JPH10134549A - Music program searching-device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve a searching capacity by providing an input means for instructing the features of a music program and a control means for selecting and outputting index information stored in a storage means based on the instruction. SOLUTION: An extraction means 12 extracts index information from power spectrum data that are created by a conversion means 6 for converting a plurality of music program data into power spectrum data to store it in a storage means 8. Then, by referring to index information stored in the input means 12 for instructing the features of a desired music program and the above storage means 8, music program data stored in the storage means 8 are selected and outputted, thus detecting tempo data based on the power spectrum data extracted from the music program and extracting a musical instrument pattern and musical interval data based on the tempo data and hence searching for desired music program data according to, for example, the index information of a plurality of data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a search device suitable for detecting desired audio information from a plurality of audio information groups, and more particularly to a music search device which is effective when applied to music search of a CD or the like.

[0002]

2. Description of the Related Art Conventionally, in a karaoke apparatus, a predetermined song is selected by designating a song designated by operating means or a song number corresponding to a singer's name based on a songbook.

In Japanese Patent Application Laid-Open No. 61-93712,
There is disclosed a sound quality adjustment device which varies a sound quality circuit constant in accordance with the presence / absence of periodicity of a reproduced audio signal waveform and, for example, discriminates between a tone signal and a conversation signal and automatically adjusts the sound quality. .

Further, JP-A-4-30382 discloses that
Correlation degree calculation means for calculating the degree of correlation of each frequency characteristic of each of a plurality of music source groups at predetermined time intervals, and tune selection means for selecting and instructing the tune of music, the music source calculated by the correlation degree calculation means A sound device is disclosed in which the music source is selected by associating the degree of correlation with the tune designated by the tune selection command means.

[0005]

In order to select a desired song from a songbook as described in the above-mentioned conventional configuration, it is necessary to know at least the name of the singer or the song.
In the sound quality adjusting device as disclosed in Japanese Patent No. 3712, it is conceivable to search for a predetermined music piece by using the periodicity of the waveform of the music piece as a parameter for each music piece to be reproduced when the music piece is selected. There was a problem that only the judgment of music and conversation could be made only by judgment using only the parameter as a parameter.

Further, as disclosed in Japanese Patent Application Laid-Open No. Hei 4-30382, the correlation of the frequency characteristic of each music source in the music source group is calculated at predetermined time intervals, and the calculated value is selected by the tune selection command means. In the case of selecting a different tune other than the desired tune in association with the specified tune, the user can select and reproduce a desired image only to understand a plurality of genres (tunes) such as jazz and classical music. Although it is possible to use a vehicle acoustic device for listening, there is a problem that index information is insufficient to select a specific song number of a song having a predetermined song tone from a karaoke device including songs of more genres.

An object of the present invention is to provide a music retrieval apparatus which solves the above-mentioned problems. The first problem to be solved is to extract power spectrum data from a plurality of pieces of music information, , A plurality of index information is extracted from the song, and a desired song is searched based on the index information. Therefore, a desired song can be selected from the image of the music, and a karaoke device can be used to sing the song. Can be obtained.

A second object of the present invention is to provide a music search apparatus capable of searching for desired music data based on a plurality of index information such as tempo information, musical instrument pattern information, and pitch information.

A third problem to be solved by the present invention is to store a piece of index information in a storage means, and to obtain a song search device capable of shortening the index creation time of a new song by using this index information. It is in.

A fourth object of the present invention is to correct the signal level of music data to be newly searched so as to match the average signal level of music data of index information already stored in the storage means, It is an object of the present invention to obtain a music search device capable of correcting the variation in signal level and reducing the number of processes at the time of search.

A fifth problem to be solved by the present invention is that when searching for an index already stored in the storage circuit, the pitch of the pitch index information is increased by a semitone and repeatedly compared and searched, thereby ensuring a predetermined search. The purpose of the present invention is to obtain a music search device capable of detecting the music data of the above.

[0012]

As shown in FIG. 1, the music retrieval apparatus of the present invention extracts music features from a plurality of music data and searches for a desired music. In the music search device 1 described above, a conversion means 6 for converting music data into power spectrum data, detection means 6 and 12 for detecting the power spectrum data converted by the conversion means 6, and a detection means A storage unit 8 for storing the power spectrum data detected in the storage units 8 and 12; an extraction unit 12 for extracting index information from the power spectrum data stored in the storage unit 8 and storing the index information in the storage unit 8; Input means 15 capable of designating the characteristics of
And a control unit 12 for referring to the index information stored in the storage unit 8 based on the instruction from the input unit 15 and selectively outputting the music data stored in the storage unit 8.

According to the music search apparatus of the present invention, the tempo data is detected based on the power spectrum data extracted from the music, and the musical instrument pattern and the pitch data are extracted based on the tempo data. Since the desired music data can be searched by the index information or the like, it is possible to obtain a desired music piece from many genres of music without fail.

[0014]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a music retrieval apparatus according to the present invention.

FIG. 1 is a system diagram showing the overall configuration of a music retrieval apparatus 1 according to the present invention. An analog music signal and digital music data are input to input terminals T ₁ and T _2. Is done.

The analog signal supplied to the input terminal T ₁ is supplied to an analog-digital conversion circuit (hereinafter referred to as ADC) 2, converted into digital data, and supplied to an input selector 3.

Similarly, the CD-R supplied to the input terminal T ₂
Digital data from an OM player or the like is decoded by a CD-ROM decoder 4 or the like and supplied to the input selector 3.

The output terminal of the input selector 3 is connected to a band-pass filter (hereinafter, referred to as PBF) 5, for example, 10H
The signal is limited to the audio audible band of z to 20 kHz and supplied to a digital, signal, and processor (hereinafter referred to as DSP) 6.

The DSP 6 calculates an auto power spectrum density (power spectrum) described later at a high speed, and
And a hard disk (HD) 8a, and a DSP 6 connected to a drive (HDD) 8 and the like.
Exchanges data with the first and second storage means comprising the HD 8a and the RAM 7.

The DSP 6 converts digital data into an analog signal by a digital-analog conversion circuit (hereinafter, referred to as DAC) 9 and supplies the signal to the input terminal T ₁ or T ₂ from the speaker 11 via the amplifier circuit 10 and the output terminal T _3. The analog or digital music is played.

The DSP 6, the HDD 8, the operation unit 15 such as a key input, and the display device 14 such as a cathode ray tube (CRT) are controlled between control means such as a computer (CPU) 12 through a control line shown by a broken line. It is made to control. C
PU12, HDD8, RAM and ROM1 for work
Data is exchanged between the third storage means 3 via a bus.

According to the present invention, the input terminal T
Automatically creates stone index information necessary for the search of music from a CD player or the like that is input based on the music ₁ or T _2, by searching the index information, an image to an input device (operating section) 15 It is an object of the present invention to obtain a music search device that can surely search for a plurality of music pieces (song numbers) specified by the user.

In the present invention, the music data digitized by the ADC 2 or the digital data decoded by the CD-ROM decoder 4 is selected by the input selector 3 and the BPF
5 and is supplied to the DSP 6. This DS
At P6, the power spectrum of the music data is extracted. The power spectrum represents, for example, the root mean square of the above-mentioned music data amount which fluctuates temporally or spatially as a distribution of frequency components.

That is, for example, 10 Hz to 20
As shown in FIG. 2, the musical tone data 16 band-limited to kHz has a plurality of band-limiting BPFs 17a, 17b to 17b.
n, and divides a band of, for example, 10 Hz to 20 kHz into 30 and sampling circuits 18 a, 18 b to 18 b for each frequency band selected by the BPFs 17 a, 17 b to 17 n.
18n and sampled for each specific time axis (unit time).

As shown in FIG. 4, the unit time at the time of sampling is, for example, 0.1 to 0.5 seconds within a predetermined bandwidth 19 selected by the BPF 17a, and the level (amplitude) is 1/256 to 1 /. Select a resolution of 4096. Using this level as table data, the tables 20a of 1 to n,
For example, the music data is created as shown in Table 1 as 20b to 20n, and the frequency and time are used as parameters, and power spectrum data in which the levels are quantified is created.

[0026]

[Table 1]

The above BPFs 17a, 17b to 17n have D
An example in which frequency analysis is performed in SP6, high-speed operation processing is performed, and sampling is performed has been described.
b to 17n may be configured by hardware as shown in FIG.

The power spectrum table data as shown in Table 1 is transferred from the RAM 7 or the like to the HD 8a in the HDD 8 by the CP.
It is stored based on the control of U12.

The power spectrum data obtained as described above represents the distribution of the average energy per unit time, and a signal obtained by extracting the music signal x (t) only in a section of −T ≦ tT is defined as xr (l). The power spectrum φ _xx (ω) of the signal (x) t at the time is expressed by Equation 1.

[0030]

Where xr (ω) is the Fourier transform of xr (t), E · [•] is the initial value, and the autocorrelation function of signal x (t) is φ _xx (2).

## EQU2 ##

As described above, based on the power spectrum data obtained by frequency-analyzing the music data stored in the HD 8a,
Various data serving as various parameters are extracted according to an index information extraction flowchart as shown in FIG.

The operation for individually extracting such index information serving as various parameters will be described below.
In the above description, the audible band frequency of 10 Hz to 20 kHz is BP
F17a, 17b to 17n divided into 30 bands, but actually divided into 15 bands, and the sampling unit time was set to 0.1.
It is enough to divide the resolution (quantization amount) in the level direction by 256 for 1 second.

First, an overall configuration of a method for detecting various parameters will be described with reference to FIG.

The operation of the DSP6 based on the control from the first step S ₁ 16 in the music in the FIG. 5 CPU 12 or BPF17a hardware configuration, the power spectrum data, such as shown in Table 1 based on 17b~17n frequency Analyze.

In general, a long-term average power spectrum of a program such as music data shows a specific frequency characteristic in Western music, light music and the like. Therefore, each music piece may be considered to have a unique power spectrum.

The power spectrum data analyzed by such a DSP is stored in the HD 8a of the HDD 8.

The power spectrum data stored in the HD 8a is called from the tables 20a to 20n of Table 1, and
Detecting the tempo based on the power spectrum data stored in the second step S ₂ in the table 20 a to 20 n.

That is, from the power spectrum, which will be described later,
The tempo of a plurality of musical instruments is detected by sampling a time equal to or more than a specific level with respect to a specific BPF (frequency unique to an instrument). How to identify instruments compared to pre-stored tempo patterns are sampled around the BPF with the maximum level value of the power spectrum, are stored in HD8a tempo and time signature as an index (seventh step S ₇ and 12 step S _12).

The extracts a musical instrument pattern from the data corresponding to the BPF for the next third step S specific instrument in _3. This stored as instrument index HD8a by comparing the tempo data obtained by instruments pattern data and the second step S ₂ that is stored in advance in HD8a (eighth step S ₈ and 13 step S _13).

The fourth step S ₄ the pitch extraction is performed. Pitch extraction were obtained in the second step S _2, it was obtained in the tempo extraction data and the third step S _3, create a chart to be described later based on instrument pattern data, corresponding to beat time B
The PF extracts a dense band as a pitch, compares it with a pre-stored pitch pattern, and uses the HD as a pitch index.
Stored in the 8a (ninth step S ₉ and 14 step S _14).

[0041] The fifth step S ₅ the chords extraction is performed. The chord is the chord extracted by extracting data corresponding to the meter time is compared with the chord patterns stored in advance, are stored as chord index D8a (10th step S ₁₀ and the fifteenth step S ₁₅₎ without seeing the further presence of the BPF chords corresponding to the tempo at the sixth step S _6, and stores the rhythm index HD8a compared with rhythm pattern stored in advance (11 steps S ₁₁ and 16 step S ₁₆ ).

The above-mentioned five types of indexes are grouped for each song to become a song-specific index, and a song search is enabled by assigning a number to each group.

The method of extracting the above-mentioned various indexes by the CPU 12 will be described in detail with reference to FIGS.

FIG. 6 shows a flowchart for detecting a tempo index from a power step of a music piece.

In the first step S _2a of FIG.
Looking for BPF indicating a maximum value at the sampling time t ₁ of the table data of the power spectrum in the table 20a~20n stored in D8a (time axis (Table 1 were tabulated data in which the frequency) as a parameter as a level) .
That is, the CPU 12 determines which BPF has the maximum value at the first sampling time t ₁ (see FIG. 4).

In the next second step S _2b , the same kind of processing is repeated until the sampling time t _n (see FIG. 4), whereby n
BPFs are selected.

In the next third step S _2c , the center frequency is extracted by extracting the maximum value among the selected n BPF values.

In the fourth step S _2d , BPFs are grouped. That is, the BPF having the maximum value is grouped and a specific BPF is selected.

In the fifth step S _{2 e} , the upper three BPFs are specified.

In the sixth step S _2f , a time equal to or longer than a specified level is extracted from the specified BPF. That is, as shown in FIG. 8, the sampling time is extracted for each of a plurality of BPF ₁ , BPF ₂ . In this case, the specific level is a threshold value, and the initial value of the threshold value is the maximum value × 0.9.

In the seventh step _S2g , it is determined whether or not the sampling time has regularity. If NO is returned to the head of the sixth step S _2f certain level after 10% down in the eighth step S _2h.

If YES in the seventh step S _2g , the process proceeds to a ninth step S _2i to calculate a regular time interval.

In the tenth step S _2j , the ninth step S _2i
Is stored in the HD 8a of the HDD 8 as tempo data, and the process reaches the end.

Based on the tempo data obtained in this manner, a musical instrument pattern is extracted as shown in FIG. 5 to create a musical instrument index. This flowchart is shown in FIG.

[0055] In FIG 7, identifies a particular time t ₁ ~t _n from Tempo index In the first step S _3a.
For example, t _1, t _3, t can be set a specific time from the tempo data as of _5.

In the second step S _3b , a specific level is set. This specific level initial value is set to the maximum value 0.5.

In the third step S _3c , the sampling time t
Selecting a more specific level of the BPF at ₁ ~t _n.

In the fourth step _S3d , the BPF of the stored data is
Is performed, and if they match, the fifth step S _3e
To specify the instrument, and set the instrument index to H in HDD8.
D8a. In this case, the stored data differs depending on the size of the system.

If the fourth step _S3d does not match, the sixth
_Proceeding to step _S3f , the specific level is changed and the level is raised.

In the seventh step _S3g , it is determined whether or not the level is the maximum. If YES, the process returns to the end, and if NO, the process _{returns to} the head of the fourth step _S3d .

Next, as shown in FIG. 5, pitches are extracted by creating a graph shown in FIG. 10 using the tempo index and the musical instrument pattern index. A flowchart of this pitch extraction will be described with reference to FIG.

In the first step _S4a of FIG. 9, the tempo index is transferred from the HD 8a to the RAM / ROM 13 of the CPU 12.
To load.

In the second step S _{4 b} , the musical instrument index is loaded from the HD 8 a into the RAM / ROM 13 of the CPU 12.

In the third step _S4c , the RAM / ROM 13
The graph shown in FIG. 10 is tabulated above. In the graph of FIG. 10, the vertical axis represents each BPF ₁ to BPF _n corresponding to the musical instrument pattern,
The horizontal axis is the tempo extraction data {beat (time)}.

In the fourth step S _4d , a band with a high BPF for the tempo data is selected from the graph of FIG.
The band of the PF is set as a pitch, and the process proceeds to a fifth step _S4e to store the pitch index in the HD of the HDD 8 as a pitch index.

Next, the operations from the fourth step S _4d to the sixth step S _5a are performed to extract a chord index.
In the sixth step _S5a , the BPF corresponding to the chord is compared with the graph of FIG. 10, and for a specific BPF, data corresponding to the beat time on the horizontal axis of the graph of FIG. 10 is extracted.

[0067] That is, the seventh step S determines whether there is BPF chord for tempo _5b, "YES" is long if H 8 proceed to step S _5c as chord index
It is stored in the HD 8a of the DD 8.

If the seventh step S _{5 b} is “NO”, the ninth step S _{5 b}
Compared to the rhythm pattern proceeds to step S _6a, HD8a matched HDD8 as Rhythm Index if
And if they do not match, it reaches the end.

As described above, the first to fifth steps S in FIG.
The detected tempo ₂ to S _6, instrument pattern, pitch, chord, rhythm is compared as a HD8a prestored in which patterns and sixth to tenth step S ₇ to S ₁₀ in the HDD 8, or 11 the 15 H as step S ₁₂ to S ₁₆
It is stored in the HD 8a of the DD 8.

A label (file name) is added to each index as shown in Table 2 below and stored in the HD 8.

[0071]

[Table 2]

FIG. 11 is a flowchart showing a process in which various indexes stored in the HD 8a of the HDD 8 are compared with those stored in the HD 8a in advance.

[0073] The first step ST ₁ tempo 11,
An index extracted from musical tones such as musical instrument patterns, pitches, chords, rhythms, and the like is compared with indexes corresponding to various tempos, musical instrument patterns, pitches, chords, and rhythms stored in the HD 8a of the HDD 8 in advance. In this case, the label A, the label B, and the label n are compared in this order for all the labels, but if they are simply compared, they cannot be naturally matched. Compare (feature comparison).

(1) Assuming that the music source to be compared and the signal level at the time of obtaining the stored data are different, the comparison is performed so that the average value of the signal level data of the comparison source and the average value of the stored data are matched. The level of the source is calculated, and the correction processing is started. (2) The variation in the signal level of the stored data is calculated and stored in advance at an average level assumed for the system. As the assumed average level, for example, as shown in FIG. 12, an average value as shown by a solid line 21 of about 10 samples of a song is taken. (3) Since the main BPF band differs between a case where a man sings and a case where a woman sings the same music source, if there is a mismatch in the comparison process, the voice formant is changed and the comparison process is performed again. If it is, register it as a new song. (4) In the pitch index comparison, if they do not match, the chromatic scale is raised or lowered on the data, and the comparison is performed again to increase the accuracy.

After such comparison processing, the second step ST
_{As shown in 2} , the CPU 12 determines whether the source exists.
If "YES" is determined that the third music source like a step ST _3. The determination in this case is determined by majority rule, and a match of only one pattern is regarded as a new song.

In the fourth step ST ₄ , each corresponding information is displayed. In this case, input or display can be performed in accordance with the added label (file name).

[0077] The second step ST ₂ is made so as to be registered as a new song proceeds to the fifth step ST ₅ If "NO", leading to the end.

Since the present invention is constructed as described above, the features are extracted from the music and a plurality of various indexes are automatically created and automatically stored in the storage means. It is possible to search for a predetermined music piece only by designating a musical feature or an image of the music piece.

When a new index is newly created, the index information stored in the storage means is prioritized, compared with new music data, and the created index is stored in the storage means. It can be created quickly. When a new frequency or index is created, the signal level of the music data to be newly compared is adjusted to match the average value level of the music data signal of the index information stored in the storage means, and the comparison process is performed. Therefore, variations in signal level are corrected, and the number of processes at the time of retrieval can be reduced.

Further, when searching the pitch index stored in the storage means, the pitch index is raised by a chromatic scale and the search is repeated, so that the music data of the predetermined pitch can be detected with certainty. A music search device suitable for performing image search for music selection of a device or the like can be obtained.

[0081]

According to the music search apparatus of the present invention, a desired music can be quickly selected from a music image and sung by a karaoke apparatus.

[Brief description of the drawings]

FIG. 1 is a system diagram of a music search device of the present invention.

FIG. 2 is a system diagram for power spectrum extraction used in the music search device of the present invention.

FIG. 3 is an explanatory diagram of band division of a power spectrum extraction axis according to the present invention.

FIG. 4 is an explanatory diagram of sampling when extracting a power spectrum according to the present invention.

FIG. 5 is a flowchart of extracting various indexes of the music search device of the present invention.

FIG. 6 is a flowchart of tempo index detection of the music search device of the present invention.

FIG. 7 is a flowchart of musical instrument index detection of the music search device of the present invention.

FIG. 8 is an explanatory diagram at the time of extracting a musical instrument pattern.

FIG. 9 is a flowchart of extracting a pitch, a chord, and a rhythm index of the music search device of the present invention.

FIG. 10 is an explanatory diagram at the time of pitch extraction.

FIG. 11 is a flowchart of storing and comparing an extraction index.

FIG. 12 is an average value explanatory waveform diagram.

[Explanation of symbols]

 5, 17a, 17b to 17n BPF 6 DSP 8 HDD 8a HD 12 CPU 13 RAM / ROM 18a, 18b to 18n Sampling circuit

Claims (5)

[Claims] 1. A music search device configured to extract a feature of a music piece from a plurality of music data pieces and search for a desired music piece, wherein said conversion means converts said music piece data into power spectrum data. Detecting means for detecting the power spectrum data converted by the converting means; storing means for storing the power spectrum data detected by the detecting means; extracting index information from the power spectrum data stored in the storing means; Extraction means to be stored in the storage means; input means capable of instructing the characteristics of the desired music; and reference to the index information stored in the storage means based on an instruction from the input means; A music search device comprising: control means for selecting and outputting music data. 2. The index information includes tempo information,
The music search device according to claim 1, wherein the music search device is at least one of musical instrument pattern information, pitch information, chord information, and rhythm pattern information. 3. The control means, when newly creating the index information, gives priority to the index information stored in the storage means and compares the index information with new music data to create new index information. The music search device according to claim 1 or 2, wherein the music piece is stored in the storage means. 4. The control means, when newly creating the index information, a signal of music data to be newly compared so as to match an average value level of a signal of music data of the index information stored in the storage means. 4. The music search device according to claim 1, wherein the music search device corrects the level of the music and stores the corrected music in the storage means. 5. The control means, when retrieving the index information stored in the storage means, raises the pitch of the pitch index information by a chromatic scale and repeatedly compares the pitches.
The music search device according to any one of claims 1 to 4, wherein the music search device outputs desired music data by searching.