706. The latch circuit 709 is reset by the clear signal Sclr, and each time a "1" pulse arrives during the latch signal Slatch, the latch circuit 709 is reset.
The pitch matching number data latched at 07 is latched. The latch circuit 710 is similarly reset by the clear signal Sclr, and the latch circuit 710 is reset by the clear signal Sclr.
Each time a "1" pulse arrives during Slatch, the shift number data latched in the latch circuit 708 is latched. Then, the output of the latch circuit 709 becomes the matching pitch number data PDeq-1, and the latch circuit 71
The output of 0 becomes shift count data PDline sml-1. As a result, if each of the circuits described above operates normally, the 27th
Shift control is performed sequentially as shown in Figures B to J.
In the parallel pitch coincidence determination circuit 702, as shown in FIG. 28A to I, data PDline (ref) and data
A match with PDline (1 to 7) is determined. In the example shown in FIG. 28, the pitch string partial data PDline (1 to 7) corresponding to 0 shifts is detected as the most similar pitch string partial data, and as a result, the data output from the latch circuit 710
The content of PDline sml-1 becomes "0", and the content of data PDeq-1 becomes "3". Next, details of the k-th similar pitch string detection circuit 700-k are shown in FIG. In the figure, the shift register 751 responds to the load signal Sload-k by
Input pitch string data PDline (full) is loaded, and the shift is controlled in response to the "1" pulse included in the shift signal Sshift-k, and the contents of all stages are cleared at the same time by the clear signal Sclr. . Further, each data in the first to seventh stages of the shift register 751 is configured to be individually resettable. Next, a parallel pitch match determination circuit 752, a pitch match number detection circuit 753, a match number comparison circuit 755, an AND gate 756, a counter 754, a latch circuit 7
59, each operation of the latch circuit 760 is as follows:
Each operation of the pair 59 and the latch circuit 760 in the first similar high pitch sequence detection circuit 700-1 is as follows.
This circuit is exactly the same as the corresponding circuit in the first similar pitch sequence detection circuit 700-1, and will not be repeatedly described here. Match determination circuits 761-1 to 761-(k-1)
are the shift number data output from the counter 754 and the shift number data PDline sml-1 to PDline sent from the similar pitch string detection circuits 700-1 to 700-(k-1) at the previous stage, respectively.
sml-(k-1), and outputs "1" only when it is determined that both data match. The OR gate 762 includes each of the coincidence determination circuits 76
The logical sum of the outputs of 1-1 to 761-(k-1) is output. The gate circuit 763 is controlled to open and close by the output from the OR gate 762, thereby supplying each coincidence output of the parallel pitch coincidence determination circuit 752 to the reset terminal R of the first to seventh stages of the shift register 751. do. AND gate 764 is controlled to open and close by the output of OR gate 762 which is inverted by inverter 765, thereby inhibiting the output of AND gate 756. The latch circuit 757 receives the determination enable signal
In response to the "1" pulse that is reset by the Sjudge and passes through the AND gate 764, the pitch match number data output from the pitch match number detection circuit 753 is latched. The latch circuit 758 is similarly reset by the determination enable signal Sjudge and latches the shift number data output from the counter 754 in response to the "1" pulse passed through the AND gate 764. As a result, when each of the above circuits operates, the 27th
As shown in Figures B to J, the input pitch data stored in each stage of the shift register 751 is sequentially shifted leftward in the figure in response to the "1" pulse of the shift signal Sshift-k, and at the same time, the input pitch data stored in each stage of the shift register 751 is In the pitch match determination circuit 752, as shown in FIGS. 29 and 30, a match determination process is performed between the pitch string partial data and the reference pitch string data for each number of shifts. Here, FIGS. 28, 29, and 30 show the first, second, and third similar pitch sequence detection circuits 70, respectively.
It shows the operation of each parallel pitch coincidence determination circuit 702, 752 in 0-1, 700-2, 700-3. As is clear from these figures, the second to kth
In each parallel pitch coincidence determination circuit 752 in the similar pitch sequence detection circuit, when the shift timing of the similar pitch pitch series partial data PDline (1 to 7) already detected in the previous stage similar pitch series detection circuit arrives, , portions included in these data that match the reference pitch string data PDline (ref) are individually reset by the output of the gate circuit 763. As a result, a portion that matches the reference pitch string data included in the already detected similar pitch string partial data will not be recognized as being duplicated with another similar pitch string partial data. Furthermore, when the shift timing of the similar pitch string partial data already detected in each preceding stage similar pitch string detecting circuit arrives, the OR gate 7
The AND gate 764 is inhibited by the output of 62, and therefore the k-th similar pitch sequence detection circuit 700-k
, data with the lowest degree of similarity is always detected than the similar tone high pitch string partial data PDline (1 to 7) that has already been detected in the first to k-1th similar tone high tone string detection circuits. It becomes. Next, the details of the matching pitch partial superposition circuit 8 will be explained in the seventh section.
As shown in the figure. In the same figure, a shift register 801
loads the input pitch string data PDline (full) in response to the load signal Sload-(k+1). Then, in response to the "1" pulse included in the shift signal Sshift-(k+1), the shift control is performed in the downward direction in the figure, and the contents of each stage are simultaneously cleared by the clear signal Sclr. A parallel pitch coincidence determination circuit 802 receives pitch string partial data PDline (1 to 7) output in parallel from the first to seventh stages of the shift register and output from the reference tone pitch string generating circuit 1. A match determination process is performed for each stage with the reference pitch pitch sequence data PDline (ref), and the determination result is outputted to terminals EQ1 to EQ7 as a 1-bit signal for each stage. The counter 803 is a judgment enable signal Sjudge.
At the same time, the shift signal Sshift-(k+1) is counted, thereby outputting shift number data corresponding to the number of shifts of the shift register 801. The coincidence determination circuits 804-1 to 804-k each receive the shift number data output from the counter 803 and the respective similar tone high pitch sequence detection circuits 700.
It is used to determine whether the shift count data PDline sml-1 to PDline sml-k outputted from -1 to 700-k match, and outputs "1" only when it is determined that they match. The OR gate 805 includes each of the coincidence determination circuits 80
4-1 to 804-k, and the output of this OR gate 805 controls data superimposition processing, which will be described later. AND gates 806-1 to 806-7 are respectively controlled to open and close by the output of the OR gate 805, and allow each output of the parallel pitch coincidence determination circuit to pass therethrough. The latch circuits 807-1 to 807-7 are latch-controlled by each coincidence output of the parallel pitch coincidence determination circuit 802, which is supplied via AND gates 806-1 to 806-7, respectively. Among the outputs from the first stage to the seventh stage outputted from the register 801, only the portion that matches the reference pitch string data PDline (ref) is latched at each shift. The series of pitch data latched by these latch circuits 807-1 to 807-7 constitutes the most similar pitch string data PDsample. As a result, when each circuit explained above operates,
As shown in FIG. 35, from among the similar tone string data detected by the similar tone pitch string detection circuits 700-1 to 700-k of each similar stage, only the portion that matches the reference tone pitch string data is extracted. , are superimposed on each other to form the most similar pitch string data PDsample. Next, the details of the similar note length string extracting circuit 10 are shown in FIG. In the figure, the configurations of the first to k-th similar note length string detection circuits 1000-1 to 1000-k are as follows:
The configuration is approximately the same as that of the first to k-th similar pitch tone sequence detection circuits 700-1 to 700-k shown in FIG. 4 described above, that is, from each similar note length tone sequence detection circuit,
Shift count data LDline sml-1 to LDline sml-k corresponding to a plurality of similar note length tone string partial data selected in order of similarity,
Code length matching number data LDeq-1 to LDeq-k are output. FIG. 9 shows details of the first similar note length string detection circuit 1000-1. In the figure, a shift register 1001, a code length match detection circuit 1003, a counter 1004, a match number comparison circuit 1005, an AND gate 1006, a latch circuit 1007, a latch circuit 1008, a latch circuit 1009, a latch circuit 10
10 is the same as the first similar pitch sequence detection circuit 700-1 shown in FIG. 5, except that the data handled is changed from pitch data to note length data. The explanation will not be repeated here. On the other hand, parallel code length matching circuit 1002
The configuration is slightly different from that of the parallel pitch coincidence determination circuit 702 shown in FIG. 5 above. That is, as shown in FIG. 25C, the reference note length tone string data output from the reference note length tone string generation circuit 10
Each of the constituent notes of LDline (ref) has a certain standard length, such as an eighth note, a quarter note, a dotted quarter note, a half note, etc. On the other hand, the note length string partial data output from the first to seventh stages of the shift register 701 are as shown in FIGS. 31B to 31J.
Each has errors Δ1 to Δ9. Here, for convenience of explanation, the values of errors Δ1 to Δ9 attached to each note are assumed to be within a predetermined tolerance range that can be recognized as each corresponding note. Therefore, when the parallel pitch match determining circuit 702 accurately determines the match between the corresponding stages in both tone string data, it is almost impossible for both data to completely match. Therefore, in the parallel note length coincidence determination circuit 1002, the note string data LDline (ref) and the note string data
When comparing with PDline (1 to 7), if the error from the reference extension length data for each stage is within a predetermined tolerance range, this is considered to be a match. In other words, matching outputs are obtained at the terminals EQ1 to EQ7 only when the differences between A1 to A7 and B1 to B7 are within a predetermined tolerance range. As a result, when each circuit shown in FIG. 9 operates,
Inside the shift register 1001, each stage is shifted as shown in FIGS.
As shown in Figure 32, the reference note length string data LDline (ref) and note length string partial data are
PDlines (1 to 7) are compared in magnitude, and it is determined for each stage whether the length of each note is within a predetermined allowable range. Then, the note length matching number detection circuit 1003 outputs numerical data corresponding to each number of matching notes, and at the same time, the number of shifts is outputted to the counter 1004. Finally, the latch circuits 1009 and 1010 output the most The number of matching notes and its shift count data LDeq-1 and LDline sml-1 corresponding to the note length note string partial data having a large number of matching notes are output. Next, the k-th similar note length tone string detection circuit 1000-k
The details are shown in FIG. In the figure, a shift register 1051, a code length matching number detection circuit 105
3, counter 1054, match number comparison circuit 105
5, AND gate 1056, latch circuit 105
7, latch circuit 1058, latch circuit 1059,
Latch circuit 1060, match determination circuit 1061-1
~1061-(k-1), the configuration of OR gate 1062, gate circuit 1063, AND gate 1064, and inverter 1065 is such that the type of data handled has changed from pitch data to note length data, and other details are shown in FIG. The configuration of the parallel note length match determination circuit 1052 is the same as the k-th similar pitch string detection circuit 700-k shown in FIG.
Since it is the same as the similar note length sound sequence detection circuit 1000-1, repeated explanation will be avoided here. As a result, when each circuit explained above operates,
For example, the second and third similar note length string detection circuits 1000
-2,1000-3, as shown in Figures 33 and 34, the reference note length string data LDline
(ref) and the note length string partial data PDline (1 to 7) is performed, and at this time, in the same way as in the case of the pitch data described above, the similar pitch string data that has already been selected is checked. Excluded from match determination. Then, a second similar note length string detection circuit 1000-
2, the note length note string partial data with the second largest number of matching notes is extracted, and the third similar note length note string detection circuit 10
The third most similar note length note sequence partial data is extracted from 00-3. Then, each extracted note length note string partial data is the input note length note string data LDline (full).
is converted to shift count data for LDline
It is output as sml-1 ~ LDline sml-k, and the number of matching notes for each sound string data is output as LDeq-1 ~
It is output as LDeq-k. Next, the details of the matching code length partial overlapping circuit 11 will be explained in the first section.
Shown in Figure 1. In the figure, shift register 11
01, counter 1103, match determination circuit 1104
-1 to 1104-k, blue gate 1105, AND gates 1106-1 to 1106-7, and latch circuits 1107-1 to 1107-7, the only difference is that the data handled has changed from pitch data to note length data. It is completely the same as the matching pitch partial superimposition circuit 8 shown in FIG. Since this is the same as the circuit, a repeated explanation will be avoided here. As a result, when the circuit described above operates, the latch circuits 1107-1 to 1107-7 detect the notes detected by the first, second, and third similar note length string detection circuits, respectively, as shown in FIG. Each matching note in the long note string partial data is superimposed on each other, thereby creating the most similar note long note string data LDsample.
will be formed. Next, details of the performance result display device 13 are shown in FIG. 17. In the figure, reference melody generation circuit 1
301 is each pitch data PD (ref) and note length data that compose the reference melody according to fixed timing.
It outputs LD (ref) and outputs a switching signal Scos in response to each output timing. The multiplexer 1302 selectively outputs each pitch data constituting the most similar pitch string data LDsample in response to the switching signal Scos. The multiplexer 1303 selectively outputs each note length data constituting the most similar note length tone sequence data LDsample in response to the switching signal Scos. A pitch match determination circuit 1304 receives input pitch data sequentially output from the multiplexer 1302.
It is determined whether PD(in) matches the reference pitch data PD(ref) outputted from the reference melody generation circuit 1301, and outputs "1" only when the two do not match. The code length match determination circuit 1305 receives the input code length data sequentially output from the multiplexer 1303.
It is determined whether LD(in) matches the reference note length data LD(ref) outputted from the reference melody generation circuit 1301, and outputs "1" only if they do not match. The outputs of the pitch match determination circuit 1304, note length match determination circuit 1305, and OR gate 1306 are sent to the CRT controller 1307 as pitch blinking signals, note length blinking signals, and staff score blinking signals, respectively.
supplied to The pitch display data ROM 1308 outputs pitch display data corresponding to the reference pitch data PD (ref) that is sequentially output in a time-division manner from the reference melody generation circuit 1301. The note length display data ROM 1309 is standard note length data output from the standard melody generation circuit 1301.
Based on LD(ref), output the corresponding note length display data. The note display data ROM 1310 outputs corresponding staff display data based on the reference pitch data PD (ref) and the reference note length data LD (ref) output from the reference melody generation circuit 1301. The CRT controller 1307 displays the upper row on the screen of the CRT 1311 based on the pitch blinking signal, note length blinking signal, staff blinking signal, pitch display data, note length display data, and note display data. , interruption, and lower part, and display the staff notation, pitch, and note length corresponding to the standard melody, respectively. Then, the pitch display and note length display displayed in the middle and lower rows will flash the incorrect part in the display, and the staff display in the upper row will have a pause in either pitch or note length. If they are different, a blinking display is performed in the same way. As a result, the CRT1311 screen allows performers or singers to reliably recognize mistakes in their performances, etc., and improve performance by concentrating on only those mistakes. It becomes possible to effectively improve technology. In addition, with the three types of displays on this display screen, it is possible to reliably recognize whether the pitch or note length is incorrect, and it is also possible to display only the error location based on the standard melody. Therefore, as shown in Figure 26A, there is no indication of parts where the performer has already recognized a mistake and has replayed it, and there is no indication of the part where the performer has unconsciously made a mistake. can be reliably informed only. Next, details of the performance time/time difference detection circuit 22 are shown in FIG. In the figure, selector 2201
is determined by pitch matching number data PDeq corresponding to the number of matching notes between the most similar pitch string data PDsample output from the individual pitch matching judgment circuit 14 and the reference pitch string data PDline (ref). It is configured so that switching can be controlled individually for each stage, and therefore the contents of each stage, either data PDsample or data PDline (ref), is output to output terminals OUT1 to OUT7. The gate circuit 2204 is controlled to open and close by a latch signal Slatch, and the gate circuit 2204 is controlled to open and close by a latch signal Slatch.
Note length matching number data LDeq corresponding to the number of matching notes between the most similar note length data LDsample outputted from 6 and the reference note length note string data LDline (ref) is passed. The OR gates 2202-1 to 2202-7 are the pitch matching number data PDeq and the note length matching number data.
LDeq and LDeq are output for each stage, and latch circuits 2203-1 to 2203-7, which will be described later, are latch-controlled by the outputs of these OR gates. The latch circuits 2203-1 to 2203-7 are specially latch-controlled by the outputs of the OR gates 2202-1 to 2202-7, so that each of the outputs OUT1 to OUT of the selector 2202
Latch the note length data output from 7. The adder circuit 2205 is connected to the latch circuit 2203.
All the latched code length data from -1 to 2203-7 are added and the total sum is output. The addition circuit 2206 receives the reference note length tone string data.
Add all the note length data that make up LDline (ref) and output the total sum. The output of this adder circuit 2206 is the performance time data D
(T). The subtraction absolute value circuit 2207 is connected to the addition circuit 22.
The difference between the addition data outputted from the adder circuit 2206 and the addition data outputted from the adder circuit 2206 is determined, and the absolute value thereof is output, and this absolute value becomes the performance time difference data D(ΔT). As a result, when each circuit explained above operates,
In the latch circuits 2203-1 to 2203-7, as shown in FIG.
If the pitch is incorrect in PDsample, if the note length is also incorrect in the most similar note length note string data LDsample, the blank area corresponding to the note length error will be replaced with the standard note length note string data. This will be corrected using the correct note length data present in the corresponding location. In other words, pitch is a particularly important element when practicing melody, and therefore, the note length of a note pressed at the wrong pitch is considered to be the correct note length when determining the note length, and this allows the correct pitch to be pressed. Judging the note lengths only for key tones results in accurate scoring that is more similar to that of a music teacher. Next, the details of the pitch similar tone sequence number detection circuit 18 are shown in FIG. In the figure, a zero data generation circuit 1801 outputs k zero data in parallel. The coincidence determination circuit 1802 outputs the zero data output in parallel from the zero data generation circuit 1801 and the k pitch coincidence number data PDeq-1 to PDeq-k output from the similar pitch sequence extraction circuit 7. The matching results are individually determined and each determination result is output to terminals EQ1 to EQk as a 1-bit signal. The mismatch number detection circuit 1803 detects the number of "0" signals outputted from EQ1 to EQk in correspondence with the match determination circuit 1802, and outputs this as pitch similar sound sequence number data D (N3). Next, the details of the note length similar note sequence detection circuit 19 will be explained in the 14th section.
As shown in the figure. In the figure, the zero data generation circuit 19
01 outputs k zero data in parallel. A coincidence determination circuit 1902 receives k zero data outputted in parallel from the zero data generation circuit 1901 and k note length coincidence number data LDeq-1 outputted in parallel from the similar note length string extraction circuit 10. ~
Matching with LDeq-k is individually determined, and each determination result is output to terminals EQ1 to EQk as a 1-bit signal. The mismatch number detection circuit 1903 detects the number of signal "0" outputted from the coincidence determination circuit 1902, and outputs this as note length similar note sequence set number data D (N4). Next, the details of the reference pitch number detection circuit 20 are shown in FIG. In the figure, zero data generation circuit 2001
outputs 14 zero data in parallel. A coincidence determination circuit 2002 separates the 14 zero data outputted from the zero data generation circuit 2001 and the reference tone pitch sequence data PDline (ref) outputted from the reference tone pitch sequence generation circuit 1 for each stage. Individually determine whether they match and send the results to terminals EQ1~
A 1-bit signal is output in parallel to the EQ14. The mismatch number detection circuit 2003 detects "0" among the outputs output from the match determination circuit 2002.
Detect the number of pitches, and use this as the reference pitch number data D (N5)
Output as . Next, the details of the reference mark length number detection circuit 21 will be explained in the 16th section.
As shown in the figure. In the same figure, the zero data generation circuit 21
01 outputs 14 zero data in parallel. A coincidence determination circuit 2102 converts the 14 zero data outputted from the zero data generation circuit 2101 and the reference note length tone string data LDline (ref) outputted from the reference note length tone string generation circuit 12 at each stage. The results are determined individually and the results are sent to terminal EQ1.
~Output to EQ14 in parallel as a 1-bit signal. The mismatch number detection circuit 21 includes the match determination circuit 2
The number of "0"s is detected among the signals outputted from the 102 terminals EQ1 to EQ14 and outputted as reference code length number data D (N6). Next, in the similarity calculation circuit 23, the thus obtained matching pitch number data D (N1), matching note length number data D (N2), pitch similar note string set number data D
(N3), note length similar note sequence set number data D (N4), standard pitch number data D (N5), standard note length number data D (N
6) Based on the performance time difference data D(ΔT) and the performance time data D(T), calculate the score for the melody performance performed by the student using the following calculation formula. Score X = {[100×(N1/N5)×(N2/N6)×(T-
|ΔT|)/T]-Y}÷5×5 However, N1; Number of matching pitches N2; Number of matching note lengths N5; Reference number of pitches N6; Reference number of note lengths T; Model performance time ΔT; Performance time difference Y ;Score data obtained by the above calculation formula: Number of pitch-similar note sequence sets N3 or note length similar note sequence set N4
Dscore is supplied to the score display 24, so that the performance score is displayed on the score display 24, for example, as shown in FIG. Thus, in the melody performance training device shown in this embodiment, the melody played by the student on the keyboard 4 is detected by the key press detection circuit 5, the input pitch string forming circuit 6, and the input note length string forming circuit 9.
is converted into two-dimensional tone sequence data consisting of pitch and note length, and then these tone sequence data are processed for each dimension by similar tone sequence extraction circuits 7 and 10 and matching part superimposition circuits 8 and 11. This method uses a method to perform pattern recognition processing through the melody, and finally extracts the phrase most similar to the standard melody, from among the melodies played by the students.
It becomes possible to reliably extract only the vaguely similar melodic parts that the music teacher (human) unconsciously feels. In other words, when comparing some kind of melody against a standard melody, human beings make an analog judgment, not only whether the two match or not, but also whether they are somewhat similar or not. According to the apparatus of this embodiment, it is possible to reliably extract only those portions that are somewhat similar to such a reference melody from the input performance melody. Further, according to this embodiment device, the most similar pitch or note length data PDsample, LDsample output from the matching portion superimposing circuits 8 and 11, respectively.
Based on this, the system further displays which parts of the extracted melody parts that are somewhat similar actually differ from the reference melody, so by displaying the error parts, the performer or It becomes possible for the student to reliably know the parts of the performance that he or she has made unconsciously. Further, according to this embodiment device, the most similar sound sequence data extracted from the matching part superimposition circuits 8 and 11
PDsample and LDsample are used as basic data, and these data are converted into matching pitch number data D (N1), matching note length number data D (N2), and pitch similar note sequence set number data D
(N3), note length similar note sequence set number data D (N4), standard pitch number data D (N5), standard note length number data D (N
6) Since the performance score is calculated by comparing performance time difference data D (ΔT) and model performance time data D (T), etc., it is possible to obtain scoring results that are extremely close to the sense of music teachers. . That is, by taking the ratio between the number of matching pitches N1 and the reference number of pitches N5, and the ratio between the number of matching note lengths N2 and the reference number of note lengths N6, the melody part that is somewhat similar to that mentioned above and the reference are obtained. It can express the difference from the melody, and also has a model performance time T.
By finding the ratio between the model performance time T and the performance time difference ΔT, we can find out how much error there is in the total performance time of the somewhat similar melody part from the reference melody performance time. Furthermore, by knowing the number N3 of pitch-similar note string sets and the number N4 of note-length similar note string sets, respectively, it is possible to know how many parts of the performance melody are redrawn. Therefore,
By using these as grading factors, it is possible to obtain grading results that are extremely close to the senses of music teachers (humans). In the above embodiment, the melody performance performed using the keyboard 4 is displayed, and the performance results are displayed or points are obtained for the performance results. Although the key press detection circuit 5 known in so-called electronic musical instruments is shown as a means for obtaining single note data, the application of the present invention is not limited thereto, and, for example, the output signal of a microphone or an appropriate electronic musical instrument, etc. It is also possible to display or score similar performance results based on the performance results. In this case, instead of the key press detection circuit 5, the second
A pitch data detection circuit 26 as shown in FIG. 4 may be provided. In FIG. 24, the fundamental wave detection circuit 26
01 detects the fundamental wave contained in the input audio signal or musical tone signal, converts it into a corresponding analog voltage, and outputs it. The A/D conversion circuit 2602 converts the analog voltage output from the fundamental wave detection circuit 2601 into a digital signal. The pitch reference data generation circuit 2603 generates C1, C
A series of reference pitch data consisting of 1#, D1...Cn is output in parallel. Permissible determination circuits 2604-1 to 2604-p receive the input pitch data output from the A/D conversion circuit 2602 and the pitch reference data generation circuit 260.
The pitch difference corresponding to the results of these comparisons is compared with each standard pitch data output from 3.
Each terminal is
Output “1” to EQ. Gate circuits 2605-1 to 2605-p are permissible determination circuits 2604-1 to 2604-p, respectively.
Opening/closing is controlled by each matching output, thereby allowing pitch data corresponding to C1 to Cn to pass through. The gate 2606 is connected to each of the gate circuits 260
5-1 to 2605-p, that is, this OR gate 2
From 606, sound string data corresponding to the input audio signal is alternatively output. The or gate 2607 is the or gate 2606
The logical sum of each bit is calculated, and it is thereby detected that some kind of tone data has been input. Monomulti 2608 is the or gate 260
In response to the output "1" of the OR gate 2606, a minute width "1" is output, and this "1" pulse becomes the key press timing signal Skon' shown in FIG. This results in the input pitch data PD (in) shown in the figure. Further, in the embodiment, the key press detection circuit 5
Alternatively, although the pitch data detection circuit 26 is shown as converting each input note into two-dimensional data consisting of pitch and note length, the application of the present invention is not limited to this; for example, as described above. Of course, in addition to the pitch and note length described above, it is also possible to convert and detect the accent dimension, timbre dimension, volume dimension, etc. For example, to give a concrete explanation using tones as an example, the tone sequence pattern of the reference pitch pitch sequence generation circuit 1 shown in FIG. 1 and the tone sequence pattern sequentially input to the shift register 607 shown in FIG. As,
The timbre information of each component tone of a tone string, for example, flute, piano, violin, trumpet, etc., may be sequentially configured into the tone color string pattern. That is, although digital pitch information is input from the shift register 607, it is only necessary to configure this so that formant data possessed by each constituent note of the tone sequence is input to the shift register 607. And to be more specific,
Each constituent note of the tone sequence is sequentially filtered through each band filter (for example, center frequency is 500Hz, 1000Hz,
The analog voltage level output from each band filter is
When converted to digital information, the value is, for example, "4",
If the values are "3" or "2", the digital data may be input into the shift register sequentially for each constituent tone. A device configured in this manner can be used, for example, for training in timing alignment of each musical instrument sound by an orchestra. Furthermore, although it is the same idea as this kind of thing,
Comparatively leading formalants such as cat's cry, dog's cry, bird's cry, etc. are extracted, and this kind of tone-based pattern extraction can be easily realized in the same way. Further, the sound volume can also be easily realized by converting each volume level of each constituent sound of the sound string into digital information and inputting the digital information to the shift register 607. The device configured in this manner is suitable for use in training to produce emotional feelings as the music progresses. Furthermore, when configuring a chord or rhythm practice device, the following steps may be taken. That is, each detected performance information is filtered by a plurality of parallel bandpass filters similar to those described above, thereby converting the constituent note data of each chord into digital data, and All you have to do is input the digital data into the shift register. Furthermore, when constructing an accent practice device, it is sufficient to similarly detect dynamic data from the input performance data, convert this into digital data, and then input it to the shift register. In this way, in addition to the above tone,
It is suitable for use in training to create a more emotional feeling. In addition, in the above-mentioned embodiment, for notes whose pitches are already incorrect in the performance time and time difference detection circuit 22, the note lengths are considered to be correct, thereby giving a score that emphasizes the unique characteristics of the melody, that is, the pitches. However, if you configure the system so that the pitch is considered correct for notes with the wrong note length, it is of course possible to obtain performance scoring results that emphasize rhythm. . Next, details of another embodiment (hereinafter referred to as the second embodiment) including the first and third inventions of this application will be described with reference to FIGS. 38 to 40. The device of this second embodiment has the appearance of a portable keyboard device, and in addition to its functions as a normal electronic musical instrument, it can also play scale roulette games, scale golf games, scale tennis games, etc. It is equipped with various scale game functions such as, and in addition to these games, it is also equipped with a melody lesson function according to the present invention. Further, control of this electronic musical instrument is performed by a so-called microprocessor, and the system configuration thereof is shown in FIG. In the same figure, the system program ROM 27
A system program that defines various operations performed by the CPU 28 is stored in the CPU 28. The working RAM 29 is used as a working area when executing the various system programs mentioned above. The pitch data ROM 30 stores a plurality of sets of pitch data for, for example, two measures, which constitute a melody to be composed when performing the composition operation processing described later. teeth,
One of them is selectively read out depending on the random number data. The note data ROM 31 contains the pitch data.
A plurality of sets of a series of musical note data, for example two measures, composed in the same manner as the ROM 30 are stored, and as in the case of the pitch data ROM 30, these melodies are selectively read out using random number data. It happens. The musical score data RAM 32 temporarily stores, for example, a two-bar melody composed by a melody composition operation process to be described later. The keyboard section 33 includes a keyboard and a key press detection circuit for detecting keys pressed on the keyboard, and outputs so-called pitch data, key-on timing signals, etc. from the keyboard section 33. The operating section 34 includes an operation mode changeover switch, a power on switch, and other various switches for switching various operations of this device, and the outputs of these switches are outputted. The score display section 35 is composed of a character display device constituted by, for example, a liquid crystal display, and the score obtained by the scoring process described later is displayed on this character display device. Become. The musical tone forming section 36 is constituted by a conversion processing device that converts pitch data into a musical tone signal, which is well known in this type of electronic musical instrument, and the musical tone signal outputted from the musical tone forming section 36 is After being amplified via the amplifier 37, the speaker 38
It will be pronounced from Next, the basic operation of this embodiment device will be explained in the 39th section.
The general flowchart in Figure. First, the execution contents of each step shown in this general flowchart will be individually explained in detail. Step (1): In response to the mode changeover switch being switched to melody lesson in the operation section 34, a melody for two measures is composed. In this composing operation, the pitch data ROM 30, the note data, and the note data are read out at random using random number data, and the melody for two measures combined by these is created using the musical score data.
It is stored in RAM32. The details of this composing motion are provided by the applicant in Japanese Patent Application No. 158437/1983.
No. 132494, Patent Application No. 125603, Patent Application No. 1982-
Since the application has already been filed under No. 132493, etc., it will not be described in detail here. Step (2): Two bars of composed melody data stored in the musical score data RAM 32 are sequentially transferred to the musical tone forming section 36 in a time-sharing manner, and converted into musical tone signals. The converted musical tone signal is then amplified via an amplifier 37 and then outputted from a speaker 38. Step (3): Determine whether any key switch has been operated on the operation unit 34, and if the determination result is YES, proceed to step (8); if NO, proceed to step (4). . Step (4): Determine whether any key has been pressed on the keyboard section 33, and check the determination result.
If YES, proceed to step (6); if NO, proceed to step (5). Step (5): Determine whether or not no key has been pressed in the keyboard section 33 for 3 seconds or more. If the determination result is YES, proceed to step (7); if NO, proceed to step (7). Return to 3). Step (6); Calculate the result of the performance performed according to the first embodiment with respect to the key press data (consisting of pitch data and note length data) input from the keyboard section 33. A score calculation is performed, and the score obtained thereby is output to the score display section 35. The details of the answer processing (6) are shown in FIG. 40. Step (7): Execute model calculation processing similar to step (2). Step (8); Determine whether or not the key pressed on the operation unit 34 is the start key. If the determination result is YES, return to step (1) and press NO.
In this case, the melody lesson ends. Next, details of the answer processing performed in step (6) will be explained according to the flowchart of FIG. Step (601): Initialize the location where the pitch data and note length data input from the keyboard section 33 are taken. Step (602): Jitter is removed from the key press data sequentially fetched from the keyboard section 33, and the remaining data is fetched into a predetermined area in the initially set working RAM 29. Note that the note length that can be considered as jitter here corresponds to a key press for a short time of about 100 ms. Step (603): The key press data obtained by the latest key press to the past 12 tones are retained, and the old data before that is deleted. Step (604): Determine whether or not input of key press data from the keyboard section 33 has been completed, and if the determination result is YES, proceed to step (605);
If so, return to step (602). Step (605); Pitch string data and note length string data stored in the musical score data RAM 32 are selected from a series of pitch string data and note string data stored in the set area in the working RAM 29. Extract the most similar sound string data. Note that when extracting this similar sound string, the pattern processing operation described in the first embodiment is performed, so a repeated explanation will be avoided here. Step (606): As explained in the first embodiment, the most similar note sequence and the musical score data RAM are
For performance parts where the pitch is incorrect compared to the reference tone string stored in the memory, the notes are assumed to be correct and the note lengths are corrected. Step (607): Compare the model performance time of the standard melody stored in the musical score data RAM 32 with the performance time indicated by the most similar note string extracted in step (605); Find the time and use it as a scoring element. For example, here, specifically, if the total performance time after note length correction is shorter than the model performance time, 2 points will be deducted every 0.1 seconds, and if it is longer, 5 points will be deducted across the board. . Step (608): Calculate the pitch correct answer rate according to the formula: score x number of correct answers ÷ number of questions. Step (609): Compare each note data for two measures stored in the data RAM 32 with the note data of the most similar note sequence extracted in the step (605) for each note, and compare all the note data for each note. If the length of the note matches the reference note, no points will be deducted; otherwise, a uniform 5 points will be deducted, and this will be used to score variations in notes. Step (610); Each note data stored in the musical score data RAM 32 and the step (605)
) is individually matched with each note data constituting the most similar note string extracted in (a), and five points are deducted for each incorrect note. Step (611): The total score is calculated by adding up all the scores obtained in each of the above steps, and rounding down the portions that are less than 5 points. Step (612): The total score data obtained in step (611) is sent to the score display section 35, thereby displaying numerical values on a predetermined display device. Next, the operation of the flowchart consisting of each step explained above will be systematically explained. First, in the operating section 34 shown in FIG. 31, the mode changeover switch is set to the melody lesson side. Then, in the flowchart of Figure 39,
Steps (1) and (2) are executed in sequence, and automatically composed melody sounds (for example, two measures) are produced from the speaker 38. Next, the student or performer performs a melody corresponding to the answer based on the melody sounds heard with the keyboard section 33. Then, step (3) → step (4) → step (6)
Then, the answer processing program is executed. In the answer processing program (6), as shown in FIG. 40, first step (601) is executed, and then, while the key is kept pressed on the keyboard section 33, the steps of step (602) → step (603) → step are executed. (604)→
(602) is repeatedly executed, and as a result, pitch string data and note string data from which jitters and the like have been removed are formed in a predetermined setting area of the working RAM 29. Next, when the performance on the keyboard section 33 is completed, the aforementioned steps (605) to (612) are executed, and as a result, the score display section 35 displays the scoring results for the melody returned from the keyboard section 33. It happens. Next, when the execution of step (6) is completed, the steps are repeated again from step (3) → step (4) → step (5) → step
(7) is repeatedly executed, and the same melody is sounded from the speaker 38 again. In other words, if the melody asked and the melody answered are different,
In order to be able to practice the same melody repeatedly, the same melody is repeatedly sounded unless the start key is pressed at the end of the answer process (6). Next, when the start key is pressed on the operation unit 34, the execution result of step (3) becomes YES, and then step (8) → step (1) → step (2)
is executed, and another newly composed melody is sounded from the speaker 38, and by repeating the above, it is possible to repeatedly perform the same melody or a different new melody. It becomes. Then, each time the answer is completed, the score display section 35 shows the following:
Scores corresponding to the performance results on the keyboard section 33 are displayed repeatedly, allowing the student to effectively improve his melody practice. Thus, according to the second embodiment, by repeating the operation of memorizing automatically composed short melody tones, giving answers to them via the keyboard section 33, and checking the scores for the performance results, This type of melody lesson can be carried out extremely effectively without the need of a teacher. As is clear from the explanation of the first and second embodiments explained above, according to the first invention related to this application, when a music teacher usually grades some kind of melody, he/she unconsciously extracts it. It is possible to automatically extract melody parts that are somehow similar to the model melody, and if a student notices a part he made a mistake in the middle of playing the melody and plays it again, the mistake and the part can be automatically extracted. The replayed portions are automatically deleted, making it possible to reliably extract only the melody portions necessary for scoring, etc. Furthermore, according to the second invention related to this application,
Among the melody parts played by the students, the rewritten and corrected parts mentioned above are deleted, ensuring that only the parts that the students unconsciously made mistakes are displayed, thereby making melody practice more effective. can be improved. Furthermore, according to the third invention related to this application,
It is possible to obtain scoring results that are extremely close to the senses of a music teacher, and by displaying such analogical scores, it is possible to provide even more effective criticism to students.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the overall configuration of a tone string pattern extraction method including the first to third inventions of this application, a performance result display device using the same, and a performance result scoring device, and FIG. 2 is a block diagram showing an input sound FIG. 3 is a block diagram showing details of the input note length string formation circuit, FIG. 4 is a block diagram showing details of the similar pitch string extraction circuit, and FIG. FIG. 6 is a block diagram showing details of the K-th similar pitch string detection circuit, and FIG. 7 is a block diagram showing details of the matching pitch partial superimposition circuit. , FIG. 8 is a block diagram showing the details of the similar note length tone sequence extraction circuit, and FIG.
The figure is a block diagram showing details of the first similar note length tone string detection circuit, FIG. 10 is a block diagram showing details of the k-th similar note length tone string detection circuit, and FIG. 12 is a block diagram showing details of the performance time/time difference detection circuit, FIG. 13 is a block diagram showing details of the circuit for detecting the number of note strings with similar pitches, and FIG. 14 is a block diagram showing the details of the number of note strings with similar note lengths. FIG. 15 is a block diagram showing details of the reference pitch number detection circuit; FIG. 16 is a block diagram showing details of the reference note length number detection circuit;
Fig. 17 is a block diagram showing details of the performance result display circuit, Fig. 18 is a diagram of a CRT screen showing the status of performance results displayed on the CRT, Fig. 19 is a block diagram showing details of the control circuit, and Fig. 20 is a block diagram showing details of the control circuit. 21 is a block diagram showing details of a similar stage designation signal generation circuit, FIG. 21 is a block diagram showing details of a shift signal generation circuit,
FIG. 22 is a block diagram showing details of the latch signal generation circuit, FIG. 23 is a time chart showing the states of each control signal output from the control circuit 25, and FIG.
The figure is a block diagram showing another example of pitch detection means.
Fig. 25 is a diagram showing an example of each tone sequence data corresponding to the standard melody, Fig. 26 is a time chart showing the states of various timing signals corresponding to the input melody, and Figs. 27 to 30 are input pitch notes. Explanatory diagrams showing details of tone sequence data processing performed in the sequence forming circuit and similar pitch tone sequence extraction circuit, Figures 31 to 34
The figure is an explanatory diagram showing the flow of tone sequence data processing carried out in the input note length forming circuit and the similar note length tone string extraction circuit, and Figure 35 shows the flow of tone sequence data processing carried out in the matching pitch partial superposition circuit. An explanatory diagram, FIG. 36 is an explanatory diagram showing the flow of note string data processing performed in the matching note length partial superimposition circuit, and FIG. 37 is an explanatory diagram showing the flow of note length data correction processing performed in the performance time/time difference detection circuit. An explanatory diagram, FIG. 38 is a block diagram showing the entire system configuration of the portable keyboard device including the first and third inventions related to this application, and FIG. 39 is a block diagram showing the melody lesson in each mode of operation of the device. FIG. 40 is a general flowchart schematically showing the related operations, and FIG. 40 is a flowchart showing details of the answer processing program in the general flowchart. DESCRIPTION OF SYMBOLS 1...Reference tone pitch sequence generation circuit, 2...Musical tone formation circuit, 3...Speaker, 4...Keyboard, 5...Key press detection circuit, 6...Input pitch pitch sequence formation circuit, 7...
Similar pitch tone string detection circuit, 8... Matching pitch partial superposition circuit, 9... Input note length tone string formation circuit, 10...
Similar note length note string extraction circuit, 11... Matching note length partial superposition circuit, 12... Reference note length note string generation circuit, 13
... Performance result display circuit, 14 ... Individual pitch coincidence discrimination circuit, 15 ... Matching tone sequence number detection circuit, 16 ...
Individual note length coincidence discrimination circuit, 17... Matching note length number detection circuit, 18... Pitch similar tone sequence number detection circuit, 19...
...Note length similar note sequence number detection circuit, 20...Reference note length number detection circuit, 21...Reference note length number detection circuit, 22...
... Performance time/time difference detection circuit, 23 ... Similarity calculation circuit, 24 ... Score display, 25 ... Control circuit, 26 ... Pitch data detection circuit, PD (in) ...
Input pitch data, PDline (full)... Input pitch string data, PDline (ref)... Reference pitch string data, PD (ref)... Reference pitch data, PDline sml
-1 to PDline sml-k...shift number data corresponding to the 1st to kth similar tone high pitch string partial data, PDeq-1 to PDeq-k...to the 1st to kth similar tone high pitch string partial data, respectively Corresponding pitch matching number data, PDsample...most similar pitch sequence data, PDeq...pitch matching number data, LDline (full)
...Input note long note string data, LDline (ref)...Reference note long note string data, LDline sml-1 to LDline sml
-k...Shift count data corresponding to the first to k-th similar note length note length data, LDeq-1~
LDeq-k...note length matching number data corresponding to the first to kth similar note length note string partial data, respectively;
LDsample...Most similar note length string data, LDeq...
...note length match data, Dscore...score data,
PDline (1 to 7)...Input pitch pitch string partial data,
LD(ref)...Reference note length data, LD(in)...Input note length data, D(N1)...Consistent pitch number data,
D(N2)... Matching code length number data, D(N3)...
Pitch similar note sequence set number data, D(N4)...Note length similar note sequence set number data, D(N5)...Standard pitch number data, D(N6)...Standard note length number data, Skon...
...key press signal, Skon′...key press timing signal,
Skon″……data acquisition signal, inverted Skon″……data acquisition signal, Splay……playing signal, Sjudge
...Judgment enable signal, Sφ...clock signal,
Send...judgment end signal, Sload...load signal,
Sshift...Shift signal, Slatch...Latch signal,
Spush...Reliable key press signal, Ssml...Similar stage designation signal, Seq...Consistent signal, Sclr...Clear signal, Sic
...Initial clear signal, Scos...Switching signal.

Claims (1)

[Claims] 1. Converting performance or singing information into one or two or more dimensional note string data having the dimension of note length; Then converting the note string data and note string data corresponding to a reference note string into a note string. A method for extracting a sound string pattern, comprising: comparing patterns with each other with respect to predetermined features; and extracting a phrase most similar to a reference sound string from an input sound string based on the comparison result. 2. Single note data detection means that converts and detects each component note sequentially generated by playing or singing into one or two or more dimensional single note data having a dimension of sequential note length; and input sound sequence data forming means for storing input sound sequence data in the order of occurrence to form input sound sequence data corresponding to performance or singing; Compare the sound string parts existing in the same time period while shifting the reference points on the axis, and select one or more sets of similar sound string parts in descending order of similarity from the input sound string data. Extracting similar sound string parts: Extracting parts that match the reference sound string data of the corresponding dimension from each of the extracted similar sound string parts, superimposing them on each other, and extracting the most similar sound string parts corresponding to the most similar phrase. A performance result display device comprising: a data superimposition means for forming similar sound string data; and a means for printing or displaying the formed most similar data. 3. The input tone string data forming means removes, from the detected single note data, single note data having a certain note length or less and which cannot be a musical element. performance result display device. 4. Single note data detection means for converting and detecting each component note sequentially generated by playing or singing into one or two or more dimensional single note data having the dimension of sequential note length; and an input tone sequence data forming means that generates and memorizes input tone sequence data corresponding to performance or singing; The input sound string data is compared with one or more sets of similar sound string parts that are selected in descending order of similarity by shifting reference points on the time axis and comparing sound string parts that exist in the same time period. Similar sound string part extraction means for extracting similar sound string parts; Data superimposition means for forming reference sound string data of a corresponding dimension from each of the extracted similar sound string parts; Among each constituent sound data of the reference sound string data. , find the ratio between the total number of sound data included in the formed most similar sound sequence data and the total number of each constituent sound data of the reference sound sequence data for the necessary dimensions, and input the values of these ratios as scoring elements. 1. A performance result scoring device comprising: scoring calculation means for scoring a tone sequence. 5. According to claim 4, the input tone string data forming means removes, from the detected single note data, single note data having a certain note length or less and which cannot be a musical element. Performance result scoring device described.