JP2010152381A - Device, method, and program for acoustic signal analysis - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device, a method, and a program for acoustic signal analysis that precisely generate a feature quantity representing the depth of a sound exerting a large influence on an atmosphere of a musical piece. <P>SOLUTION: A frequency analysis section 12 divides an acoustic signal into a plurality of frequency bands to generate matrix data consisting of component strengths of the respective bands as elements in a prescribed time cycle. A stable component detection section 13 detects elements larger than a prescribed value as effective elements from the matrix data and detects, as stable components, regions in each of which the prescribed number of or more effective elements are present within a time corresponding to a plurality of time cycles, in the matrix data. A feature quantity generation section 14 generates a feature quantity representing the depth of a sound in a prescribed section by using the sum total of intensities of or the number of stable components in the prescribed section. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

The present invention relates to an acoustic signal analysis device, an acoustic signal analysis method, and an acoustic signal analysis program that extract a feature of a song from an acoustic signal recorded with the song.

In recent years, with the progress of audio compression technology and the widespread use of mass storage devices, digitized music data is widely stored in computers and the like.

Under such circumstances, conventionally, as shown in Patent Documents 1 and 2, a device that generates music tone information from an acoustic signal and controls a lighting device, an air conditioner, a toy, etc. in synchronization with the music tone information has been proposed. ing. In addition, as disclosed in Patent Document 3, an apparatus for analyzing music signals of music to determine the atmosphere and genre of the music to search for music has been proposed.

JP 2004-163767 A JP2003-228387A JP 2004-185432 A

When expressing the tone of a song or instrument, the thickness of the sound is an important factor that determines the atmosphere of the song, as can be seen from the frequent use of expressions such as “thick sound” and “slim sound”. . However, in Patent Documents 1 to 3, the sound pressure from the acoustic signal,
Although information such as beat chords was extracted, features that directly reflected the thickness of the sound were not accurately extracted.

In Patent Document 1, in addition to detecting chords (code names) such as C major and C minor from an acoustic signal, the intensity ratio of a fundamental tone and a harmonic is calculated. Although it is considered possible to use the intensity ratio between the fundamental tone and the harmonic overtone as one feature value representing the thickness of the sound, the frequency spectrum of the acoustic signal of a general music composed of a plurality of musical instruments is very complex, A technique for separating the fundamental tone and the harmonic overtone from such an acoustic signal with sufficient accuracy cannot be said to be established at present.

In other words, the intensity ratio between the fundamental tone and the harmonic overtone described in Patent Document 1 is not necessarily suitable for practical use as a feature value representing the thickness of the sound.

Also, in the part where percussion instruments etc. are pronounced in the music, the intensity of a wide frequency band is increased, and the apparent harmonic intensity is very large. I don't feel much. Since the frequency spectrum fluctuates greatly before and after the timing at which percussion instruments produce sound, the influence of such percussion instruments should be reduced by searching the acoustic signal for a place where the frequency spectrum is stable for a certain period of time. However, Patent Documents 1 to 3 do not have means or a method for determining temporal stability of the frequency spectrum.

Therefore, the method disclosed in Patent Document 1 cannot detect the feature amount related to the thickness of the sound with sufficient accuracy.

In Patent Document 2, the sense of localization of a sound image, the harmonic feeling that is the pitch of a sound, and the sound pressure level are detected to determine the height of the music. However, although these feature quantities are not irrelevant to the thickness of the sound, they are not indices that directly represent the auditory sound thickness. Further, as described above, there is no means or method for determining the temporal stability of the frequency spectrum.

In Patent Document 3, a chord (a chord name) is detected as in Patent Document 1, but a feature amount related to the thickness of the sound is not detected.

Therefore, the present invention has an object to provide an acoustic signal analysis device, an acoustic signal analysis method, and an acoustic signal analysis program that can accurately generate a feature quantity that directly reflects the thickness of a sound that has a great influence on the atmosphere of a music piece. And In particular, it is an object of the present invention to provide an acoustic signal analysis device, an acoustic signal analysis method, and an acoustic signal analysis program that can accurately generate a feature quantity by reducing the influence of a percussion instrument or the like.

Also provided are an acoustic signal analyzer, an acoustic signal analysis method, and an acoustic signal analysis program capable of generating a feature quantity that directly reflects the thickness of a sound with a small amount of processing without separating a fundamental tone and a harmonic overtone from the acoustic signal. For the purpose.

Therefore, in order to solve the above problems, the present invention provides the following apparatus, method, and program.
(1) An acoustic signal analyzer that extracts the characteristics of the music from an audio signal related to the music,
Frequency analysis means for performing frequency analysis on the acoustic signal and generating frequency component data composed of each element corresponding to time, frequency, and component intensity;
An element in which the component intensity is a predetermined value or more is detected as an effective element from the frequency component data, and a region in which a predetermined number or more of the effective elements having the same frequency are present within a predetermined time in the frequency component data Stable component detection means for detecting as a stable component;
Feature amount generating means for generating a feature amount of the predetermined section based on the total sum of the strengths of the stable components in the predetermined section or the number of the stable components;
The stable component detecting means adds a value obtained by adding, in the time direction, component intensities of other elements corresponding to the same frequency as the element to be determined as to whether the element is the effective element or a frequency in the vicinity of the same frequency. To calculate the predetermined value,
An acoustic signal analyzer characterized by that.
(2) The stable component detection means adds, in the time direction, component intensities of other elements corresponding to the same frequency as the element to be determined as to whether the element is the effective element or a frequency in the vicinity of the same frequency. Using the value obtained, an average value of the component strength is calculated, and a value obtained by multiplying the average value by a predetermined coefficient is calculated as the predetermined value.
The acoustic signal analyzer according to (1) above, wherein
(3) The stable component detection means is another element that is temporally adjacent to the element that is the determination target of whether or not it is the effective element, and the same frequency as the element that is the determination target, or the same Using the value obtained by adding the component intensities of other elements corresponding to frequencies in the vicinity of the frequency in the time direction, the predetermined value is calculated.
The acoustic signal analyzer according to (1) or (2) above, wherein
(4) The frequency analysis means may be any one of a frequency that is equally spaced on the frequency axis, a frequency that corresponds to an average temperment scale, or a frequency that is divided more finely than a semitone of the average temperament scale. Generating each element corresponding to
The acoustic signal analysis device according to any one of (1) to (3) above, wherein
(5) The feature quantity generation means is the number of the stable components in the predetermined section, or a value obtained by dividing the number of the stable components by the number of types of frequencies in the elements constituting the frequency data, or the stable components The feature amount is a value obtained by dividing the number of the above by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section.
The acoustic signal analysis device according to any one of (1) to (4) above, wherein
(6) The feature value generation means is a sum of the strengths of the stable components in the predetermined section, or a value obtained by dividing the sum of the strengths of the stable components by the number of types of frequencies in the elements constituting the frequency data, Alternatively, the feature amount is a value obtained by dividing the sum of the strengths of the stable components by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section.
The acoustic signal analysis device according to any one of (1) to (4) above, wherein
(7) The feature quantity generation means generates the feature quantity by performing a process of smoothing in the time direction.
The acoustic signal analyzer according to any one of (1) to (6) above, wherein
(8) An acoustic signal analysis method executed by an acoustic signal analyzer that extracts features of the music from an acoustic signal related to the music,
A frequency analysis step of performing frequency analysis on the acoustic signal and generating frequency component data composed of elements corresponding to time, frequency, and component intensity;
An element in which the component intensity is a predetermined value or more is detected as an effective element from the frequency component data, and a region in which a predetermined number or more of the effective elements having the same frequency are present within a predetermined time in the frequency component data A stable component detection step for detecting as a stable component;
A feature amount generation step of generating a feature amount of the predetermined section based on a sum of the strengths of the stable components in a predetermined section or the number of the stable components;
In the stable component detection step, a value obtained by adding in the time direction the component intensity of another element corresponding to the same frequency as the element to be determined whether or not it is the effective element or a frequency in the vicinity of the same frequency. To calculate the predetermined value,
An acoustic signal analysis method characterized by the above.
(9) In the stable component detection step, the component intensities of other elements corresponding to the same frequency as the element to be determined whether or not it is the effective element or a frequency near the same frequency are added in the time direction. Using the value obtained, an average value of the component strength is calculated, and a value obtained by multiplying the average value by a predetermined coefficient is calculated as the predetermined value.
The acoustic signal analysis method according to (8) above, wherein
(10) The stable component detection step is another element that is temporally adjacent to the element that is the determination target of whether or not it is the effective element, and the same frequency as the element that is the determination target, or the same Using the value obtained by adding the component intensities of other elements corresponding to frequencies in the vicinity of the frequency in the time direction, the predetermined value is calculated.
The method for analyzing an acoustic signal as described in (8) or (9) above.
(11) The frequency analysis step may be any one of a frequency that is equally spaced on the frequency axis, a frequency that corresponds to an average temperment scale, or a frequency that is more finely divided than a semitone of the average temperament scale. Generating each element corresponding to
The acoustic signal analysis method according to any one of (8) to (10) above, wherein
(12) In the feature quantity generation step, the number of the stable components in the predetermined section, or a value obtained by dividing the number of the stable components by the number of types of frequencies in the elements constituting the frequency data, or the stable components The feature amount is a value obtained by dividing the number of the above by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section.
The acoustic signal analysis method according to any one of (8) to (11) above, wherein
(13) In the feature quantity generation step, the sum of the strengths of the stable components in the predetermined section, or a value obtained by dividing the sum of the strengths of the stable components by the number of types of frequencies in the elements constituting the frequency data, Alternatively, the feature amount is a value obtained by dividing the sum of the strengths of the stable components by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section.
The acoustic signal analysis method according to any one of (8) to (11) above, wherein
(14) The feature quantity generation step generates the feature quantity by performing a process of smoothing in a time direction.
The acoustic signal analysis method according to any one of (8) to (13) above, wherein
(15) An acoustic signal analysis program for extracting features of the music from an audio signal related to the music,
A frequency analysis step of performing frequency analysis on the acoustic signal and generating frequency component data composed of elements corresponding to time, frequency, and component intensity;
An element in which the component intensity is a predetermined value or more is detected as an effective element from the frequency component data, and a region in which a predetermined number or more of the effective elements having the same frequency are present within a predetermined time in the frequency component data A stable component detection step for detecting as a stable component;
Causing the computer to execute a feature amount generation step of generating a feature amount of the predetermined section based on the total sum of the strengths of the stable components in the predetermined section or the number of the stable components;
In the stable component detection step, a value obtained by adding in the time direction the component intensity of another element corresponding to the same frequency as the element to be determined whether or not it is the effective element or a frequency in the vicinity of the same frequency. To calculate the predetermined value,
An acoustic signal analysis program characterized by that.
(16) In the stable component detection step, component intensities of other elements corresponding to the same frequency as the element to be determined whether or not it is the effective element or a frequency in the vicinity of the same frequency are added in the time direction. Using the value obtained, an average value of the component strength is calculated, and a value obtained by multiplying the average value by a predetermined coefficient is calculated as the predetermined value.
The acoustic signal analysis program according to (15) above, wherein
(17) The stable component detection step is another element that is temporally adjacent to an element that is a determination target of whether or not it is the effective element, and the same frequency as the element that is the determination target, or the same Using the value obtained by adding the component intensities of other elements corresponding to frequencies in the vicinity of the frequency in the time direction, the predetermined value is calculated.
The acoustic signal analysis program according to (15) or (16) above, wherein
(18) The frequency analysis step may be any one of a frequency that is equally spaced on the frequency axis, a frequency that corresponds to an average temperment scale, or a frequency that is more finely divided than a semitone of the average temperament scale. Generating each element corresponding to
The acoustic signal analysis program according to any one of (15) to (17) above, wherein
(19) In the feature quantity generation step, the number of stable components in the predetermined section, or a value obtained by dividing the number of stable components by the number of types of frequencies in elements constituting the frequency data, or the stable components The feature amount is a value obtained by dividing the number of the above by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section.
The acoustic signal analysis program according to any one of (15) to (18) above, wherein
(20) In the feature quantity generation step, the sum of the strengths of the stable components in the predetermined section, or a value obtained by dividing the sum of the strengths of the stable components by the number of types of frequencies in the elements constituting the frequency data, Alternatively, the feature amount is a value obtained by dividing the sum of the strengths of the stable components by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section.
The acoustic signal analysis program according to any one of (15) to (18) above, wherein
(21) The feature amount generation step generates the feature amount by performing a process of smoothing in a time direction.
The acoustic signal analysis program according to any one of (15) to (20) above, wherein

According to the acoustic signal analysis device, the acoustic signal analysis method, and the acoustic signal analysis program of the present invention,
Distinguishes between places where a pitched instrument is sounded and a constant frequency is stably maintained, and places where percussion instruments are sounded and a constant frequency is not stable, The feature quantity that represents the thickness of the sound is generated by calculating the total number of frequency components or the intensity of the component, which directly reflects the sense of thickness of the sound, which is a large factor that determines the atmosphere of the music. It is possible to accurately generate a feature amount suitable for a sense of thickness. In addition, since processing is performed in a manner that does not discriminate and separate the fundamental tone and harmonics of the musical sound, feature quantities can be generated with a relatively small amount of processing, and the cost of the acoustic signal analyzer can be reduced. It can also be executed by a computer having a small arithmetic processing capability.

It is a block diagram which shows the structure of the acoustic signal analyzer of Example 1 and Example 2 of this invention. It is a flowchart which shows the processing flow of the frequency analysis part of FIG. It is a figure which shows the flame | frame preparation operation | movement in the frequency analysis part of FIG. It is a figure which shows the characteristic of the filter group used by the frequency component calculation operation | movement in the frequency analysis part of FIG. It is a schematic diagram which shows the characteristic of the matrix data produced | generated by the frequency analysis part of FIG. 2 is a flowchart illustrating a processing flow of a stable component detection unit in FIG. 1 according to the first embodiment. It is a figure which shows the data storage format of the stable component detection part of FIG. 3 is a flowchart illustrating a processing flow of a feature amount generation unit in FIG. 1 according to the first embodiment. 6 is a flowchart illustrating a processing flow of a stable component detection unit in FIG. 1 according to a second embodiment. It is a figure which shows the data storage format of the stable component detection part of FIG. 6 is a flowchart illustrating a processing flow of a feature amount generation unit in FIG. 1 according to a second embodiment.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
Example 1
Embodiment 1 of an acoustic signal analysis device, an acoustic signal analysis method, and an acoustic signal analysis program of the present invention
Will be described with reference to FIGS. 1 is a block diagram showing a configuration of an acoustic signal analysis apparatus according to a first embodiment of the present invention, FIG. 2 is a flowchart showing a processing flow of a frequency analysis unit in FIG. 1, and FIG.
Is a diagram showing the frame creation operation in the frequency analysis unit of FIG. 1, FIG. 4 is a diagram showing the characteristics of the filter group used in the frequency component calculation operation in the frequency analysis unit of FIG. 1, FIG. 5 is a frequency analysis unit of FIG. 6 is a schematic diagram showing the characteristics of the generated matrix data, FIG. 6 is a flowchart showing the processing flow of the stable component detector of FIG. 1 in the first embodiment, and FIG. 7 is data storage of the stable component detector of FIG. FIG. 8 is a flowchart showing a processing flow of the feature quantity generation unit of FIG. 1 in the first embodiment.

As shown in FIG. 1, the acoustic signal analyzer 1 includes an acoustic signal input unit 11 and an A / D converter 11.
b, a frequency analysis unit 12, a stable component detection unit 13, a stable component memory 13b, a feature amount generation unit 14, and arithmetic processing circuits 11a to 14a.

An acoustic signal in which music is recorded is input to the acoustic signal input unit 11. A / D converter 11b
Is the acoustic signal A when the acoustic signal input to the acoustic signal input unit 11 is an analog signal.
/ D conversion.

The frequency analysis unit 12 divides the acoustic signal into a plurality of frequency bands, and generates matrix data having the component intensity of each band in a predetermined time period as an element.

The stable component detection unit 13 detects an element having a component intensity of a predetermined value or more from the matrix data generated by the frequency analysis unit 12 as an effective element, and within a time period corresponding to a plurality of times of the time period in the matrix data. A region where a predetermined number or more of the effective elements are present is detected as a stable component. The stable component memory 13b stores information on stable components detected by the stable component detector 13.

The feature quantity generation unit 14 refers to the stable component memory 13b and generates a feature quantity that represents the thickness of the sound in the section using the sum of the strengths of the stable components in the predetermined section or the number of stable components.

  The arithmetic processing circuits 11a to 14a calculate and control each part of the apparatus.

  Next, the operation of the acoustic signal analyzing apparatus 1 according to the first embodiment and the acoustic signal analyzing method will be described.

First, in the acoustic signal input unit 11, when the input acoustic signal is an analog signal, the arithmetic processing circuit 11a sends a predetermined sampling frequency Fs to the A / D converter 11b.
Control to digitize. When the input acoustic signal is a digital signal, rate conversion is performed so that the sampling frequency becomes a predetermined value Fs. The data digitized by the acoustic signal input unit 11 is hereinafter represented as acoustic data x [m] (m = 0 to L−1, L is the total number of acoustic data).

Next, in the frequency analysis unit 12, the arithmetic processing circuit 12 a performs frequency analysis on the acoustic data digitized at a predetermined sampling rate by the acoustic signal input unit 11, and each band for each predetermined time period. Is calculated, and matrix data having the component intensity as a matrix element is created.

In this embodiment, a well-known STFT (Short-time Fourier Transform) is used as a frequency analysis method.
rm) is used, but the present invention is not limited to this, and wavelet transform, filter bank, or the like may be used.

Here, the processing flow of the frequency analysis unit 12 will be described based on the flowchart shown in FIG. In this embodiment, the acoustic data is divided into fixed-length frames, and processing is performed in units of frames. In the following, it is assumed that the frame length is N and the frame shift length is S. The frame shift length S corresponds to a time period. The total number M of frames is obtained according to (Formula 1). Where floor
The function is a function that returns an integer with the decimal part truncated.

First, in step S110, the arithmetic processing circuit 12a controls the control variable i indicating the frame number.
Is set to 0.

Next, in step S120, the arithmetic processing circuit 12a creates the i-th frame.
That is, as shown in FIG. 3, N × from the position offset by i × S from the beginning of the acoustic data.
This data is cut out and multiplied by a window function w as shown in (Formula 2) to create i-th frame data y [i] [n] (n = 0 to N−1).

As the window function, for example, a Hamming window shown in (Formula 3) may be used. In addition, a rectangular window, Hanning window, Blackman window, or the like may be used.

Next, in step S130, the arithmetic processing circuit 12a calculates the discrete Fourier transform (DFT) of the i-th frame according to (Equation 4).

Next, in step S140, the arithmetic processing circuit 12a has the real part Re {a [i] [k]} of the complex sequence a [i] [k] (k = 0 to N−1) obtained in step S130. And imaginary part Im
{A [i] [k]} is used to calculate the spectrum sequence b [i] [k] (k = 0 to N / 2-1) of the i-th frame according to (Equation 5) or (Equation 6). calculate.

Next, in step S150, the arithmetic processing circuit 12a determines the frame i,
The frequency component c [i] [q] (q = 0 to Q−1, Q is the number of bands) of the band q is calculated. Here, there are the following three calculation methods in step S150.

The first method for calculating the frequency component is the spectrum sequence b [i] [k according to (Equation 7).
], A part or all of them are made to correspond to c [i] [q]. Here, λ is a predetermined integer of 0 or more, and is a parameter that determines the lowest frequency of the band. Further, the number of bands Q is set to a predetermined value which is (N / 2−λ) or less. The first method has the least amount of calculation and is simple.

The second method for calculating the frequency component is a method for obtaining the frequency component corresponding to the average tempered scale according to (Equation 8).

Here, Fs is a sampling frequency in the acoustic signal input unit 11, and Fr is a frequency serving as a reference for the average temperament scale. For example, if the “center sound” is 440 Hz and a sound 4 octaves lower than this is used as the reference for the average temperament scale, Fr = 27.5 Hz.

V is a constant that determines how many bands a 1-octave scale is divided into. For example, when one octave is divided into 12 bands, V = 12. Further, the value of V may be made larger than this and divided into bands that are finer than the semitones of the average temperament scale. The function R is a function that outputs an integer closest to the input value. K1 and K2 are constants for determining the lowest frequency (lowest musical scale) and the highest frequency (highest musical scale) of the band, and μ is a constant for setting the minimum value of the argument q representing the band to zero. The frequency components c [i] [q] are obtained by adding the spectrum series b [i] [k] by the number k corresponding to the same value q.

Compared to the first method, the second method can calculate a frequency component more reflecting the musical characteristics of the music.

The third method for calculating the frequency component is a method for obtaining the frequency component corresponding to the average tempered scale according to (Equation 9).

Here, z [q] [k] (q = 0 to Q-1, k = 0 to N / 2-1) is a filter group having band characteristics as shown in FIG. The frequency corresponds to the frequency of the average scale.

In the example shown in FIG. 4, the C1 pitch of the average temperament scale is made to correspond to the band 0, and thereafter, one semitone is made to correspond to one band, and finally the B6 pitch is made to correspond to the band Q-1. FIG.
z [0] [k] shown in a) is a filter that passes a frequency corresponding to C1 (do),
Z [1] [k] shown in FIG. 4 (b) indicates a filter that passes a frequency corresponding to C # 1 (de #) that is a semitone higher than that.

The spectrum series b [i] [k] exists at equal intervals on the frequency axis, whereas the average temperament scale has a higher frequency range so that the frequency interval between adjacent semitones increases, so the filter group z [q]
Corresponding to the center frequency of [k], the interval between the center frequencies adjacent to each other is increased as the treble part is increased. For example, z [0] [k] shown in FIG. 4 (a) and z [1] [k] shown in FIG. 4 (b).
Z [Q-2] [k] shown in FIG. 4C and z [Q shown in FIG.
−1] [k] has a larger difference in center frequency.

Similarly, the bandwidth of each filter becomes wider as the treble part increases. For example, FIG.
The bandwidth of Z [Q-1] [k] shown in FIG. 4 is wider than the bandwidth of z [0] [k] shown in FIG.

In the example shown in FIG. 4, the band matches the average temperament semitone, but a band that further divides the average temperament semitone may be obtained.

The third method can determine the frequency corresponding to the average temperament scale with higher accuracy than the second method.

Returning to the description of FIG. 2, in step S160, the arithmetic processing circuit 12a increments the value of the control variable i indicating the frame number by one.

Next, in step S170, the arithmetic processing circuit 12a controls the control variable i indicating the frame number.
It is checked whether the value of is smaller than the total number M of frames. Smaller than total number of frames M (YE
When S), the process returns to step S120 and is repeated. The total number of frames is more than M (NO)
At that time, since all the frames have been processed, the processing ends.

When the above processing is completed, the frequency analysis unit 12 forms and stores frequency components c [i] [q] (frames i = 0 to M−1, bands q = 0 to Q−1) in a matrix format. Therefore, the stable component detector 13 can be used.

Here, the feature-value which the acoustic signal analyzer 1 of this invention produces | generates is demonstrated. Various factors are related to the depth of sound that humans feel when listening to music and music, but the following two factors have a great influence.
(1) Number of notes that are sounded simultaneously (number of notes, number of fundamentals)
(2) Overtone component contained in each sound (1) relates to so-called harmony and is a matter that can be expressed by a score. Usually, it is felt that “the sound is thicker” as the number of sounds (number of notes, the number of fundamental notes) that are pronounced simultaneously increases.

(2) relates to the tone of the musical instrument and cannot be expressed in the score. Normally, it is felt that “the sound is thicker” as there are more harmonic components contained in each sound.

Since the above two factors are different, a method of calculating the overall sound thickness after separating the two factors once can be considered. However, it is difficult to recognize a fundamental tone from an acoustic signal of a general musical composition in which a plurality of unspecified types of musical instruments are mixed, and it is not practical to forcibly separate the above two factors.

Therefore, in the present invention, the feature quantity representing the sense of thickness of the sound is calculated without separating the above two factors.
For this reason, a feature amount can be obtained with a relatively small amount of calculation.

FIG. 5 is a schematic diagram showing characteristics of matrix data generated by the frequency analysis unit 12 of FIG. In FIG. 5, the frame is shown on the horizontal axis and the band is shown on the vertical axis, and the black portion (including the hatched portion in the C portion) shows an element having a strong component strength. Since a general musical piece includes a pitched musical instrument that generates a sound by connecting a certain pitch, there are many horizontal segments.

In this horizontal line segment, there is a mix of what corresponds to the fundamental tone and what corresponds to the overtone,
It is difficult to determine which is a fundamental tone and which is a harmonic.

On the other hand, for example, when comparing the A portion and the B portion, the A portion having a larger number of horizontal line segments is more likely to be a portion where the sound is perceived thicker.

Also, at locations where percussion instruments that do not have a clear pitch in the music are being pronounced, or at the locations where some instruments are pronounced, there are elements with a strong component intensity in a wide band, as shown in part C. . In such a portion, although the apparent harmonic component becomes very strong, the thickness of the sound is not so much perceived.

Therefore, the stable component detection unit 13 and the feature value generation unit 14 calculate the number of frequency components or the sum of the component strengths that are temporally stable of the musical instrument so as not to be affected by percussion instruments and the like. Generate quantity.

Next, the processing flow of the stable component detection unit 13 will be described based on the flowchart shown in FIG. In the stable component detector 13, the frequency component c stored in the frequency analyzer 12
[I] [q] (i = 0 to M-1, q = 0 to Q-1) is read and processed.

First, in step S210, the arithmetic processing circuit 13a sets a control variable p representing a frame number for starting a search to 0.

Next, in step S220, the arithmetic processing circuit 13a sets the control variable q representing the band to the minimum band Q1 (Q1 is a constant greater than or equal to 0 and less than Q) that is the target of the stable component.

Next, in step S230, the arithmetic processing circuit 13a sets a variable r for counting a frequency component meeting a condition described later as an effective component to 0.

Next, in step S240, the arithmetic processing circuit 13a controls the control variable i representing the frame number.
Is set to the value of p.

Next, in step S250, the arithmetic processing circuit 13a checks whether the frequency component c [i] [q] is a valid component. When it is determined that the component is an effective component (YES), the process proceeds to step S260, and when it is determined that the component is not an effective component (NO), the process proceeds to step S270. As a specific method for determining an effective component in step S250, 5 described below.
Any one of the two methods or an appropriate combination may be used.

A first method for determining an effective component is a method for determining an effective component using (Equation 10) when c [i] [q] is equal to or greater than a threshold value α [q].

  Here, the threshold value α [q] may be determined by a method described later.

A second method for determining an effective component is a method using (Formula 11). In addition to the condition of (Formula 10) described above, this is a frequency component of the same time as the band q (referred to as the center band) and belongs to a frequency band (referred to as a nearby band) in the vicinity of the band q. The frequency component of the band q is greater than the value obtained by multiplying the sum of these frequency components by a fixed ratio γ.
The condition that the frequency component c [i] [q] is an effective component is added. That is, when a certain frequency component is equal to or higher than a predetermined value and the frequency component is larger than a value obtained by multiplying the sum of the frequency components in the vicinity by a predetermined ratio, the frequency component is determined as an effective component. Become. In (Formula 11), the sum is calculated using only the frequency component of frame i, but the present invention is not limited to this. For example, the sum may be calculated including a frame in the vicinity of frame i. good.

Here, G1, G2, and γ are constants. The threshold value α [q] may be determined by a method described later. In general, when a musical instrument is sounded, the band component of the pitch is stronger than the adjacent band component, and when the percussion instrument is sounded, the difference between the adjacent band components is small. Is used.

That is, the condition of the two items of (Equation 11) is true in the portions A and B shown in FIG. 5, but is false in the portion C, and does not include a portion where a percussion instrument or the like is pronounced as a stable component. have. The constants G1 and G2 may be determined according to the frequency difference between the bands. In general music, two pitched musical instruments are not often pronounced with a pitch that is different from a semitone, whereas when percussion instruments are pronounced, frequency components that are different from a semitone are often strong at the same time. Therefore,
It can be determined that the musical instrument is sounded when the intensity of the central band is somewhat higher than the intensity of the adjacent band that differs from the central band by a semitone to two semitone frequency. For this purpose, G1 is set so that the frequency difference between the central band and the neighboring band corresponds to one or two semitones of the average scale, and the one-side bandwidth of the neighboring band corresponds to one or two semitones of the average scale. And G2 may be set. For example, when each band corresponds to each semitone of the average temperament scale, G1 = 1 to 1
2, G2 = 1-3 is appropriate. Of course, it is not limited to this value.

A third method for determining an effective component is a method using (Formula 12). The concept is the same as the second method for determining an effective component, but it is not the sum of the frequency components specified from the frequency components belonging to the band near band q, but the frequency components belonging to the band near band q. The maximum frequency component is used. That is, when a certain frequency component is equal to or greater than a predetermined value and the frequency component is larger than a value obtained by multiplying the maximum value of the neighboring frequency components by a predetermined ratio, the frequency component is determined as an effective component. become. (Equation 12)
In this case, the maximum value is calculated using only the frequency component of the frame i, but the present invention is not limited to this. For example, the maximum value may be calculated including a frame near the frame i.

A fourth method for determining an effective component is a method using (Formula 13). In addition to the condition of (Equation 10) described above, this specifies a frequency component belonging to the frequency band having the same time as the band q and having a harmonic relationship with the frequency band of the band q (harmonic band). If the frequency component belonging to the harmonic band is larger than the value obtained by identifying several frequency components from the frequency components near the frequency component belonging to the harmonic band and multiplying the sum of these frequency components by a fixed ratio. And c [i] [q] as an active ingredient. That is,
When a certain frequency component is greater than or equal to a predetermined value and the value of the harmonic component in harmonic relationship with the frequency component is greater than a value obtained by multiplying the sum of frequency components in the vicinity of the harmonic component by a predetermined ratio,
The frequency component is determined as an effective component.

Here, the function h (d, q) is a function that returns a band number corresponding to a frequency (d overtone) of d times the band q. This is a value obtained by subtracting the sum of the frequency components of the bands in the vicinity of the dq band from the frequency components of the band corresponding to the d harmonic overtone of the band q (referred to as the dq band) in addition to the condition of (Formula 10) described above. A condition that an active component is added when a value obtained by multiplying a value multiplied by the ratio η [d] as d = 2 to D (D is a constant of 2 or more) is larger than 0 is added.

G3 and G4 are constants determined by the frequency difference between the bands as in G1 and G2. The threshold value α [q] may be determined by a method described later.

In general, when a musical instrument is sounded, this method has a harmonic structure, and a component of a harmonic band having a frequency that is an integral multiple of the fundamental tone is in the vicinity of the harmonic band in terms of frequency (a harmonic nearby band and When the percussion instrument is pronounced,
Since there is no clear overtone structure, the property that such a condition is difficult to be satisfied with respect to the strength of the components of the overtone band and the overtone vicinity band is utilized. That is, the condition of the two items in (Equation 13) is true in the portions A and B shown in FIG. 5, but is false in the portion C, and does not include a portion where a percussion instrument or the like is pronounced as a stable component. have. In (Equation 13), the sum is calculated using only the frequency component of frame i. However, the present invention is not limited to this. For example, the sum may be calculated including frames in the vicinity of frame i. good.

A fifth method for determining an effective component is a method using (Formula 14). This is similar to the fourth method for determining the active component, but is not the sum of the frequency components of the nearby harmonics band,
The maximum value of the frequency component of the overtone vicinity band is used. That is, when a certain frequency component is greater than or equal to a predetermined value and the value of the harmonic component in harmonic relationship with that frequency component is greater than the value obtained by multiplying the maximum value of the frequency component in the vicinity of the harmonic component by a predetermined ratio In addition, the frequency component is determined as an effective component.

In (Equation 14), the maximum value is calculated using only the frequency component of frame i. However, the present invention is not limited to this. For example, the maximum value is calculated including a frame in the vicinity of frame i. May be.

Furthermore, an effective component may be determined by appropriately combining the five methods described above. For example, the second method and the fourth method are combined to satisfy (Equation 11) and (Equation 13).
) May be determined as an effective component only when the above condition is satisfied. In other words, in this case, a certain harmonic component is equal to or higher than a predetermined value, and the frequency component is larger than a value obtained by multiplying the sum of the neighboring frequency components by a predetermined ratio, and has a harmonic overtone relationship with the frequency component. When the value of the component is larger than a value obtained by multiplying the sum of frequency components in the vicinity of the harmonic component by a predetermined ratio, the frequency component is determined as an effective component. In this case, compared with the case where the second method or the fourth method is used alone, the influence of the percussion instrument or the like as described above can be further reduced, and the feature amount representing the audible sound thickness is further increased. It is possible to calculate with high accuracy.

As another example, the third method and the fifth method may be combined to determine that the component is effective only when (Expression 12) is satisfied and (Expression 14) is satisfied. That is, in this case, a certain frequency component is equal to or greater than a predetermined value, and the frequency component is larger than a value obtained by multiplying the maximum value of the frequency components in the vicinity by a predetermined ratio, and has a harmonic overtone relationship with the frequency component. When the value of the harmonic component is larger than a value obtained by multiplying the maximum value of the frequency component in the vicinity of the harmonic component by a predetermined ratio, the frequency component is determined as an effective component. In this case, compared with the case where the third method or the fifth method is used alone, it is possible to further reduce the influence of the percussion instrument or the like as described above, and to further increase the feature amount representing the audible sound thickness. It is possible to calculate with high accuracy.

  Further, an effective component may be determined by combining other methods.

Next, a method for determining the threshold value α [q] in the above (Formula 10) to (Formula 14) will be described.

A first method for determining the threshold value α [q] is a method of setting a preset constant.
This method is simple because the amount of calculation in the stable component detector 13 is the smallest. In addition, when using the 2nd-5th method in the method of determining an active ingredient mentioned above, threshold value (alpha) [q]
Is a relatively small value ("0" in extreme cases), and the threshold value α [
It is also possible to set the influence of q] to be small.

The second method for determining the threshold value α [q] is as shown in (Formula 15) for all frames (M
This is a method using an average value of frequency components for each band.

Here, β is a preset constant. The second method has a feature that it is hardly affected by variations in the magnitude of the acoustic signal for each music piece.

A third method for determining the threshold value α [q] is a method using an average value of frequency components for each band in a frame near the i-th frame, as shown in (Formula 16).

Here, φ (i) represents a set of frames belonging to the neighborhood of the i-th frame, H is the number of neighboring frames (H <M), and β is a preset constant. The third method is suitable for the case where the intensity of the acoustic signal changes greatly in one piece of music and the change in signal intensity is not desired to be reflected in the processing result.

The fourth method for determining the threshold value α [q] is a method using an average value of frequency components over a plurality of bands, as shown in (Expression 17). Here, δ is a parameter for determining the number of bands used for calculating the average value.

Returning to the description of FIG. 6, in step S260, the arithmetic processing circuit 13a increases the value of the variable r for counting the active component by one.

  Next, in step S270, the arithmetic processing circuit 13a increases the value of the control variable i by one.

Next, in step S280, the arithmetic processing circuit 13a determines whether or not the value of the control variable i is less than (p + U), and when it is less than (p + U) (YES), the process returns to step S250 to perform the process. repeat. Here, U is a constant. The value of the control variable i is (p + U) or more (N
O), the process proceeds to step S290.

In step S290, the arithmetic processing circuit 13a sets the effective component counting variable r.
Is greater than or equal to a constant V (where V ≦ U), the process proceeds to step S300 when it is greater than or equal to V (YES), and proceeds to step S310 when it is less than V (NO).

Here, when V = U, it is determined as a stable component only when U effective components exist continuously. However, in reality, even when a sound of a certain pitch is sounded for a certain period of time, there is a minute frequency fluctuation (vibrato), so that effective components are not always continuous and exist intermittently. In some cases. For this reason, a better result may be obtained when V is set to about 80 to 90% of U.

Next, in step S300, the arithmetic processing circuit 13a stores information on the frequency component that satisfies the condition in step S290 in the stable component memory 13b of the stable component detecting unit 13.
Specifically, the set of (p, q) is stored in the stable component memory 13b in the format shown in FIG. The stable component memory 13b can be referred to from the feature value generation unit 14.

  Next, in step S310, the arithmetic processing circuit 13a increases the value of the control variable q by one.

Next, in step S320, the arithmetic processing circuit 13a determines whether or not the value of the control variable q is equal to or less than Q2, and when it is equal to or less than Q2 (YES), the process returns to step S230 and repeats the process. When larger than Q2 (NO), the process proceeds to step S330. Here, Q2 is a constant representing the maximum band that is the target of the stable component.

In step S330, the arithmetic processing circuit 13a increases the value of the control variable p by P. Here, P is normally 1, but P may be a value of 2 or more in order to reduce the processing amount. However, when P is set to 2 or more, the groups (p, q) to (p + P-1, q) are stored together when the stable components are stored in step S300.

In step S340, the arithmetic processing circuit 13a determines whether or not the control variable p is less than (MU). When it is less than (MU) (YES), the process returns to step S220 to repeat the process, and when it is equal to or greater than (MU) (NO), the process is terminated.

After the processing of the stable component detection unit 13 is performed in this manner, the stable component information is stored in the stable component memory 13b.

Next, the processing flow of the feature quantity generation unit 14 will be described based on the flowchart shown in FIG. The feature amount generation unit 14 generates a feature amount for each section having a predetermined length. In the present embodiment, T times the frame shift length S is set as the section length (T is an integer of 1 or more).

First, in step S510, the arithmetic processing circuit 14a sets a control variable t representing the head of a section for generating a feature value to 0.

Next, in step S520, the arithmetic processing circuit 14a refers to the stable component memory 13b of the stable component detector 13, and counts the number E of stable components in the section t. Specifically, the number of stable components that satisfy t ≦ p <t + T in the p field of the stable component memory 13b may be counted.

Next, in step S530, the arithmetic processing circuit 14a determines the feature amount ou for the section t.
As t [t], the number E of stable components, or E / Q obtained by dividing E by the total number of bands Q, or E / (QT) obtained by dividing E by the product of the total number of bands Q and the section length T. Is output.

Next, in step S540, the arithmetic processing circuit 14a increases the control variable t by T.

Next, in step S550, the arithmetic processing circuit 14a determines that the control variable t is floor (
It is determined whether it is less than (M / T). Here, the floor function is a function that returns an integer with the decimal part truncated. If it is less than floor (M / T) (YES), step S5
Returning to 20, the process is performed, and when it is equal to or greater than floor (M / T) (NO), the feature amount generation unit 1
The process of 4 is finished.

Note that the feature amount time-series data out [t] generated by the feature amount generation unit 14 may be smoothed in the time direction to obtain a smoother output.

As described above, according to the acoustic signal analyzing apparatus and the acoustic signal analyzing method of the first embodiment, a place where a musical instrument is sounded and a constant frequency is stably maintained, and a percussion instrument or the like is sounded and constant. The location of the frequency of the musical instrument is discriminated and the number of frequency components of the musical instrument that are stable over time is calculated to generate a feature value that represents the thickness of the sound. Therefore, it is possible to accurately generate a feature value that directly reflects the thickness of sound. In addition, since the processing is performed in a manner that does not discriminate and separate the fundamental tone and the harmonic overtone of the musical tone, it is possible to generate a feature value with a simple calculation.
(Example 2)
Embodiment 2 of the acoustic signal analysis device, acoustic signal analysis method, and acoustic signal analysis program of the present invention
Will be described with reference to FIGS. 1 and 9 to 11. FIG. 1 is a block diagram showing a configuration of an acoustic signal analyzer according to a second embodiment of the present invention, FIG. 9 is a flowchart showing a processing flow of the stable component detection unit of FIG. 1 in the second embodiment, and FIG. FIG. 11 is a flowchart showing a processing flow of the feature quantity generation unit of FIG. 1 in the second embodiment.

As shown in FIG. 1, the configuration of the acoustic signal analyzer 1 in the second embodiment of the present invention is the same as that in the first embodiment.
It is the same. The acoustic signal input unit 11 and the frequency analysis unit 12 perform the same operations as those described in the first embodiment.

Next, the processing flow of the stable component detection unit 13 will be described based on the flowchart shown in FIG. In the stable component detector 13, the frequency component c stored in the frequency analyzer 12
[I] [q] (i = 0 to M−1, q = 0 to Q−1) is read and processed.

First, in step S710, the arithmetic processing circuit 13a sets a control variable p representing a frame number for starting a search to 0.

Next, in step S720, the arithmetic processing circuit 13a sets the control variable q representing the band to the minimum band Q1 (Q1 is a constant greater than or equal to 0 and less than Q) that is the target of the stable component.

Next, in step S730, the arithmetic processing circuit 13a sets a variable r for counting a frequency component meeting a condition described later as an effective component to 0.

Next, in step S740, the arithmetic processing circuit 13a sets a variable sa for calculating the sum of the strengths of the active components to 0.

Next, in step S750, the arithmetic processing circuit 13a controls the control variable i representing the frame number.
Is set to the value of p.

Next, in step S760, the arithmetic processing circuit 13a checks whether or not the frequency component c [i] [q] is an effective component. When it is determined that the component is an effective component (YES), the process proceeds to step S770, and when it is determined that the component is not an effective component (NO), the process proceeds to step S790.
The specific method of step S760 is the same as the method described in the first embodiment.

Next, in step S770, the arithmetic processing circuit 13a increases the value of the variable r for counting active components by one.

Next, in step S780, the arithmetic processing circuit 13a adds the frequency component c [i] [q] to the variable sa for calculating the sum of the intensities of the effective components.

  Next, in step S790, the arithmetic processing circuit 13a increases the value of the control variable i by one.

next. In step S800, the arithmetic processing circuit 13a determines whether or not the value of the control variable i is less than (p + U). If it is less than (p + U) (YES), the process returns to step S760 and repeats the process. Here, U is a constant. The value of the control variable i is (p + U) or more (
If NO, the process proceeds to step S810.

In step S810, the arithmetic processing circuit 13a sets the variable r for counting the effective component.
Is greater than or equal to a constant V (however V ≦ U), the process proceeds to step S820 when it is greater than or equal to V (YES), and proceeds to step S830 when it is less than V (NO).

Here, when V = U, it is determined as a stable component only when U effective components exist continuously. However, in reality, even when a sound of a certain pitch is sounded for a certain period of time, there is a minute frequency fluctuation (vibrato), so that effective components are not always continuous and exist intermittently. In some cases. For this reason, a better result may be obtained when V is set to about 80 to 90% of U.

Next, in step S820, the arithmetic processing circuit 13a (p, q,
The set of sa) is stored in the stable component memory 13b. The stable component memory 13b is a feature quantity generation unit 1
4 can be referred to.

  Next, in step S830, the arithmetic processing circuit 13a increases the value of the control variable q by one.

Next, in step S840, the arithmetic processing circuit 13a determines whether or not the value of the control variable q is equal to or less than Q2, and when it is equal to or less than Q2 (YES), the process returns to step S730 and repeats the process. When larger than Q2 (NO), the process proceeds to step S850. Here, Q2 is a constant representing the maximum band that is the target of the stable component.

In step S850, the arithmetic processing circuit 13a increases the value of the control variable p by P. Here, P is normally 1, but P may be a value of 2 or more in order to reduce the processing amount. However, when P is set to 2 or more, not only (p, q, sa) but also a set of (p, q, sa) to (p + P-1, q, sa) at the time of storing stable components in step S820. Are stored together.

In step S860, the arithmetic processing circuit 13a determines whether or not the control variable p is less than (MU). When it is less than (MU) (YES), the process returns to step S720 to repeat the process, and when it is equal to or greater than (MU) (NO), the process is terminated.

After the processing of the stable component detection unit 13 is performed in this manner, the stable component information is stored in the stable component memory 13b.

Next, the processing flow of the feature quantity generation unit 14 will be described based on the flowchart shown in FIG. The feature amount generation unit 14 generates a feature amount for each section having a predetermined length. In the present embodiment, T times the frame shift length S is set as the section length (T is an integer of 1 or more).

First, in step S910, the arithmetic processing circuit 14a sets a control variable t representing the head of a section for generating a feature value to 0.

Next, in step S920, the arithmetic processing circuit 14a refers to the stable component memory 13b of the stable component detector 13, and calculates the sum sum of the strengths of the stable components in the section t. Specifically, the stable field set θ [t] where the p field of the stable component memory 13b satisfies t ≦ p <t + T is obtained, and the sum of sa belonging to the set θ [t] as shown in (Formula 18). In search of
um.

Next, in step S930, the arithmetic processing circuit 14a determines the feature amount ou for the section t.
Sum, or sum / Q, or sum / (QT) is output as t [t].

Next, in step S940, the arithmetic processing circuit 14a increases the control variable t by T.

Next, in step S950, the arithmetic processing circuit 14a determines that the control variable t is floor (
It is determined whether it is less than (M / T). Here, the floor function is a function that returns an integer with the decimal part truncated. If it is less than floor (M / T) (YES), the process returns to step S920 to perform the process, and if it is greater than floor (M / T) (NO), the process of the feature quantity generation unit 14 is terminated.

Note that the feature amount time-series data out [t] generated by the feature amount generation unit 14 may be smoothed in the time direction to obtain a smoother output.

As described above, the acoustic signal analysis device and the acoustic signal analysis method according to the second embodiment have the stable component detection unit 13.
The feature amount generation unit 14 calculates the sum of the strengths of the stable components in a predetermined section of the time-stable frequency component of the musical instrument and generates a feature amount representing the thickness of the sound. The same effect can be obtained.

In the acoustic signal analysis apparatus described in the first and second embodiments, each processing unit is provided with an arithmetic processing circuit. However, a configuration in which one arithmetic processing circuit controls each unit is also possible. .

In addition, the acoustic signal analysis apparatus described in the first and second embodiments can be partially or entirely configured from a personal computer or the like. In this case, each part of the apparatus described above can realize the function by computer hardware or software. For example, a program for causing a computer to execute part or all of the operations described in the first and second embodiments is stored in a computer hard disk device, C
A recording medium such as a D-ROM or a computer memory or the like may be downloaded and used.

The present invention can be applied to a music search apparatus that searches using the acoustic features of music. By applying the feature amount generated in the present invention to the search tag, it is possible to perform a search that meets the user needs such as “I want to search for a song with a thick sound”. In addition, since the sense of thickness of sound is an important factor that determines the atmosphere of music, the use of this feature makes it possible to search for music that accurately reflects the atmosphere of music.

The present invention also provides a control device for controlling the screen display of lighting devices, air conditioners, toys, visual effects, etc. according to the music genre and music tone, and the sound quality, volume, sound field, etc. according to the music genre, music tone, etc. It can be applied to an audio device to be controlled. Compared to conventional control devices and audio devices, it is possible to control the sense of thickness and excitement of music more accurately.

DESCRIPTION OF SYMBOLS 1 Acoustic signal analyzer 11 Acoustic signal input part 12 Frequency analysis part 13 Stable component detection part 14 Feature-value production | generation part 11a-14a Arithmetic processing circuit 11b A / D converter 13b Stable component memory

Claims (21)

An acoustic signal analyzer for extracting the characteristics of the music from the audio signal related to the music,
Frequency analysis means for performing frequency analysis on the acoustic signal and generating frequency component data composed of each element corresponding to time, frequency, and component intensity;
An element in which the component intensity is a predetermined value or more is detected as an effective element from the frequency component data, and a region in which a predetermined number or more of the effective elements having the same frequency are present within a predetermined time in the frequency component data Stable component detection means for detecting as a stable component;
Feature amount generating means for generating a feature amount of the predetermined section based on the total sum of the strengths of the stable components in the predetermined section or the number of the stable components;
The stable component detecting means adds a value obtained by adding, in the time direction, component intensities of other elements corresponding to the same frequency as the element to be determined as to whether the element is the effective element or a frequency in the vicinity of the same frequency. To calculate the predetermined value,
An acoustic signal analyzer characterized by that. The stable component detecting means adds a value obtained by adding, in the time direction, component intensities of other elements corresponding to the same frequency as the element to be determined as to whether the element is the effective element or a frequency in the vicinity of the same frequency. And calculating an average value of the component strengths and calculating a value obtained by multiplying the average value by a predetermined coefficient as the predetermined value.
The acoustic signal analyzing apparatus according to claim 1, wherein: The stable component detection means is another element that is temporally adjacent to the element that is the determination target of whether or not it is the effective element, and the same frequency as the determination target element, or the vicinity of the same frequency The predetermined value is calculated using a value obtained by adding the component intensities of other elements corresponding to the frequency in the time direction.
The acoustic signal analyzer according to claim 1 or 2, wherein The frequency analysis means is made to correspond to any one of frequencies existing at equal intervals on the frequency axis, a frequency corresponding to an average temperament scale, or a frequency divided more finely than a semitone of the average temperament scale. Generate each element,
The acoustic signal analyzer according to any one of claims 1 to 3, wherein the acoustic signal analyzer is provided. The feature amount generation means calculates the number of the stable components in the predetermined section, the value obtained by dividing the number of the stable components by the number of types of frequencies in the elements constituting the frequency data, or the number of the stable components. The feature amount is a value divided by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section.
The acoustic signal analyzer according to any one of claims 1 to 4, wherein the acoustic signal analyzer is provided. The feature amount generation means is a sum of the strengths of the stable components in the predetermined section, or a value obtained by dividing the sum of the strengths of the stable components by the number of types of frequencies in the elements constituting the frequency data, or the stable A value obtained by dividing the sum of the component intensities by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section is the feature amount.
The acoustic signal analyzer according to any one of claims 1 to 4, wherein the acoustic signal analyzer is provided. The feature amount generation means generates the feature amount by performing a process of smoothing in the time direction.
The acoustic signal analyzer according to any one of claims 1 to 6, wherein An acoustic signal analysis method executed by an acoustic signal analysis device that extracts features of the music from an acoustic signal related to the music,
A frequency analysis step of performing frequency analysis on the acoustic signal and generating frequency component data composed of elements corresponding to time, frequency, and component intensity;
An element in which the component intensity is a predetermined value or more is detected as an effective element from the frequency component data, and a region in which a predetermined number or more of the effective elements having the same frequency are present within a predetermined time in the frequency component data A stable component detection step for detecting as a stable component;
A feature amount generation step of generating a feature amount of the predetermined section based on a sum of the strengths of the stable components in a predetermined section or the number of the stable components;
In the stable component detection step, a value obtained by adding in the time direction the component intensity of another element corresponding to the same frequency as the element to be determined whether or not it is the effective element or a frequency in the vicinity of the same frequency. To calculate the predetermined value,
An acoustic signal analysis method characterized by the above. In the stable component detection step, a value obtained by adding in the time direction the component intensity of another element corresponding to the same frequency as the element to be determined whether or not it is the effective element or a frequency in the vicinity of the same frequency. And calculating an average value of the component strengths and calculating a value obtained by multiplying the average value by a predetermined coefficient as the predetermined value.
The acoustic signal analysis method according to claim 8. The stable component detection step is another element that is temporally adjacent to the element that is the determination target of whether it is the effective element, and is the same frequency as the determination target element, or the vicinity of the same frequency The predetermined value is calculated using a value obtained by adding the component intensities of other elements corresponding to the frequency in the time direction.
The method for analyzing an acoustic signal according to claim 8 or 9, wherein: The frequency analysis step is made to correspond to any one of a frequency that exists at equal intervals on the frequency axis, a frequency corresponding to an average temperament scale, or a frequency divided more finely than a semitone of the average temperament scale. Generate each element,
The acoustic signal analysis method according to any one of claims 8 to 10, wherein: In the feature amount generation step, the number of the stable components in the predetermined section, the value obtained by dividing the number of the stable components by the number of types of frequencies in the elements constituting the frequency data, or the number of the stable components is calculated. The feature amount is a value divided by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section.
The acoustic signal analysis method according to any one of claims 8 to 11, wherein: In the feature amount generation step, the sum of the strengths of the stable components in the predetermined section, or a value obtained by dividing the sum of the strengths of the stable components by the number of types of frequencies in the elements constituting the frequency data, or the stable A value obtained by dividing the sum of the component intensities by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section is the feature amount.
The acoustic signal analysis method according to any one of claims 8 to 11, wherein: The feature amount generation step generates the feature amount by performing a process of smoothing in the time direction.
The acoustic signal analysis method according to any one of claims 8 to 13, wherein the method is an acoustic signal analysis method. An acoustic signal analysis program for extracting features of the music from an audio signal related to the music,
A frequency analysis step of performing frequency analysis on the acoustic signal and generating frequency component data composed of elements corresponding to time, frequency, and component intensity;
An element in which the component intensity is a predetermined value or more is detected as an effective element from the frequency component data, and a region in which a predetermined number or more of the effective elements having the same frequency are present within a predetermined time in the frequency component data A stable component detection step for detecting as a stable component;
Causing the computer to execute a feature amount generation step of generating a feature amount of the predetermined section based on the total sum of the strengths of the stable components in the predetermined section or the number of the stable components;
In the stable component detection step, a value obtained by adding in the time direction the component intensity of another element corresponding to the same frequency as the element to be determined whether or not it is the effective element or a frequency in the vicinity of the same frequency. To calculate the predetermined value,
An acoustic signal analysis program characterized by that. In the stable component detection step, a value obtained by adding in the time direction the component intensity of another element corresponding to the same frequency as the element to be determined whether or not it is the effective element or a frequency in the vicinity of the same frequency. And calculating an average value of the component strengths and calculating a value obtained by multiplying the average value by a predetermined coefficient as the predetermined value.
The acoustic signal analysis program according to claim 15. The stable component detection step is another element that is temporally adjacent to the element that is the determination target of whether it is the effective element, and is the same frequency as the determination target element, or the vicinity of the same frequency The predetermined value is calculated using a value obtained by adding the component intensities of other elements corresponding to the frequency in the time direction.
The acoustic signal analysis program according to claim 15 or 16, characterized in that The frequency analysis step is made to correspond to any one of a frequency that exists at equal intervals on the frequency axis, a frequency corresponding to an average temperament scale, or a frequency divided more finely than a semitone of the average temperament scale. Generate each element,
The acoustic signal analysis program according to any one of claims 15 to 17, wherein the program is an acoustic signal analysis program. In the feature amount generation step, the number of the stable components in the predetermined section, the value obtained by dividing the number of the stable components by the number of types of frequencies in the elements constituting the frequency data, or the number of the stable components is calculated. The feature amount is a value divided by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section.
The acoustic signal analysis program according to any one of claims 15 to 18, wherein the program is an acoustic signal analysis program. In the feature amount generation step, the sum of the strengths of the stable components in the predetermined section, or a value obtained by dividing the sum of the strengths of the stable components by the number of types of frequencies in the elements constituting the frequency data, or the stable A value obtained by dividing the sum of the component intensities by the product of the number of types of frequencies in the elements constituting the frequency data and the length of the predetermined section is the feature amount.
The acoustic signal analysis program according to any one of claims 15 to 18, wherein the program is an acoustic signal analysis program. The feature amount generation step generates the feature amount by performing a process of smoothing in the time direction.
The acoustic signal analysis program according to any one of claims 15 to 20, wherein the program is an acoustic signal analysis program.

Images