JP2021015138A - Performance evaluation method, performance evaluation device and program - Google Patents

Abstract

To evaluate performance by a user without a need of a sound signal indicating exemplary performance.SOLUTION: A performance evaluation device 100 comprises: a feature extraction section 20 for respectively generating a plurality of feature amount sequences for a plurality of times of performance by a user; a first sequence specification section 31 for specifying a plurality of analysis point sequences similar in the feature amounts from a plurality of analysis points corresponding to a different combination of the plurality of feature amounts in the first feature amount sequence and the plurality of feature amounts in the second feature amount sequence; a first index calculation section 32 for calculating an evaluation index Ea from the plurality of analysis point sequences; a second sequence specification section 41 for generating a performance code string where performance codes corresponding to the feature amount of the analysis point is arranged at every analysis point constituting each analysis point sequence; a code sequence classification section 42 for classifying the performance code string generated for the plurality of times of performance into one of a plurality of code string groups; and a second index calculation section 43 for calculating an evaluation index Eb for each code string group from one or more performance code strings classified into the code string groups.SELECTED DRAWING: Figure 2

Description

The present invention relates to a technique for evaluating a performance by a user.

Various techniques for evaluating the skill of performance by users have been conventionally proposed. For example, Patent Document 1 describes a technique for evaluating the skill of a user's performance by mutually comparing an acoustic parameter relating to an acoustic signal representing a performance by a user and an acoustic parameter relating to an acoustic signal representing a model performance. Has been proposed.

Japanese Unexamined Patent Publication No. 2008-40258

However, in order to evaluate a performance by a user using the technique of Patent Document 1, it is necessary to prepare an acoustic signal representing a model performance in advance. In view of the above circumstances, it is an object of the present disclosure to evaluate a performance by a user without preparing an acoustic signal representing a model performance in advance.

In order to solve the above problems, the performance evaluation method according to one aspect of the present disclosure generates a plurality of feature quantity series which are time series of feature quantities related to the performance sound for a plurality of performances by the user. From a plurality of analysis points corresponding to different combinations of a plurality of feature quantities in the first feature quantity series and a plurality of feature quantities in the second feature quantity series among the plurality of feature quantity series, the first feature quantity series A plurality of analysis point series, which are a series of analysis points having similar feature amounts between the second feature amount series and the second feature amount series, are specified, and the first evaluation index regarding the proficiency level of the performance by the user is analyzed. For each of the plurality of analysis point series specified for each feature quantity series calculated from the point series, a performance code corresponding to the feature quantity of the analysis point is arranged for each analysis point constituting the analysis point series. A plurality of code strings that are a set of one or more performance code strings having a similar or common arrangement of performance codes, which is generated for each of the plurality of performances. Classify into any one, and for each of the plurality of code string groups, a second evaluation index regarding the proficiency level of the performance by the user is calculated from one or more performance code strings classified into the code string group.

The performance evaluation device according to one aspect of the present disclosure includes a feature extraction unit that generates a plurality of feature series, which are time series of features related to the performance sound, for a plurality of performances by the user, and the plurality of features. From a plurality of analysis points corresponding to different combinations of a plurality of features in the first feature series and a plurality of features in the second feature series among the quantity series, the first feature series and the second feature The first series identification unit that specifies a plurality of analysis point series, which is a series of analysis points having similar feature quantities with the quantity series, and the first evaluation index regarding the proficiency level of the performance by the user are described. For each of the first index calculation unit calculated from the analysis point series and the plurality of analysis point series specified for each feature amount series, the feature amount of the analysis point for each analysis point constituting the analysis point series. The second series identification unit that generates a performance code string in which performance codes are arranged according to the above and the performance code string generated for each of the plurality of performances are one or more similar or common in the arrangement of the performance codes. From a code string classification unit that classifies into any of a plurality of code string groups that is a set of performance code strings, and from one or more performance code strings classified into the code string group for each of the plurality of code string groups. It is provided with a second index calculation unit that calculates a second evaluation index related to the proficiency level of performance by the user.

The program according to one aspect of the present disclosure includes a feature extraction unit that generates a plurality of feature series, which are time series of features related to the performance sound, for a plurality of performances by the user, and the plurality of feature series. From the plurality of analysis points corresponding to different combinations of the plurality of features in the first feature series and the plurality of features in the second feature series, the first feature series and the second feature series From the plurality of analysis point sequences, the first sequence identification unit for specifying a plurality of analysis point sequences, which is a sequence of analysis points having similar feature amounts, and the first evaluation index regarding the proficiency level of performance by the user are obtained from the plurality of analysis point sequences. For each of the plurality of analysis point series specified for each of the feature quantity series of the first index calculation unit to be calculated, for each analysis point constituting the analysis point series, a performance code corresponding to the feature quantity of the analysis point. A second series identification unit that generates a performance code string in which the above are arranged, and the performance code string generated for each of the plurality of performances is a set of one or more performance code strings having a similar or common arrangement of performance codes. A code string classification unit that classifies into any of a plurality of code string groups, and a performance by the user from one or more performance code strings classified into the code string group for each of the plurality of code string groups. The computer functions as a second index calculation unit that calculates the second evaluation index related to the proficiency level of.

It is a block diagram which illustrates the structure of the performance evaluation apparatus which concerns on one aspect. It is a block diagram which illustrates the functional structure of the performance evaluation apparatus. It is a flowchart which illustrates the specific procedure of the feature extraction process. It is explanatory drawing of the feature extraction processing. It is a flowchart which illustrates the specific procedure of the series identification process. It is explanatory drawing of the analysis matrix. It is explanatory drawing of the selection analysis point and the candidate analysis point. It is explanatory drawing about the condition of a candidate analysis point. It is a flowchart which illustrates the specific procedure of the 1st index calculation process. It is explanatory drawing of the index about the repetition of a performance. It is explanatory drawing of analysis point sequence and performance code string. It is a schematic diagram of a performance record table. It is a flowchart which illustrates the specific procedure of the 2nd index calculation process. It is a flowchart which illustrates the specific procedure of the performance evaluation process.

FIG. 1 is a block diagram illustrating the configuration of the performance evaluation device 100 according to one embodiment of the present disclosure. The performance evaluation device 100 is a signal processing device that evaluates the performance of the musical instrument 200 by the user U. The musical instrument 200 is a natural musical instrument that generates a playing sound according to an operation by the user U. For example, the user U plays various musical instruments 200 such as keyboard instruments, stringed instruments, and wind instruments. The user U repeats the performance of the music a plurality of times. By repeating the performance, the user U gradually becomes proficient in playing the music.

The performance evaluation device 100 is realized by a computer system including a control device 11, a storage device 12, a sound collecting device 13, and a display device 14. The performance evaluation device 100 is realized by an information terminal such as a mobile phone, a smartphone, or a personal computer.

The control device 11 is composed of, for example, a single or a plurality of processors that control each element of the performance evaluation device 100. For example, the control device 11 is one or more types such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), and an ASIC (Application Specific Integrated Circuit). It consists of a processor.

The display device 14 displays an image under the control of the control device 11. For example, the display device 14 displays the result of evaluating the performance of the musical instrument 200 by the user U. The sound collecting device 13 collects the performance sound generated from the musical instrument 200 by the performance of the user U.

The storage device 12 is a single or a plurality of memories composed of a recording medium such as a magnetic recording medium or a semiconductor recording medium. The storage device 12 stores, for example, a program executed by the control device 11 and various data used by the control device 11. The storage device 12 may be configured by combining a plurality of types of recording media. Further, a portable recording medium that can be attached to and detached from the performance evaluation device 100, or an external recording medium (for example, online storage) that the performance evaluation device 100 can communicate with via a communication network may be used as the storage device 12. ..

The storage device 12 stores a plurality of acoustic data X representing the waveforms of the performance sounds collected by the sound collecting device 13. One acoustic data X is stored in the storage device 12 for each performance by the user U. That is, the acoustic data X is stored for each of the plurality of performances. The start point and end point of one performance are instructed by, for example, the user U. In a configuration in which an electric musical instrument (for example, an electric guitar) that generates an electric signal according to a performance by the user U is used as the musical instrument 200, acoustic data X representing an electric signal supplied from the musical instrument 200 is stored in the storage device 12. Will be done. Therefore, the sound collecting device 13 may be omitted.

FIG. 2 is a block diagram illustrating a functional configuration of the control device 11. The control device 11 has a plurality of functions (feature extraction unit 20, first performance evaluation unit 30, second performance evaluation unit 20) for evaluating the performance by the user U by executing the program stored in the storage device 12. 40 and the display control unit 50) are realized. The function of the control device 11 may be realized by a plurality of devices configured separately from each other.

<Feature extraction unit 20>
The feature extraction unit 20 analyzes the acoustic features of the performance sound represented by each acoustic data X. FIG. 3 is a flowchart illustrating a specific procedure of the process (hereinafter referred to as “feature extraction process”) Sa executed by the feature extraction unit 20. Further, FIG. 4 is an explanatory diagram of the feature extraction process Sa.

The feature extraction unit 20 calculates the time series (that is, spectrogram) of the frequency spectrum by analyzing the acoustic data X (Sa1). A known frequency analysis such as a discrete Fourier transform is used to calculate each frequency spectrum. The feature extraction unit 20 may calculate a chromagram representing the intensity of the acoustic component for each of the 12 semitones constituting the equal temperament, instead of the frequency spectrum.

The feature extraction unit 20 identifies a plurality of pronunciation times T [n] (n = 1, 2, ...) From the time series of the frequency spectrum (Sa2). As illustrated in FIG. 4, the pronunciation time T [n] is the time when the nth performance sound is sounded. Generally, the pronunciation time T [n] corresponds to the start point of each note in the music. For example, the time at which the amplitude at an arbitrary frequency increases to a range exceeding a predetermined threshold value is specified as the sounding time T [n]. That is, even when the amplitude increases, if the amplitude after the increase falls below the threshold value, it does not correspond to the pronunciation time T [n]. Assuming that the musical instrument 200 in which the performance sound is produced immediately after the operation by the user U, the pronunciation time T [n] is the time when the user U operates the musical instrument 200 (for example, the keystroke time of the keyboard instrument or the stringed instrument). It is also called the string striking time).

The feature extraction unit 20 specifies the feature amount F [n] (Sa3). The feature amount F [n] is an acoustic feature related to the nth performance sound. Specifically, the feature extraction unit 20 selects the frequency spectrum immediately after the pronunciation time T [n] as the feature amount F [n]. That is, the feature quantity F [n] is the frequency spectrum of the nth performance sound.

As can be understood from the above explanation, in the feature extraction process Sa, the pronunciation time T [n] and the feature amount F [n] are specified for each performance of the performance sound for each of the plurality of performances by the user U. .. In the following description, the time series (F [1], F [2], F [3], ...) Of the feature quantity F [n] extracted by the feature extraction unit 20 for one performance is referred to as "feature quantity series Y". Is written. That is, the feature extraction unit 20 generates a feature sequence Y for each of a plurality of performances by the user U. Although the music is common among the plurality of acoustic data Xs, the number of pronunciation times T [n] may differ from performance to performance due to a performance error by the user U or the like. Therefore, the content and number of each feature amount F [n] constituting the feature amount series Y may differ for each feature amount series Y.

<1st performance evaluation unit 30>
The first performance evaluation unit 30 in FIG. 2 calculates the evaluation index Ea using the feature quantity series Y of each acoustic data X. The evaluation index Ea is an index showing the proficiency level of the performance by the user U. "Familiarity" means the degree of proficiency in playing. The degree of improvement in performance by practice (degree of improvement) and the degree of effect by practice of performance (hereinafter referred to as "practice effect") are also included in the concept of "proficiency". In the following description, among the plurality of feature series Y generated by the feature extraction unit 20, the feature series Y1 and the feature series Y2 corresponding to different performances will be focused on. The feature quantity series Y1 is a feature quantity sequence Y specified from the acoustic data X corresponding to the first performance, and the feature quantity series Y2 is specified from the acoustic data X corresponding to the second performance separate from the first performance. It is a feature quantity series Y. The feature quantity series Y1 is an example of the "first feature quantity series", and the feature quantity series Y2 is an example of the "second feature quantity series".

The first performance is a performance performed by the user U before the second performance. For example, the second performance is the latest performance among a plurality of performances, and the first performance is the performance immediately before the second performance. The performance of the musical instrument 200 by the user U tends to be gradually mastered by repeating the performance. Based on the above trends, it is estimated that the second performance has a higher degree of proficiency than the first performance. The evaluation index Ea is an index of the proficiency level of the second performance based on the first performance.

The first performance evaluation unit 30 includes a first series identification unit 31 and a first index calculation unit 32. The first series specifying unit 31 analyzes the relationship between the feature series Y1 and the feature series Y2. The first index calculation unit 32 calculates the evaluation index Ea from the result of the analysis by the first series identification unit 31.

FIG. 5 is a flowchart illustrating a specific procedure of a process (hereinafter referred to as “series identification process”) Sb in which the first series identification unit 31 analyzes the relationship between the feature quantity sequence Y1 and the feature quantity sequence Y2.

The first sequence identification unit 31 generates an analysis matrix M from the feature sequence Y1 and the feature sequence Y2 (Sb1). FIG. 6 is a schematic diagram of the analysis matrix M. The feature series Y1 is composed of N1 feature quantities F [1] to F [N1], and the feature quantity series Y2 is composed of N2 feature quantities F [1] to F [N2]. The analysis matrix M corresponds to a combination of N1 feature quantities F [1] to F [N1] of the feature quantity series Y1 and N2 feature quantities F [1] to F [N2] of the feature quantity series Y2. It is a matrix of N2 rows × N1 columns in which a similarity index D [n1, n2] is set for each analysis point P [n1, n2] (n1 = 1 to N1, n2 = 1 to N2). The analysis points P [n1, n2] are each of N1 features F [1] to F [N1] of the feature series Y1 and N2 features F [1] to F [N2] of the feature series Y2. ] Is a virtual point corresponding to a different combination with each of. The horizontal direction in the analysis matrix M corresponds to the time axis of the feature series Y1, and the vertical direction in the analysis matrix M corresponds to the time axis of the feature series Y2.

The similarity index D [n1, n2] is an index of similarity between the feature amount F [n1] of the feature amount series Y1 and the feature amount F [n2] of the feature amount series Y2. Specifically, the similarity index D [n1, n2] is, for example, the distance between the feature amount F [n1] and the feature amount F [n2] (for example, the Euclidean distance). Therefore, the more similar the feature amount F [n1] and the feature amount F [n2] are to each other, the smaller the similarity index D [n1, n2] becomes. Since the music is common between the feature series Y1 and the feature series Y2, roughly, in the arrangement of the analysis points P [n1, n2] from the lower left to the upper right of the analysis matrix M, the similarity index D [n1 , n2] tends to be a small number. As illustrated below, the first series identification unit 31 is a series of analysis points P [n1, n2] in which the feature quantity F [n] is similar between the feature quantity series Y1 and the feature quantity series Y2 (hereinafter, "" (Analysis point series) "Z is specified (Sb2 to Sb7).

First, the first series identification unit 31 selects one analysis point (hereinafter referred to as “selection analysis point”) P [n1, n2] from the analysis matrix M (Sb2). As illustrated in FIG. 7, the first series identification unit 31 is derived from a plurality of analysis points P in a predetermined range (hereinafter referred to as “analysis range”) A based on the selected analysis point P [n1, n2]. In the analysis point series Z, a plurality of candidate analysis points (hereinafter referred to as “candidate analysis points”) P that follow the selected analysis points P [n1, n2] are specified (Sb3). The analysis range A is located behind the selective analysis point P [n1, n2] in the feature sequence Y1 and a predetermined length range a1 located behind the selective analysis point P [n1, n2] in the feature sequence Y2. It is a range in which the range a2 having a predetermined length overlaps with each other.

Specifically, the first series identification unit 31 sets the analysis points P [m1, m2] that satisfy the following conditions among the plurality of analysis points P in the analysis range A as candidate analysis points P [m1, m1]. Specify as m2]. As illustrated in FIG. 8, the interval Δ [n1, m1] is the pronunciation time T [m1] corresponding to the analysis point P [m1, m2] in the analysis range A in the feature series Y1 and the selection analysis point. This is the interval from the pronunciation time T [n1] corresponding to P [n1, n2]. Similarly, the interval Δ [n2, m2] is the pronunciation time T [m2] corresponding to the analysis point P [m1, m2] in the analysis range A in the feature quantity series Y2, and the selective analysis point P [n1, n2]. ] Is the interval with the pronunciation time T [n2] corresponding to.
Condition 1: m1-n1 ≦ 2
Condition 2: m2-n2 ≦ 2
Condition 3: Δ [n1, m1] = T [m1] −T [n1] ≧ 30 msec
Condition 4: Δ [n1, m1] ≤ 4000 msec
Condition 5: Δ [n2, m2] = T [m2] −T [n2] ≧ 30 msec
Condition 6: Δ [n2, m3] ≤ 4000 msec
Condition 7: D [n1, n2] ≤ 0.3
Condition 8: D [m1, m2] ≤ 0.3

The first series identification unit 31 calculates the analysis index R for each of the plurality of candidate analysis points P [m1, m2] (Sb4). The analysis index R is expressed by, for example, the following mathematical formula (1).

As understood from the mathematical formula (1), the larger the distance between the selective analysis point P [n1, n2] and the candidate analysis point P [m1, m2], the smaller the analysis index R becomes. That is, the larger the candidate analysis point P [m1, m2] that is correspondingly closer to the selective analysis point P [n1, n2], the larger the analysis index R becomes. Further, the larger the similarity index D [n1, n2] of the selection analysis point P [n1, n2] or the similarity index D [m1, m2] of the candidate analysis point P [m1, m2], the smaller the analysis index R becomes. .. That is, the more similar the feature amount F [n1] (F [m1]) of the feature amount series Y1 and the feature amount F [n2] (F [m2]) of the feature amount series Y2 are, the larger the analysis index R becomes. Become. The first series identification unit 31 selects one candidate analysis point P [m1, m2] having the maximum analysis index R among the plurality of candidate analysis points P [m1, m2] as a selection analysis point P [n1, n2]. ] Is selected as the analysis point P following (Sb5).

The first series identification unit 31 determines whether or not the above processing has been executed for all the analysis points P in the vertical N2 row × horizontal N1 column (Sb6). When there is an unprocessed analysis point P (Sb6: NO), the first series identification unit 31 sets an analysis point P different from the current selection analysis point P [n1, n2] to a new selection analysis point P [n1]. , N2] is selected (Sb2), and processing (Sb3 to Sb5) for selecting a subsequent analysis point P for the updated selected analysis point P [n1, n2] is executed. As understood from the above description, for each of the plurality of analysis points P in the vertical N2 row × horizontal N1 column, a pair of the analysis point P and another analysis point P following the analysis point P is set. .. In addition, there are some analysis points such as the upper right analysis point P [N1, N2] in the analysis matrix M, for which a pair with another analysis point P [n1, n2] is not set.

On the other hand, when the process is executed for all the analysis points P (Sb6: YES), the first sequence identification unit 31 identifies the analysis point sequence Z by connecting the sets set for each analysis point P to each other (Sb6: YES). Sb7). For example, when a set of analysis points P [1,1] and analysis points P [2,2] and a set of analysis points P [2,2] and analysis points P [3,3] are set, analysis is performed. An analysis point sequence Z in which the points P [1,1], the analysis points P [2,2], and the analysis points P [3,3] are arranged is specified.

By the sequence identification processing Sb exemplified above, a plurality of analysis point sequences Z which are a sequence of analysis points P having similar feature quantities F [n] between the feature quantity sequence Y1 and the feature quantity sequence Y2 are specified. As can be understood from the above description, the first series identification unit 31 has N1 feature quantities F [1] to F [N1] in the feature quantity series Y1 and N2 feature quantities F [1] in the feature quantity series Y2. ] ~ F [N2] From multiple analysis points P (vertical N2 rows x horizontal N1 columns) corresponding to different combinations, the feature amount F [n] is between the feature amount series Y1 and the feature amount series Y2. A plurality of analysis point sequences Z, which are sequences of similar analysis points P, are specified.

The first index calculation unit 32 of FIG. 2 calculates the index Ea from the plurality of analysis point series Z specified by the series identification process Sb. FIG. 9 is a flowchart illustrating a specific procedure of a process in which the first index calculation unit 32 calculates the index Ea (hereinafter referred to as “first index calculation process”). The evaluation index Ea includes four types of indexes Ea1 to Ea4.

The first index calculation unit 32 counts the number K of the analysis points P constituting each analysis point series Z, and calculates the index Ea1 according to the representative value of the number K in the plurality of analysis point series Z (Sc1). Specifically, the average value (for example, simple average or weighted average) of the number K of the analysis points P over the plurality of analysis point series Z is calculated as the index Ea1. The number K of the analysis points P constituting each analysis point series Z is small at the beginning when the user U starts practicing the music, and tends to increase as the user U becomes proficient in playing the music. Therefore, the evaluation index Ea1 is used as an index that appropriately represents the proficiency level of the performance by the user U. Specifically, the larger the index Ea1, the more proficient the user U can be evaluated in playing the music.

The first index calculation unit 32 calculates the gradient α of the straight line represented by each analysis point series Z (Sc2). The gradient α with respect to one analysis point sequence Z is a numerical value obtained by averaging the gradient (slope) between two analysis points P adjacent to each other over the entire analysis point sequence Z. Therefore, the faster the playing speed of the second performance is higher than the playing speed of the first performance, the smaller the gradient α becomes.

The first index calculation unit 32 calculates the index Ea2 according to the representative value of the gradient α in the plurality of analysis point series Z (Sc3). Specifically, the average value (for example, simple average or weighted average) of the gradient α in the plurality of analysis point series Z is calculated as the index Ea2. As understood from the above explanation, the index Ea2 is an index related to the degree of change in the playing speed. In general musical instrument practice, there is a tendency to start with a tempo that is slow enough to play comfortably, and gradually approach the target tempo (for example, the tempo specified in the song) to become more proficient. Therefore, the gradient α of each analysis point series Z tends to decrease as the performance of the music by the user U improves. Therefore, the index Ea2 is used as an index appropriately expressing the degree of improvement of the performance by the user U from the viewpoint of the change in the playing speed. For example, it can be evaluated that the smaller the index Ea2 is, the greater the degree of improvement in the performance of the music by the user U.

Further, the first index calculation unit 32 calculates the index Ea3 according to the degree of dispersion of the gradient α in the plurality of analysis point series Z (Sc4). The degree of dispersion is a statistic for the dispersion of the gradient α, such as standard deviation or variance. Specifically, the first index calculation unit 32 calculates the reciprocal of the dispersal degree of the gradient α in the plurality of analysis point series Z as the index Ea3. The more the user U becomes proficient in playing music, the more stable the playing speed tends to be. That is, as the user U becomes more proficient in playing, the degree of dispersion of the gradient α (that is, the fluctuation of the playing speed) tends to decrease. Therefore, the index Ea3 is used as an index that appropriately represents the proficiency level of the performance by the user U from the viewpoint of the stability of the performance speed. For example, the larger the index Ea3, the more proficient the user U can be evaluated in playing the music.

As illustrated in FIG. 10, for a specific section y1 in the feature sequence Y1, when the analysis point sequence Z is specified between a plurality of sections y2 in the feature sequence Y2, the section y1 in the music is , In the second performance, it means that the performance was repeated multiple times. The first index calculation unit 32 calculates the index Ea4 according to the number of sections y2 in the feature series Y2 in which the analysis point series Z is specified with the specific section y1 in the feature series Y1 (Sc5). .. Specifically, the number of the section y2 is calculated as the index Ea4. The more the user U repeats the performance of a specific section in the music, the easier it is for the user U to become proficient in the performance of the music. Therefore, the index Ea4 is used as an index that appropriately represents the practice effect of the performance by the user U from the viewpoint of repeating the performance with respect to a specific section in the music. For example, it can be evaluated that the larger the index Ea4 is, the higher the effect of practicing the performance by the user U is.

When the evaluation indexes Ea (Ea1 to Ea4) are calculated by the above procedure, the display control unit 50 causes the display device 14 to display the evaluation indexes Ea (Ea1 to Ea4). The user U can refer to the index Ea for each of the plurality of viewpoints relating to the performance of the music. As can be understood from the above explanation, the evaluation index Ea used the acoustic data X relating to the first performance and the second performance to evaluate the performance proficiency (including the degree of progress and the practice effect) of the entire music. It is an index.

<Second performance evaluation unit 40>
The second performance evaluation unit 40 in FIG. 2 uses the feature sequence Y of each acoustic data X and the analysis point sequence Z generated by the first performance evaluation unit 30 (specifically, the first performance evaluation unit 31). The evaluation index Eb is calculated. The evaluation index Eb, like the evaluation index Ea, is an index indicating the proficiency level of the performance by the user U. The second performance evaluation unit 40 includes a second series identification unit 41, a code string classification unit 42, and a second index calculation unit 43.

The second sequence specifying unit 41 generates a performance code string W for each of the plurality of analysis point sequences Z specified by the first sequence specifying unit 31. A performance code string W is generated for each of the plurality of analysis point series Z specified from the feature sequence Y of each of the plurality of performances.

FIG. 11 is a schematic diagram for explaining the relationship between the analysis point sequence Z and the performance code string W. The performance code string W corresponding to one analysis point sequence Z is an array of performance codes Q arranged for each analysis point P constituting the analysis point sequence Z. The performance code Q arranged for each analysis point P is a code corresponding to the feature amount F [n] corresponding to the analysis point P. Specifically, a symbol for abstracting each of the N2 feature quantities F [1] to F [N2] of the feature quantity series Y2 is used as the performance code Q. The performance code Q is, for example, 12 note names (C, C #, D, D #, E, F, F #, G, G #, A, A #, B) constituting equal temperament. As can be understood from the above description, a plurality of feature quantities F [n2] that are similar to each other in the analysis point sequence Z are replaced with a common performance code Q in the performance code string W. The performance code string W is also paraphrased as a time series (that is, a specific melody) of the note names played by the user U.

The code string classification unit 42 of FIG. 2 classifies the performance code string W generated by the second series identification unit 41 for each of the plurality of performances into any of the plurality of code string groups G. Each of the plurality of code string groups G is a set of one or more performance code strings W similar to or common to a specific array of performance code Q (hereinafter referred to as "performance phrase"). That is, the performance code string W when the user U plays a specific performance phrase is classified into the common code string group G corresponding to the performance phrase.

Specifically, the code string classification unit 42 generates the performance recording table Tb illustrated in FIG. 12 in the storage device 12. The performance recording table Tb is composed of a plurality of unit data g corresponding to different code string groups G (G1, G2, ...). In the unit data g corresponding to any one code string group G, in addition to the performance phrase (array of performance code Q) corresponding to the code string group G, the number of elements ν, the number of occurrences λ, and the importance ω And are registered. The number of elements ν of each code string group G is the number of performance codes Q constituting the performance phrase corresponding to the code string group G (ν ≧ 1).

The number of occurrences λ of each code string group G is the number of times the performance phrase of the code string group G appears in a plurality of performances by the user U (λ ≧ 1). The number of occurrences λ is also paraphrased as the number of times the user U repeats the performance of the performance phrase in a plurality of performances. The code string classification unit 42 increases the number of occurrences λ of the code string group G by 1 each time one performance code string W specified by the second series identification unit 41 is classified into each code string group G.

The importance ω of each code string group G is an index of importance regarding the performance phrase of the code string group G. The importance ω is used, for example, to evaluate whether or not the performance phrase is intentionally repeated by the performer. The importance ω is calculated according to the number of elements ν and the number of occurrences λ. For example, the importance ω is expressed as a power of the number of elements ν whose number of occurrences λ is a power (ω = ν ^λ ). That is, the importance ω increases monotonically with respect to the number of elements ν and the number of occurrences λ.

The second index calculation unit 43 calculates the evaluation index Eb indicating the proficiency level of the performance by the user U for each of the plurality of code string groups G. Specifically, the second index calculation unit 43 evaluates using the analysis point series Z generated by the first series identification unit 31 and the classification result (performance record table Tb) by the code string classification unit 42. The index Eb is calculated for each code string group G. The evaluation index Eb includes four types of indexes Eb1 to Eb4. FIG. 13 is a flowchart illustrating a specific procedure of the process (hereinafter referred to as “second index calculation process”) Se in which the second index calculation unit 43 calculates the evaluation index Eb.

The second index calculation unit 43 calculates the index Eb1 for each code string group G according to the importance ω registered in the performance record table Tb (Se1). The second index calculation unit 43 calculates the index Eb1 of the code string group G by a predetermined calculation for the importance ω of each code string group G. The importance ω may be used as the index Eb1.

The second index calculation unit 43 calculates the gradient α of the analysis point sequence Z corresponding to each performance code string W classified into the code string group G for each of the plurality of code string groups G (Se2). The gradient α calculated from the analysis point sequence Z corresponding to each code string group G is an index relating to the performance speed at which the user U plays the performance code string W belonging to the code string group G. Specifically, the faster the performance speed of the performance phrase in the second performance is faster than the performance speed of the performance phrase in the first performance, the smaller the gradient α becomes.

The second index calculation unit 43 corresponds to the representative value of the gradient α in one or more analysis point series Z corresponding to one or more performance code strings W of the code string group G for each of the plurality of code string groups G. The index Eb2 is calculated (Se3). Specifically, the average value (for example, simple average or weighted average) of the gradient α in one or more analysis point series Z is calculated as the index Eb2. Therefore, the faster the playing speed of the playing phrase in the second performance is faster than the playing speed of the playing phrase in the first performance, the smaller the gradient α (further, the index Ea2) becomes.

As understood from the above description, the index Eb2 is an index relating to the degree of change in the playing speed with respect to the playing phrase corresponding to the code string group G. As described above, the gradient α of each analysis point series Z tends to decrease as the user U becomes proficient in playing the performance phrase. Therefore, the index Eb2 is used as an index that appropriately represents the proficiency level (particularly the degree of improvement) of the performance by the user U from the viewpoint of the change in the performance speed for each performance phrase. For example, it can be evaluated that the smaller the index Eb2 is, the greater the degree of improvement in the performance of the performance phrase by the user U.

Further, the second index calculation unit 43 disperses the gradient α in one or more analysis point series Z corresponding to one or more performance code strings W of the code string group G for each of the plurality of code string groups G. The index Eb3 is calculated according to (Se4). Specifically, the second index calculation unit 43 calculates the reciprocal of the dispersal degree of the gradient α in the analysis point series Z of 1 or more as the index Eb3. The more the user U becomes proficient in playing the playing phrase, the more stable the playing speed tends to be. That is, as the user U becomes more proficient in playing the performance phrase, the degree of dispersion of the gradient α (that is, the fluctuation of the performance speed) tends to decrease. Therefore, the index Eb3 is used as an index that appropriately represents the proficiency level of the performance by the user U from the viewpoint of the stability of the performance speed for each performance phrase. For example, the larger the index Eb3, the more proficient the user U is in playing the performance phrase.

The second index calculation unit 43 calculates the index Eb4 for each of the plurality of code string groups G according to the number of performance code strings W classified into the code string group G (Se5). Specifically, the second index calculation unit 43 specifies the number of performance code strings W classified in the code string group G as the index Eb4 for each code string group G. The more the performance of the performance phrase is repeated, the easier it is for the user U to become proficient in playing the performance phrase. Therefore, the index Eb4 is used as an index that appropriately expresses the practice effect of the performance by the user U from the viewpoint of repeating the performance for each performance phrase. For example, it can be evaluated that the larger the index Eb4 is, the higher the effect of practicing the performance by the user U is.

When the evaluation indexes Eb (Eb1 to Eb4) are calculated by the above procedure, the display control unit 50 causes the display device 14 to display the evaluation indexes Eb for each code string group G (that is, for each performance phrase). The user U can refer to the index Eb for each performance phrase for each of the plurality of viewpoints relating to the performance of the music. As can be understood from the above explanation, the evaluation index Eb is an index for evaluating the proficiency level for each performance phrase using the acoustic data X of the performance over a plurality of times.

FIG. 14 is a flowchart illustrating a specific procedure of a process in which the performance evaluation device 100 evaluates the performance of the user U (hereinafter referred to as “performance evaluation process”). The performance evaluation process is started every time the user U plays the musical instrument 200 (that is, every time the acoustic data X is generated). In the performance evaluation process, the latest performance among a plurality of performances is regarded as the second performance, and the performance immediately before the second performance is regarded as the first performance.

When the performance evaluation process is started, the feature extraction unit 20 generates a feature sequence Y2 from the acoustic data X of the second performance by the feature extraction process Sa illustrated in FIG. The first performance evaluation unit 30 generates a plurality of analysis point sequences Z from the feature quantity sequence Y1 of the first performance and the feature quantity sequence Y2 of the second performance by the sequence identification process Sb illustrated in FIG. The first index calculation unit 32 calculates the evaluation index Ea (Ea1 to Ea4) for evaluating the proficiency level of the second performance based on the first performance by the first index calculation processing Sc illustrated in FIG. ..

The second sequence identification unit 41 generates a performance code sequence W for each of the plurality of analysis point sequence Z identified by the sequence identification process Sb (Sd1). Further, the code string classification unit 42 sets the performance code string W generated by the second series identification unit 41 for each of the plurality of performances by the user U as any of the plurality of code string groups G corresponding to different performance phrases. Classify as (Sd2). The second index calculation unit 43 calculates the evaluation index Eb (Eb1 to Eb4) for each of the plurality of code string groups G by the second index calculation processing Se illustrated in FIG.

The display control unit 50 causes the display device 14 to display the evaluation indexes Ea (Ea1 to Ea4) calculated by the first index calculation unit 32 and the evaluation indexes Eb (Eb1 to Eb4) calculated by the second index calculation unit 43 ( Sf). The four types of indexes Ea1 to Ea4 constituting the evaluation index Ea and the four types of indexes Eb1 to Eb4 constituting the evaluation index Eb are displayed in parallel on the display device 14. The evaluation index Ea and the evaluation index Eb may be selectively displayed on the display device 14. The evaluation index Ea is an example of the "first evaluation index", and the evaluation index Eb is an example of the "second evaluation index".

As described above, among the multiple performances by the user U, a plurality of analysis point sequences Z are specified from the feature quantity sequence Y1 related to the first performance and the feature quantity sequence Y2 related to the second performance, and a plurality of analysis point sequences. Using Z, the evaluation index (Ea, Eb) regarding the proficiency level of the performance by the user U is calculated. Therefore, there is an advantage that it is not necessary to prepare an acoustic signal representing a model performance in advance.

The evaluation index Ea is calculated according to the result of analyzing the relationship between the feature quantity series Y1 and the feature quantity series Y2 corresponding to different performances among a plurality of performances. Therefore, it is possible to calculate the evaluation index Ea that appropriately represents the overall tendency of the performance between two different performances. Further, the performance code string W generated from each analysis point sequence Z for each of the plurality of performances is classified into any of the plurality of code string groups G, and the evaluation index Eb is calculated for each code string group G. Therefore, there is also an advantage that the evaluation index Eb that appropriately represents the tendency of each code string group G (that is, each performance phrase) in a plurality of performances can be calculated.

<Modification example>
Specific modifications added to each of the above-exemplified embodiments will be illustrated below. Two or more embodiments arbitrarily selected from the following examples may be appropriately merged to the extent that they do not contradict each other.

(1) In the above-described form, the latest performance among the multiple performances by the user U is set as the second performance, and the performance immediately before the second performance is set as the first performance. However, the first performance and the second performance are selected. The method is not limited to the above examples. For example, the first performance of a plurality of performances may be the first performance. Of the plurality of performances, the performance selected by the user U as a reference may be the first performance, and the performance selected by the user U as the evaluation target may be the second performance. As can be understood from the above examples, the first performance and the second performance are not limited to performances that are in tandem with each other. Further, the user U may select the first performance and the second performance from a plurality of performances. Further, the first performance, which is a reference for the second performance, is not limited to the performance of the user U himself. For example, a performance by a performance expert may be used as the first performance.

(2) In the above-described embodiment, each of the plurality of note names (12 semitones) is used as the performance code Q, but the performance code Q is not limited to the above examples. For example, in addition to the note name of a single note, the note name of the constituent notes of the chord (for example, CEG or the like) may be used as the performance code Q. Further, a combination of the acoustic intensity represented by the feature amount F [n] and the note name of a single note or chord (for example, C10, C5E3G2, etc.) may be used as the performance code Q. The performance code Q does not have to be a note name. For example, an identification number (for example, a serial number assigned to each set) assigned to each of a plurality of sets in which the feature quantity F [n] is classified (clustered) may be used as the performance code Q.

(3) In the above-described embodiment, the distance between the feature amount F [n1] and the feature amount F [n2] is illustrated as the similarity index D [n1, n2], but the method for calculating the similarity index D [n1, n2] Is not limited to the above examples. For example, the correlation between the feature amount F [n1] and the feature amount F [n2] may be calculated as the similarity index D [n1, n2]. In a configuration in which the correlation between the feature amount F [n1] and the feature amount F [n2] is set as the similarity index D [n1, n2], the similarity index is obtained so that the feature amount F [n1] and the feature amount F [n2] are similar. D [n1, n2] is a large value. Therefore, the sequence of analysis points P having a large similarity index D [n1, n2] is specified as the analysis point sequence Z by the sequence identification process Sb.

(4) In the above-described form, the four indexes Ea1 to Ea4 constituting the evaluation index Ea are calculated and displayed, but one evaluation index Ea corresponding to the four indexes Ea1 to Ea4 is calculated and displayed. May be good. For example, the weighted sum of the four indexes Ea1 to Ea4 is calculated as the evaluation index Ea. Similarly, for the evaluation index Eb, one evaluation index Eb corresponding to the four indexes Eb1 to Eb4 may be calculated and displayed. For example, the weighted sum of the four indexes Eb1 to Eb4 is calculated as the evaluation index Eb.

(5) In the above-described form, both the evaluation index Ea and the evaluation index Eb are calculated, but only one of the evaluation index Ea and the evaluation index Eb may be calculated. In the configuration in which only the evaluation index Eb is calculated, the first index calculation unit 32 illustrated in the above-described embodiment is omitted, and the second performance evaluation unit 40 is the feature quantity series Y of each acoustic data X and the first series identification unit 31. The evaluation index Eb is calculated using the analysis point series Z generated by. On the other hand, in the configuration in which only the evaluation index Ea is calculated, the second performance evaluation unit 40 illustrated in each of the above-described forms is omitted, and the first performance evaluation unit 30 uses the feature quantity series Y of each acoustic data X. Calculate the evaluation index Ea.

(6) Although the performance of the musical instrument 200 by the user U has been evaluated, the performance to be evaluated by the performance evaluation device 100 is not limited to the performance of the musical instrument 200. The "performance" in the present disclosure is a performance in a narrow sense that includes a musical sound produced by operating the musical instrument 200, as well as a singing that the user U utters due to the vibration of the vocal cords.

(7) The performance evaluation device 100 may be realized by a server device that communicates with a terminal device such as a mobile phone or a smartphone. For example, the performance evaluation device 100 executes a performance evaluation process on a plurality of acoustic data X received from the terminal device, and transmits the evaluation index Ea and the evaluation index Eb calculated by the performance evaluation process to the requesting terminal device. To do. The evaluation index Ea and the evaluation index Eb are displayed on the terminal device.

(8) As described above, the function of the performance evaluation device 100 illustrated above is realized by the cooperation of the single or a plurality of processors constituting the control device 11 and the program stored in the storage device 12. The program according to the present disclosure may be provided and installed on a computer in a form stored in a computer-readable recording medium. The recording medium is, for example, a non-transitory recording medium, and an optical recording medium (optical disc) such as a CD-ROM is a good example, but a known arbitrary such as a semiconductor recording medium or a magnetic recording medium. Recording media in the format of are also included. The non-transient recording medium includes any recording medium except for a transient propagation signal (transitory, propagating signal), and a volatile recording medium is not excluded. Further, in the configuration in which the distribution device distributes the program via the communication network, the storage device that stores the program in the distribution device corresponds to the above-mentioned non-transient recording medium.

<Additional notes>
From the forms exemplified above, for example, the following configuration can be grasped.

In the performance evaluation method according to one aspect (first aspect) of the present disclosure, a plurality of feature quantity sequences, which are time series of feature quantities related to the performance sound, are generated for each of a plurality of performances by the user, and the plurality of performance evaluation methods are generated. From the plurality of analysis points corresponding to different combinations of the plurality of feature quantities in the first feature quantity series and the plurality of feature quantities in the second feature quantity series among the feature quantity series, the first feature quantity series and the second feature quantity series. A plurality of analysis point sequences, which are a series of analysis points having similar feature amounts with the feature quantity series, are specified, and a first evaluation index regarding the proficiency level of performance by the user is calculated from the plurality of analysis point sequences. Then, for each of the plurality of analysis point sequences specified for each feature quantity series, a performance code string in which performance codes corresponding to the feature quantities of the analysis points are arranged for each analysis point constituting the analysis point series. Is generated, and the performance code string generated for each of the plurality of performances is classified into one of a plurality of code string groups which is a set of one or more performance code strings having a similar or common arrangement of performance codes. Then, for each of the plurality of code string groups, a second evaluation index regarding the proficiency level of the performance by the user is calculated from one or more performance code strings classified into the code string group.

According to the above aspect, a plurality of analysis point sequences are specified from the first feature quantity series related to one performance and the second feature quantity series related to the other performance among the multiple performances by the user, and a plurality of analysis points are specified. Using the series, the first evaluation index and the second evaluation index regarding the proficiency level of the performance by the user are calculated. Therefore, there is an advantage that it is not necessary to prepare an acoustic signal representing a model performance in advance. In addition, the first evaluation index is calculated according to the result of analyzing the relationship between the first feature quantity series and the second feature quantity series corresponding to different performances among a plurality of performances. Therefore, it is possible to calculate a first evaluation index that appropriately represents the overall tendency of a performance between two different performances. Further, the performance code string generated from each analysis point sequence for each of a plurality of performances is classified into one of a plurality of code string groups, and a second evaluation index is assigned to each code string group according to the result of the classification. It is calculated. Therefore, there is also an advantage that the second evaluation index that appropriately represents the tendency of each code string group in a plurality of performances can be calculated.

In the specific example of the first aspect (second aspect), in the calculation of the first evaluation index, the number of the analysis points constituting each analysis point series is counted, and the number of the analysis points in the plurality of analysis point series is counted. The first evaluation index is calculated according to the representative value. As the user becomes more proficient in playing, the number of analysis points constituting one analysis point series tends to increase. Therefore, according to the above aspect, the first evaluation index that appropriately represents the proficiency level of the performance by the user can be calculated.

In the specific example (third aspect) of the first aspect or the second aspect, in the calculation of the first evaluation index, the gradient of each analysis point series is calculated, and the representative value of the gradient in the plurality of analysis point series. The first evaluation index is calculated according to the above. The gradient of the analysis point series is used as an index of the relationship between the performance speed of the performance related to the first feature series and the performance speed of the performance related to the second feature series. On the other hand, as the performance by the user improves, the playing speed tends to increase. That is, the gradient of the analysis point series tends to decrease as the performance by the user improves. Therefore, according to the above aspect, the first evaluation index that appropriately represents the proficiency level of the performance by the user (particularly the degree of improvement in the performance) can be calculated from the viewpoint of the change in the performance speed.

In any specific example (fourth aspect) of the first to third aspects, in the calculation of the first evaluation index, the gradient of each analysis point series is calculated, and the gradient in the plurality of analysis point series. The first evaluation index is calculated according to the degree of spraying. As described above, the gradient of the analysis point series is used as an index of the relationship between the performance speed of the performance related to the first feature series and the performance speed of the performance related to the second feature series. On the other hand, as the user becomes more proficient in playing, the fluctuation (variation) in the playing speed tends to be suppressed. That is, as the user becomes more proficient in playing, the degree of dispersion of the gradient of the analysis point series tends to decrease. Therefore, according to the above aspect, the first evaluation index that appropriately represents the proficiency level of the performance by the user can be calculated from the viewpoint of the stability of the performance speed.

In any specific example (fifth aspect) of the first to fourth aspects, in the calculation of the first evaluation index, a plurality of specific sections in the first feature amount series are used in the second feature amount series. When the analysis point series is specified between the two sections, the first evaluation index is calculated according to the number of the sections in the second feature series. When an analysis point series is specified between a specific section in the first feature quantity series and each of a plurality of sections in the second feature quantity series, the section is included in the performance related to the first feature quantity series. It means that the performance was repeated multiple times. Therefore, according to the above aspect, the first evaluation index that appropriately represents the proficiency level (particularly the practice effect) of the performance by the user can be calculated from the viewpoint of repeating the practice for a specific section of the music.

In any specific example (sixth aspect) of the first to fifth aspects, in the calculation of the second evaluation index, each of the plurality of code string groups is classified into each performance. The gradient of the analysis point sequence corresponding to the code string is calculated, and the second is performed according to the representative value of the gradient in one or more analysis point sequences corresponding to one or more performance code sequences classified in the code string group. Calculate the evaluation index. The gradient of the analysis point series is used as an index of the performance speed of the performance. On the other hand, as the performance by the user improves, the playing speed tends to increase. That is, the gradient of the analysis point series tends to decrease as the performance by the user improves. Therefore, according to the above aspect, it is possible to calculate the second evaluation index that appropriately represents the proficiency level of the performance by the user (particularly the degree of improvement in the performance) for each code string group from the viewpoint of the change in the performance speed.

In any specific example (seventh aspect) of the first to sixth aspects, in the calculation of the second evaluation index, for each of the plurality of code string groups, each performance classified into the code string group. The gradient of the analysis point sequence corresponding to the code string is calculated, and the second is performed according to the degree of dispersion of the gradient in one or more analysis point sequences corresponding to one or more performance code sequences classified in the code string group. Calculate the evaluation index. The gradient of the analysis point series is used as an index of the performance speed of the performance. On the other hand, as the user becomes more proficient in playing, the fluctuation (variation) in the playing speed tends to be suppressed. That is, as the user becomes more proficient in playing, the degree of dispersion of the gradient of the analysis point series tends to decrease. Therefore, according to the above aspect, it is possible to calculate the second evaluation index that appropriately represents the proficiency level of the performance by the user for each code string group from the viewpoint of the stability of the performance speed.

In any specific example (eighth aspect) of any one of the first to seventh aspects, in the calculation of the second evaluation index, each of the plurality of code string groups is a performance code classified into the code string group. The second evaluation index is calculated according to the number of columns. According to the above aspect, it is possible to calculate a second evaluation index that appropriately represents the proficiency level (particularly the practice effect) of the performance by the user for each code string group from the viewpoint of repeating the practice for a specific section of the music.

The performance evaluation device according to one aspect (9th aspect) of the present disclosure includes a feature extraction unit that generates a plurality of feature quantity series, which are time series of feature quantities related to the performance sound, for a plurality of performances by the user. From a plurality of analysis points corresponding to different combinations of a plurality of feature quantities in the first feature quantity series and a plurality of feature quantities in the second feature quantity series among the plurality of feature quantity series, the first feature quantity series A first series identification unit that specifies a plurality of analysis point sequences that are a series of analysis points having similar feature quantities between the second feature quantity series and the second feature quantity series, and a first evaluation index regarding the proficiency level of performance by the user. For each of the first index calculation unit that calculates from the plurality of analysis point series and each of the plurality of analysis point series specified for each feature quantity series, for each analysis point constituting the analysis point series, the said The arrangement of the performance codes is similar or similar to the second series identification unit that generates the performance code sequence in which the performance codes are arranged according to the feature amount of the analysis point and the performance code string generated for each of the plurality of performances. A code string classification unit that classifies into any of a plurality of code string groups that is a set of one or more common performance code strings, and one or more code string groups classified into the code string group for each of the plurality of code string groups. It is provided with a second index calculation unit that calculates a second evaluation index related to the proficiency level of performance by the user from the performance code string.

The program according to one aspect (10th aspect) of the present disclosure includes a feature extraction unit that generates a plurality of feature quantity sequences, which are time series of feature quantities related to performance sounds, for a plurality of performances by a user. From a plurality of analysis points corresponding to different combinations of a plurality of features in the first feature series and a plurality of features in the second feature series among the feature series of the above, the first feature series and the first feature series. The first series identification unit that specifies a plurality of analysis point sequences that are a series of analysis points having similar feature quantities with the two feature quantity series, and the plurality of first evaluation indexes related to the proficiency level of performance by the user. For each of the plurality of analysis point series specified for each of the feature quantity series and the first index calculation unit calculated from the analysis point series of the above, the feature quantity of the analysis point for each analysis point constituting the analysis point series. A second series identification unit that generates a performance code string in which performance codes are arranged according to the above, and one or more performances in which the arrangement of the performance codes is similar or common to the performance code strings generated for each of the plurality of performances. From the code string classification unit that classifies into any of a plurality of code string groups that is a set of code strings, and from one or more performance code strings classified into the code string group for each of the plurality of code string groups. The computer functions as a second index calculation unit that calculates a second evaluation index related to the user's proficiency in playing.

100 ... Performance evaluation device, 200 ... Musical instrument, 11 ... Control device, 12 ... Storage device, 13 ... Sound collection device, 14 ... Display device, 20 ... Feature extraction unit, 30 ... First performance evaluation unit, 31 ... First series Specific unit, 32 ... 1st index calculation unit, 40 ... 2nd performance evaluation unit, 41 ... 2nd series identification unit, 42 ... code string classification unit, 43 ... 2nd index calculation unit, 50 ... display control unit.

Claims (10)

For multiple performances by the user, multiple feature quantity series, which are time series of feature quantities related to the performance sound, are generated.
From the plurality of analysis points corresponding to different combinations of the plurality of features in the first feature series and the plurality of features in the second feature series among the plurality of feature series, the first feature series and the first feature series A plurality of analysis point series, which are a series of analysis points having similar feature amounts with the second feature amount series, are specified.
The first evaluation index regarding the proficiency level of the performance by the user is calculated from the plurality of analysis point series, and is calculated.
For each of the plurality of analysis point sequences specified for each feature quantity series, a performance code string in which performance codes corresponding to the feature quantities of the analysis points are arranged is generated for each analysis point constituting the analysis point series. And
The performance code string generated for each of the plurality of performances is classified into one of a plurality of code string groups which are a set of one or more performance code strings having a similar or common arrangement of performance codes.
A performance evaluation method realized by a computer for calculating a second evaluation index regarding the proficiency level of performance by the user from one or more performance code strings classified into the code string group for each of the plurality of code string groups. .. In the calculation of the first evaluation index, the number of the analysis points constituting each analysis point series is counted, and the first evaluation index is calculated according to the representative value of the number in the plurality of analysis point series. The performance evaluation method of claim 1. In the calculation of the first evaluation index, claim 1 or claim in which the gradient of each analysis point series is calculated and the first evaluation index is calculated according to the representative value of the gradient in the plurality of analysis point series. Performance evaluation method of 2. In the calculation of the first evaluation index, the gradient of each analysis point series is calculated, and the first evaluation index is calculated according to the degree of dispersion of the gradient in the plurality of analysis point series. Any of the performance evaluation methods of 3. In the calculation of the first evaluation index, when the analysis point series is specified between a plurality of sections in the second feature quantity series for a specific section in the first feature quantity series, the second feature The performance evaluation method according to any one of claims 1 to 4, wherein the first evaluation index is calculated according to the number of the sections in the quantity series. In the calculation of the second evaluation index, for each of the plurality of code string groups,
Calculate the gradient of the analysis point series corresponding to each performance code string classified in the code string group, and calculate
Any of claims 1 to 5 for calculating the second evaluation index according to the representative value of the gradient in one or more analysis point series corresponding to one or more performance code strings classified in the code string group. How to evaluate the performance. In the calculation of the second evaluation index, for each of the plurality of code string groups,
Calculate the gradient of the analysis point series corresponding to each performance code string classified in the code string group, and calculate
Any of claims 1 to 6 for calculating the second evaluation index according to the degree of dispersion of the gradient in one or more analysis point series corresponding to one or more performance code strings classified into the code string group. How to evaluate the performance. In the calculation of the second evaluation index, claims 1 to claim 1 to calculate the second evaluation index according to the number of performance code strings classified into the code string group for each of the plurality of code string groups. Any of the performance evaluation methods of 7. A feature extraction unit that generates multiple feature series, which are time series of features related to the performance sound, for multiple performances by the user.
From the plurality of analysis points corresponding to different combinations of the plurality of features in the first feature series and the plurality of features in the second feature series among the plurality of feature series, the first feature series and the first feature series A first-series identification unit that specifies a plurality of analysis-point sequences, which are a series of analysis points having similar features with the second feature-quantity series,
The first index calculation unit that calculates the first evaluation index regarding the proficiency level of the performance by the user from the plurality of analysis point series, and
For each of the plurality of analysis point sequences specified for each feature quantity series, a performance code string in which performance codes corresponding to the feature quantities of the analysis points are arranged is generated for each analysis point constituting the analysis point series. 2nd series specific part to be
A code string classification that classifies the performance code string generated for each of the plurality of performances into any of a plurality of code string groups that are a set of one or more performance code strings having a similar or common arrangement of performance codes. Department and
For each of the plurality of code string groups, a second index calculation unit for calculating a second evaluation index regarding the proficiency level of performance by the user is provided from one or more performance code strings classified into the code string group. Performance evaluation device. A feature extraction unit that generates multiple feature series, which are time series of features related to the performance sound, for multiple performances by the user.
From the plurality of analysis points corresponding to different combinations of the plurality of features in the first feature series and the plurality of features in the second feature series among the plurality of feature series, the first feature series and the first feature series A first sequence identification unit that specifies a plurality of analysis point sequences, which are a series of analysis points having similar feature quantities with the second feature quantity series.
The first index calculation unit, which calculates the first evaluation index regarding the proficiency level of performance by the user from the plurality of analysis point series,
For each of the plurality of analysis point sequences specified for each feature quantity series, a performance code string in which performance codes corresponding to the feature quantities of the analysis points are arranged is generated for each analysis point constituting the analysis point series. 2nd series identification part,
A code string classification that classifies the performance code string generated for each of the plurality of performances into any of a plurality of code string groups that are a set of one or more performance code strings having a similar or common arrangement of performance codes. Department and
For each of the plurality of code string groups, a computer is used as a second index calculation unit for calculating a second evaluation index regarding the proficiency level of performance by the user from one or more performance code strings classified into the code string group. A program to make it work.