JP2604400B2 - Pitch extraction method and extraction device - Google Patents

Abstract

PURPOSE:To extract a pitch with high accuracy without raising sampling frequency, by correcting pitch data corresponding to the judgement based on the pitch data extracted from the max. value data of the autocorrelation function of a digital acoustic signal before outputting the same. CONSTITUTION:The digital acoustic signal through an acoustic signal input apparatus 8 and an A/D converter 7 is stored in an auxiliary memory device 6 and processed by a CPU 1 operated corresponding to the program from a main memory device 3. The autocorrelation function of the acoustic signal is calculated and pitch data is detected from the max. value data of the calculated autocorrelation function. The presence of the max. autocorrelation function value in the vicinity of the frequency component two times the pitch data is judged and, when said value is present, the value corresponding to the max. point thereof is outputted as corrected pitch data. Therefore, the autocorrelation function in the vicinity of the second max. point becomes larger than that in the vicinity of the first max. point and it is prevented that the pitch data twice real pitch data is obtained and a pitch is extracted with high accuracy without raising sampling frequency.

Description

Description: TECHNICAL FIELD The present invention relates to a pitch extraction method and a pitch extraction device,
The present invention is suitably applied to an automatic music transcription device that converts an acoustic signal into musical score data.

[Prior Art] An acoustic signal such as a singing voice, a humming voice, or a musical instrument sound has a repeating waveform of a basic waveform. In an automatic music transcription device that converts such an acoustic signal into musical score data, in order to accurately determine various information such as a pitch and a pitch in the acoustic signal, first, a repetition frequency (hereinafter, referred to as a pitch frequency) of a basic waveform in the acoustic signal. , And the corresponding cycle is called a pitch cycle, and the combined concept is called pitch) for each analysis cycle.

Such extraction methods include frequency analysis and autocorrelation analysis developed in the field of speech synthesis and speech recognition. The pitch can be extracted without being affected by ambient noise, and the autocorrelation processing is simple. Analysis has traditionally been employed.

[Problems to be Solved by the Invention] By the way, in a conventional automatic transcription apparatus, an autocorrelation function is obtained after converting an acoustic signal into a digital signal. Therefore, an autocorrelation function can be obtained only for each sampling period.

Therefore, the pitch can be extracted only at the resolution determined by this sampling period. If the resolution of the extracted pitch is low, the accuracy such as the pitch and the pitch determined in the processing described later is low.

Therefore, it is conceivable to increase the sampling frequency, but the amount of data to be calculated, such as the autocorrelation function, also increases, and the configuration of the automatic transcription apparatus becomes large and hard, and real-time processing may not be realized. There is.

The present invention has been made in consideration of the above points, and has as its object to provide a pitch extracting method and an extracting device capable of extracting a pitch of an audio signal with high accuracy without increasing a sampling frequency. .

[Means for Solving the Problems] In order to solve the problems, according to a first aspect of the present invention, an analog / digital conversion means for converting an input audio signal into a digital signal, and a predetermined processing procedure are stored. Storage means, and control means for executing a processing procedure stored in the storage means, the control means takes in the audio signal converted to a digital signal, calculates an autocorrelation function of the audio signal, and calculates Pitch information is detected from the detected maximum value information of the autocorrelation function, and it is determined whether or not a maximum point of the autocorrelation function exists near a frequency component twice as large as the detected pitch information. A value corresponding to the maximum point is output as pitch information.

Further, in the second aspect of the present invention, an analog / digital conversion means for converting an input audio signal into a digital signal, an autocorrelation function calculation means for calculating an autocorrelation function of the audio signal converted into a digital signal, Pitch information detecting means for detecting pitch information from the calculated maximum value information of the autocorrelation function, and determining whether or not a maximum point of the autocorrelation function exists near a frequency component twice as large as the detected pitch information. And a pitch information correction unit that outputs a value corresponding to the maximum point as pitch information when it exists.

[Operation] Both the first and second aspects of the present invention are intended to extract pitch information of an audio signal using an autocorrelation function. Here, since the autocorrelation function is obtained from the acoustic signal converted into a digital signal, the autocorrelation function can be obtained only at each sampling period, and the pitch information is extracted from the maximum value information of the autocorrelation function. Then, the autocorrelation function value near the second maximum point may be larger than the autocorrelation function value near the first maximum point, and twice the pitch information of the true pitch information may be extracted.

Therefore, in the first aspect of the present invention, the pitch information obtained from the autocorrelation function is not immediately used as the output pitch information, but the maximum point exists near the frequency component twice as large as the detected pitch information. It is determined whether or not there is, and if there is, a value corresponding to the maximum point is used as output pitch information.

In the present invention shown in FIG. 2, the autocorrelation function calculating means calculates the autocorrelation function of the audio signal converted into the digital signal via the analog / digital conversion means, and calculates the autocorrelation function of the calculated autocorrelation function. The pitch information is extracted from the maximum value information by the pitch information detecting means, and it is determined by the pitch information correcting means whether or not the maximum point of the autocorrelation function exists near the frequency component twice as high as the extracted pitch information. In this case, a value corresponding to the maximum point is output as pitch information.

Hereinafter, an embodiment in which the present invention is applied to an automatic transcription apparatus will be described in detail with reference to the drawings.

Automatic Music Transcription Device First, an automatic music transcription device to which the present invention is applied will be described.

In FIG. 3, a central processing unit (CPU) 1 controls the entire automatic transcription apparatus, and is stored in a main storage device 3 connected via a bus 2 as shown in FIG. This is to execute a music transcription processing program.
In the bus 2, in addition to the CPU 1 and the main storage device 3, a keyboard 4 as an input device, a display device 5 as an output device,
An auxiliary storage device 6 and an analog / digital converter 7 used as a working memory are connected.

An audio signal input device 8, for example, a microphone, is connected to the analog / digital converter 7. The acoustic signal input device 8 captures an acoustic signal such as a singing or humming uttered by a user or a musical tone generated from a musical instrument and converts the signal into an electric signal, and converts the electric signal into an analog / digital converter 7. Is output to

When a process is instructed by the keyboard 4, the CPU 1 starts the transcription process, executes a program stored in the main storage device 3, and executes the analog / digital converter 7.
The audio signal converted into a digital signal by the above is temporarily stored in the auxiliary storage device 6, and thereafter, the above-described audio signal is converted into musical score data by executing the above-mentioned program, and is then displayed on the display device 5.
Output.

Next, the transcription process performed by the CPU 1 after capturing the audio signal will be described in detail with reference to the flowchart shown in FIG.

First, the CPU 1 performs an autocorrelation analysis of the acoustic signal, extracts a pitch frequency of the acoustic signal at each analysis cycle, and performs a square-sum processing on the acoustic signal to extract power, and then performs noise removal, smoothing processing, and the like. Execute post-processing (steps SP1, SP
2). Thereafter, for the pitch information, the CPU 1 calculates a shift amount of the pitch axis of the acoustic signal with respect to the absolute pitch axis, and executes a tuning process of shifting the obtained pitch information according to the shift amount (step SP3). . That is, the pitch information is corrected so that the difference between the pitch axis and the absolute pitch axis of the singer or musical instrument that has generated the acoustic signal is small.

Next, the CPU 1 obtains a continuous period of pitch information in which the obtained pitch information is considered to indicate the same pitch,
A process (hereinafter, referred to as segmentation) for dividing the acoustic signal into sections (segments) for each sound is performed, and a segmentation is performed based on the obtained change in power information (steps SP4 and SP5). On the basis of these two pieces of segment information thus obtained, the CPU 1 calculates a reference length corresponding to a time length such as a quarter note length or an eighth note length, and performs more detailed segmentation based on this reference length. (Step SP6).

Based on the pitch information of the segment thus segmented, the CPU 1 identifies the pitch of the segment as a pitch on the absolute pitch axis that can be determined that the pitch information is the closest, and further, determines the pitch of the identified continuous segment. The segmentation is performed again based on whether or not the pitches are the same (steps SP7 and SP8).

Thereafter, the CPU 1 obtains a product sum of the frequency of appearance of the pitch obtained by summing the pitch information and a predetermined weighting coefficient determined according to the key, and based on the maximum information of the product sum, for example, C major, Determine the key of the music of the input audio signal, such as a minor key, and review the pitch for pitch information for the predetermined pitch on the scale in the determined key, and confirm the pitch,
Correct (steps SP9 and SP10). Next, the CPU 1 executes a review of the segmentation based on whether or not there is the same continuous segment from the finally determined pitch or whether there is a change in power between the continuous segments, Final segmentation is performed (step SP11).

When the pitch and duration are determined in this manner, the CPU 1 determines that the music starts from the first beat, the last sound of the phrase does not extend to the next measure, and there is a break between measures. A beat is determined from the extracted bar information and the segmentation information, and a tempo is determined from the determined beat information and the length of the bar (steps SP12 and SP13).

Then, the CPU 1 arranges information on the determined pitch, pitch (segment), tone, time signature, and tempo so far, and finally creates musical score data (step SP14).

Extraction of pitch information Next, extraction of pitch information in an automatic transcription apparatus that performs transcription by performing such processing (see step SP1)
Will be described in more detail.

FIG. 1 shows a detailed flowchart of such pitch information extraction processing. First, the CPU 1 first sets the N acoustic signals y in the analysis window including the sampling point of interest s and thereafter.
(T) (t = 0,..., N−1;
From the sampling number), the following equation of the audio signal and the audio signal obtained by shifting the audio signal by τ with respect to the target sampling point s Autocorrelation function φ (τ) (τ = 0,..., N−1; u =
0,..., N−1−τ) (step SP20). FIGS. 2A and 2B show the autocorrelation function curves obtained in this manner.

Next, the CPU 1 expresses the deviation amount z which gives the maximum value of the function φ (τ) with a deviation amount other than 0 from the values of the N autocorrelation functions φ (τ), that is, the scale of the sampling number. Detect the pitch period of the sound signal (step SP2
1).

Thereafter, the CPU 1 determines a shift amount z-
Autocorrelation functions φ (z−1), φ of three points of 1, z + 1
(Z) and φ (z + 1) are taken out, and a parameter A shown in the following equation A = {φ (z−1) + 2 · φ (z) + φ (z + 1)} 4 (1) is calculated (step SP22) , SP2
3). Note that the parameter A is an autocorrelation function φ (z−
1), a weighted average of φ (z) and φ (z + 1).

After the end of such processing, the CPU 1 determines whether the autocorrelation function φ of any of the amounts y and y + 1 is closest to the shift amount z / 2 that is half of the shift amount z.
(Y) and φ (y + 1) are taken out, and a parameter B shown in the following equation B = {φ (y) + φ (y + 1)} 2 (2) is calculated (steps SP24 and SP2).
Five). The parameter B is an autocorrelation function φ (y),
It is the average of φ (y + 1).

After that, the CPU 1 compares the two parameters A and B,
If the parameter A is larger than the parameter B, the shift amount z is selected as the shift amount τp corresponding to the pitch (steps SP26 and SP27). On the other hand, parameter B is
If it is larger, the shift amount z / 2 is selected as the shift amount τp corresponding to the pitch (step SP28).

The reason why the shift amount giving the maximum value of the autocorrelation function is not immediately set as the pitch period is that, as shown in FIG. 2A, the shift amount is twice the shift amount giving the true maximum value. If the amount is approximately equal to the sampling point and the amount of deviation giving the true maximum is halfway between the sampling points, the autocorrelation function near the second maximum point may be detected as giving the maximum. Because of this, it is checked whether or not this is the case based on the magnitudes of the parameters A and B. If the case does not correspond to the shift amount that gives the true maximum value, half of the shift amount corresponds to the pitch period. To do that.

FIG. 2B shows a case where the value near the first maximum point is detected as the maximum value from the beginning. In this case, as shown in FIG. A is always larger than the parameter B, and the obtained shift amount z is directly used as a pitch period in the following processing.

The CPU 1 calculates the pitch frequency fp from the pitch period τp expressed in the scale of the sampling number obtained in this way according to the following formula fp = fs / τp (3) and shifts to the next processing ( Step SP29). Here, fs is a sampling frequency.

Therefore, according to the above-described embodiment, even when the autocorrelation function near the second maximum point becomes the maximum value due to the sampling period, this is detected to correct the pitch period. As a result, pitch information can be extracted more accurately than before without increasing the sampling frequency, and subsequent segmentation, pitch identification processing, key determination processing, and the like can be executed accurately.

Other Embodiments In the above-described embodiment, the parameters A and B used when determining whether or not the shift amount giving the maximum value is a shift amount corresponding to the vicinity of the true peak are weighted average values. Although shown, the determination may be made using other parameters.

Further, in the above-described embodiment, the present invention is applied to an automatic transcription apparatus. However, the present invention can be applied to various apparatuses which require a process of extracting pitch information from an audio signal.

Further, in the above-described embodiment, the CPU 1 executes all the processes shown in FIG. 4 according to the program stored in the main storage device 3. However, some or all of the processes are executed by hardware. The configuration may be executed. For example, as shown in FIG. 5 in which the same reference numerals are given to the corresponding parts in FIG. 3, the sound signal from the sound signal input device 8 is amplified via the amplifier circuit 10 and then further passed through the pre-filter 11. The digital signal is converted to a digital signal by an analog / digital converter 12, and the acoustic signal converted to the digital signal is subjected to autocorrelation analysis by a signal processor 13 to extract pitch information, and is processed by sum-of-squares processing to obtain power. Information may be extracted and provided to the software processing system by the CPU 1. As the signal processor 13 used in such a hardware configuration (10 to 13), a processor (for example, a processor that can process a signal in a voice band in real time and is provided with an interface signal with a host CPU 1)
NEC Corporation μPD7720) can be applied.

[Effects of the Invention] As described above, according to the present invention, there is provided a pitch extraction method and an extraction apparatus that can extract pitch information more accurately than before using the autocorrelation function without increasing the sampling frequency. Obtainable.

[Brief description of the drawings]

FIG. 1 is a flowchart showing pitch extraction processing according to one embodiment of the present invention, FIG. 2 is a schematic diagram showing an autocorrelation function curve used for pitch extraction processing, and FIG. 3 is an automatic transcription system to which the present invention is applied. FIG. 4 is a block diagram showing the structure of the automatic transcription apparatus, FIG.
The figure is a block diagram showing another configuration of the automatic music transcription device. 1 ... CPU, 3 ... main storage device, 6 ... auxiliary storage device, 7
... Analog / digital converter, 8 ... Acoustic signal input device.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Masaki Fujimoto 5-7-15 Shiba, Minato-ku, Tokyo Inside NEC Technical Information System Development Co., Ltd. (72) Inventor Masanori Mizuno 5-7-1 Shiba, Minato-ku, Tokyo No. 15 Examiner, NEC Technical Information Systems Development Co., Ltd. Shigeo Arai

Claims (2)

(57) [Claims] 1. An analog / digital converter for converting an input acoustic signal into a digital signal, a storage for storing predetermined processing, and a control for executing a processing procedure stored in the storage. The control means takes in the audio signal converted to a digital signal, calculates an autocorrelation function of the audio signal, detects pitch information from the maximum value information of the calculated autocorrelation function, and performs detection. A pitch extraction method for determining whether or not a maximum point of the autocorrelation function exists near a frequency component twice as large as the pitch information obtained above, and outputting a value corresponding to the maximum point as pitch information when it exists. 2. An analog / digital conversion means for converting an input audio signal into a digital signal, an autocorrelation function calculation means for calculating an autocorrelation function of the audio signal converted into a digital signal, Pitch information detecting means for detecting pitch information from the maximum value information of the correlation function; and determining whether or not a maximum point of the autocorrelation function exists near a frequency component twice as large as the detected pitch information. A pitch information correcting unit for outputting a value corresponding to the maximum point as pitch information in a case.