JP5480842B2 - Clip noise detection device, clip noise detection method, program - Google Patents

Description

  The present invention relates to a clip noise detection device, a clip noise detection method, and a program for detecting a clip noise occurrence time from a recorded signal.

  When recording voice or music, if the input sound pressure level becomes excessive and exceeds the allowable signal level, a unique distortion occurs in the waveform of the recorded sound source. This distortion is called clip noise.

  In general, in order to prevent the occurrence of clip noise, record the sound after measuring the dynamic range of the sound to be recorded in advance or checking whether the maximum volume level that can be recorded by recording the sound is exceeded. There is. In addition, while recording, monitor the recording volume with a peak meter, etc., and in the unlikely event that it exceeds the maximum volume level, re-record, etc., basically clip by exceeding the maximum volume by human monitoring Noise is not generated. On the other hand, a clip noise prevention technique has been developed in which the maximum amplitude value of a signal is monitored and the headroom is automatically adjusted using the maximum amplitude value as a feature amount (see Non-Patent Document 1).

  In the past, when checking whether clip noise has occurred in the recorded sound source, a method of reproducing the recorded sound source and listening to it for confirmation by humans has been common.

Atsushi Miura, Hiroshi Nakajima, Shoji Makino, Takeshi Yamada, Kazuhiro Nakajo, "Restoration of Clipped Acoustic Signals", Proceedings of the Acoustical Society of Japan, March 11, 2011, p.941-944

  When the occurrence of clip noise is manually monitored, even an amateur, such as a professional audio engineer such as a professional mixer, may often cause inadvertent monitoring omission or level over due to the accuracy of the display device itself. Even if the clip noise is inaudible to human hearing, there is a case where large distortion occurs in the spectrum. When various digital analysis processing and statistical processing are performed using the recorded sound, waveform distortion due to clip noise that is inaudible in terms of audibility causes an unexpected error. On the other hand, in the case of digital recording, there is a conventional technique for mechanically monitoring samples that take values in the vicinity of the maximum range, but the existence of samples having values in the vicinity of the maximum range of the quantization width is examined. However, due to problems such as device settings, there is a problem that it is impossible to detect when clip noise occurs outside the maximum range. Therefore, the present invention provides a clip noise detection apparatus that can automatically analyze a recorded sound source and accurately detect a clip noise occurrence time.

  The clip noise detection apparatus of the present invention includes a filter bank processing unit, an orthogonal detection unit, a clip strength calculation unit, and a determination unit. The filter bank processing unit receives a signal, passes the input signal through a filter bank including N filters (N is an integer of 2 or more), and outputs N subband signals. The quadrature detection unit receives N subband signals, performs quadrature detection on the N subband signals, and outputs N complex analysis signals. The clip strength calculation unit receives N complex analysis signals as input, obtains a correlation between any two complex analysis signals selected from the N complex analysis signals, and clips each time based on the obtained correlation. The intensity value is calculated and the calculated clip intensity value is output. The determination unit receives the clip strength value, and outputs the time when the excess occurs as the clip noise occurrence time when the clip strength value exceeds a predetermined threshold.

  According to the clip noise detection apparatus of the present invention, the recorded sound source can be automatically analyzed to accurately detect the clip noise occurrence time.

1 is a block diagram illustrating a configuration of a clip noise detection device according to Embodiment 1. FIG. FIG. 3 is a block diagram illustrating a configuration of a filter bank processing unit according to the first embodiment. FIG. 3 is a block diagram illustrating a configuration of a quadrature detection unit according to the first embodiment. FIG. 3 is a block diagram illustrating a configuration of a clip strength calculation unit according to the first embodiment. 5 is a flowchart illustrating the operation of the clip noise detection apparatus according to the first embodiment. The figure explaining the passage characteristic of the filter bank of a filter bank process part. The figure which shows the example of clip noise generation time detection.

  Hereinafter, embodiments of the present invention will be described in detail. In addition, the same number is attached | subjected to the structure part which has the same function, and duplication description is abbreviate | omitted.

  First, an outline of the clip noise detection apparatus 10 according to the first embodiment will be described with reference to FIGS. 1 and 5. FIG. 1 is a block diagram illustrating a configuration of a clip noise detection apparatus 10 according to the present embodiment. FIG. 5 is a flowchart showing the operation of the clip noise detection apparatus 10 according to the present embodiment. The clip noise detection apparatus 10 according to the present exemplary embodiment includes a filter bank processing unit 11, a quadrature detection unit 12, a clip strength calculation unit 13, a determination unit 14, and a threshold storage unit 15. The filter bank processing unit 11 receives a signal, passes the input signal through a filter bank including N filters (N is an integer of 2 or more), and outputs N subband signals (S11). The quadrature detection unit 12 receives N subband signals, performs quadrature detection on the N subband signals, and outputs N complex analysis signals (S12). The clip strength calculation unit 13 receives N complex analysis signals, obtains a correlation between two arbitrary complex analysis signals selected from the N complex analysis signals, and calculates the correlation between the obtained correlations for each time. The clip strength value is calculated, and the calculated clip strength value is output (S13). The determination unit 14 receives the clip strength value, and when the clip strength value exceeds a predetermined threshold, outputs the time when the excess has occurred as the clip noise occurrence time (S14). At this time, the threshold used by the determination unit 14 is stored in the threshold storage unit 15 in advance. The threshold value used by the determination unit 14 can be obtained as follows.

1) Unsupervised discrimination threshold In order to uniquely determine the presence or absence of a clip from data, a threshold is determined in advance without prior learning, and when the clip intensity value exceeds the unsupervised discrimination threshold, the time is set as clip noise. Time of occurrence.

2) Supervised discrimination threshold When the signal data that the clip noise is known in advance is present to such an extent that it can be used for learning, the clip strength of the clipped portion and the unclipped portion The value is learned in advance by a machine learning method such as SVM (Support Vector Machine), and the clip strength value of the actual signal is input to the SVM for discrimination.

  As for which method of 1) and 2) should be adopted, if there is a considerable amount of clip noise pattern data in advance, the method of 2) will be used. If it is unclear how clip noise occurs, the method 1) may be considered. However, even if clip noise pattern data exists, if the variation is large, the method 1) is desirable.

  Next, the configuration and operation of the filter bank processing unit 11 will be described in detail with reference to FIGS. FIG. 2 is a block diagram illustrating a configuration of the filter bank processing unit 11 according to the present embodiment. FIG. 6 is a diagram for explaining the pass characteristics of the filter bank of the filter bank processing unit 11. As shown in FIG. 2, the filter bank processing unit 11 includes N filter groups (filter banks). The filter bank is composed of one low-pass filter, (N−2) band-pass filters, and one high-pass filter. Input signals are simultaneously input to the filters. For each filter, the correlation of the output signal may be obtained, and it is simplest to use an FIR filter having a linear phase characteristic. However, an IIR filter may be used in consideration of the phase characteristic. Next, the cutoff frequency of each filter will be described. Regarding the frequency characteristics of the filter, there are two types of configuration methods as shown in FIGS.

(A) In this case, the passbands are overlapped. As shown in FIG. 6A, the cutoff frequency of each filter is the center frequency of the passbands of adjacent filters. In other words, if the i-th filter has a low-pass cutoff frequency Fl _i , a high-pass cutoff frequency Fh _i , and a center frequency Fc _i.

となる。

It becomes.

(B) The passbands are not overlapped. In this case, as shown in FIG. 6B, the low-frequency cutoff frequency of each filter is the same as the high-frequency cutoff frequency of the adjacent filter on the low frequency side. Further, the high-frequency cutoff frequency of each filter is the same as the low-frequency cutoff frequency of the adjacent filter on the high frequency side. That is

となる。

It becomes.

Which of (a) and (b) is suitable depends on the type and purpose of the input signal. If the input signal is voice and the purpose is to detect a clip sound that a person is interested in hearing, the center frequency of the filter is not set as an equal division point of the frequency band as shown in (a), but the auditory sense It is desirable to set equal division points on a typical mel frequency axis. When the input signal is music and it is desired to detect clip noise that is a problem in digital processing, the center frequency can be set to an equal division point on the linear frequency axis as shown in (b).

  Next, the configuration and operation of the quadrature detection unit 12 will be described in detail with reference to FIGS. 3 and 5. FIG. 3 is a block diagram illustrating a configuration of the quadrature detection unit 12 according to the present embodiment. As shown in FIG. 3, the quadrature detection unit 12 includes quadrature detection means 12-a and low-pass means 12-b. Quadrature detection is one method for obtaining an analytic signal by temporal processing of a signal. Generally, it is expressed by the following formula.

Here, x (t) is an input signal, and ω _k is a frequency shift amount. A complex analysis signal is obtained by cutting a band above half of the bandwidth of the input signal with a low-pass filter with respect to B (t). The amplitude of the complex analysis signal is an estimated value of the instantaneous amplitude. Orthogonal detection means 12-a, as a time of t, as input N sub-band signals, any i (i = 1,2,3, ..., N) for, i th sub-band signals _x i ( When the center frequency of the i-th filter corresponding to t) is Fc _i , B _i (t) = x _i (t) exp (−jFc _i t) is calculated and B _i (t) is calculated. Output (SS12a). The low-pass means 12-b receives the B _i (t) for an arbitrary i, performs low-pass filtering on the B _i (t), and outputs a complex analysis signal A _i (t) (SS12b).

Next, the configuration and operation of the clip strength calculation unit 13 will be described in detail with reference to FIGS. 4 and 5. FIG. 4 is a block diagram illustrating the configuration of the clip strength calculation unit 13 according to the present embodiment. As shown in FIG. 4, the clip strength calculation unit 13 includes a correlation value calculation unit 13-a and an intensity value calculation unit 13-b. Correlation value calculation means 13-a receives N complex analysis signals (A ₁ (t), A ₂ (t),..., A _N (t)) and inputs any k (k = 1, 2, 3, ..., N-1)

And the R _k (t) is output as a correlation value (SS13a). The intensity value calculation means 13-b receives the correlation value R _k (t) for (k = 1, 2, 3,..., N−1) and uses k ₀ as a predetermined coefficient.

And C (t) is output as a clip strength value (SS13b). Here, k ₀ should be appropriately set according to the type of the input signal, but is preferably a value from 5 to N-5. As described above, the determination unit 14 receives the clip strength value, and when the clip strength value exceeds a predetermined threshold, outputs the time when the excess has occurred as the clip noise occurrence time (S14).

  The clip noise actually detected by the clip noise detection apparatus 10 of the present embodiment will be described with reference to FIG. FIG. 7 is a diagram showing an example of clip noise occurrence time detection. FIG. 7A shows an example of an input signal. FIG. 7B illustrates one of the subband signals obtained by performing filter bank processing (S11) on the input signal of FIG. FIG. 7C illustrates the instantaneous amplitude of the complex analysis signal obtained by performing low-pass filtering (SS12b) after quadrature detection (SS12a) on the subband signal of (b). FIG. 7D illustrates the clip strength value (SS13b) obtained by calculating the correlation value of the complex analysis signal of (c) (SS13a) and calculating from the average value of the correlation values. As shown by a thick solid line in FIG. 7D, a threshold (either unsupervised or supervised) is set in advance, and a time indicating a clip intensity value exceeding the threshold is detected as a clip noise occurrence time. Can do. Since the clip location in FIG. 7A and the estimated clip position in FIG. 7D substantially coincide, the clip noise detection apparatus 10 of this embodiment can accurately detect the clip noise occurrence time. I know that there is.

  As described above, the clip noise detection apparatus 10 according to the present embodiment divides the input signal into a plurality of bands, and observes all or a part of the correlation of the detection signals for each band, so that the clip noise is detected in all bands. It is possible to capture a phenomenon in which noise is superimposed across. A normal music signal or audio signal has a signal in a part of the band, and the signal does not exist over the entire band. On the other hand, in the case of simple noise (white noise), signals exist in all bands, but the signal characteristics of each band are random and have no correlation. Therefore, the clip noise occurrence time can be accurately detected by calculating the correlation of the band-divided signal and detecting the one having a high correlation. Also, the time resolution can be increased by executing all the band division processing and the estimation of the instantaneous amplitude of the signal after the band division by the filter processing on the time axis. For this reason, even a very instantaneous clip can be detected.

  In addition, the various processes described above are not only executed in time series according to the description, but may be executed in parallel or individually according to the processing capability of the apparatus that executes the processes or as necessary. Needless to say, other modifications are possible without departing from the spirit of the present invention.

  Further, when the above-described configuration is realized by a computer, processing contents of functions that each device should have are described by a program. The processing functions are realized on the computer by executing the program on the computer.

  The program describing the processing contents can be recorded on a computer-readable recording medium. As the computer-readable recording medium, for example, any recording medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, and a semiconductor memory may be used.

  The program is distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM in which the program is recorded. Furthermore, the program may be distributed by storing the program in a storage device of the server computer and transferring the program from the server computer to another computer via a network.

  A computer that executes such a program first stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its own storage device. When executing the process, the computer reads a program stored in its own recording medium and executes a process according to the read program. As another execution form of the program, the computer may directly read the program from a portable recording medium and execute processing according to the program, and the program is transferred from the server computer to the computer. Each time, the processing according to the received program may be executed sequentially. Also, the program is not transferred from the server computer to the computer, and the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes the processing function only by the execution instruction and result acquisition. It is good. Note that the program in this embodiment includes information that is used for processing by an electronic computer and that conforms to the program (data that is not a direct command to the computer but has a property that defines the processing of the computer).

  In this embodiment, the present apparatus is configured by executing a predetermined program on a computer. However, at least a part of these processing contents may be realized by hardware.

Claims (8)

A clip noise detection device for detecting clip noise occurrence time,
A filter bank processing unit that takes a signal as an input, passes the input signal through a filter bank including N (N is an integer of 2 or more) filters, and outputs N subband signals;
A quadrature detection unit that receives the N subband signals, performs quadrature detection on the N subband signals, and outputs N complex analysis signals;
Using the N complex analysis signals as input, obtain the correlation between any two complex analysis signals selected from the N complex analysis signals, and calculate the clip strength value for each time from the obtained correlation A clip strength calculation unit for outputting the calculated clip strength value;
A clip noise detection comprising: a determination unit that receives the clip intensity value as input and outputs a time when the excess occurs as a clip noise occurrence time when the clip intensity value exceeds a predetermined threshold value apparatus. The clip noise detection device according to claim 1,
The orthogonal detection unit is
t is time, and the N subband signals are input. For any i (i = 1, 2, 3,..., N), i-th corresponding to the i-th subband signal x _i (t) the center frequency of the filter when the Fc _i, and orthogonal detection means to calculate a _{B i (t) = x i} (t) exp (-jFc i t), and outputs the _B i (t),
For any i, clip as input the B _{i (t),} to the B _{i (t),} subjected to low-pass filtering, characterized in that it comprises a low-pass means for outputting a complex analytic signal A _{i (t)} Noise detection device. The clip noise detection device according to claim 1 or 2,
The clip strength calculator
The N complex analysis signals (A ₁ (t), A ₂ (t),..., A _N (t)) are input and arbitrary k (k = 1, 2, 3,..., N−1). Against

And a correlation value calculation means for outputting R _k (t) as a correlation value;
For (k = 1, 2, 3,..., N−1), the correlation value R _k (t) is input, and k ₀ is a predetermined coefficient.

Intensity value calculating means for calculating C (t) as a clip intensity value;
A clip noise detection apparatus comprising: The clip noise detection device according to any one of claims 1 to 3,
A clip noise detection apparatus, wherein the threshold is learned in advance by supervised learning. A clip noise detection method for detecting clip noise occurrence time,
A filter bank processing step that takes a signal as an input, passes the input signal through a filter bank composed of N (N is an integer of 2 or more) filters, and outputs N subband signals;
A quadrature detection step for receiving the N subband signals, performing quadrature detection on the N subband signals, and outputting N complex analysis signals;
Using the N complex analysis signals as input, obtain the correlation between any two complex analysis signals selected from the N complex analysis signals, and calculate the clip strength value for each time from the obtained correlation A clip strength calculation step for outputting the calculated clip strength value;
A clip noise detection comprising: a step of receiving the clip strength value as input and outputting a time when the excess occurs as a clip noise occurrence time when the clip strength value exceeds a predetermined threshold value Method. The clip noise detection method according to claim 5,
The orthogonal detection step includes:
t is time, and the N subband signals are input. For any i (i = 1, 2, 3,..., N), i-th corresponding to the i-th subband signal x _i (t) the center frequency of the filter when the _{Fc i, B i (t)} = x i (t) exp a (-jFc _i t) calculated, a quadrature detection substep of outputting the _B i (t) ,
And a low-pass sub-step for inputting B _i (t) for an arbitrary i, performing low-pass filtering on the B _i (t), and outputting a complex analysis signal A _i (t). Clip noise detection method. The clip noise detection method according to claim 5 or 6,
The clip strength calculation step includes:
The N complex analysis signals (A ₁ (t), A ₂ (t),..., A _N (t)) are input and arbitrary k (k = 1, 2, 3,..., N−1). Against

And a correlation value calculation sub-step for outputting the R _k (t) as a correlation value;
For (k = 1, 2, 3,..., N−1), the correlation value R _k (t) is input, and k ₀ is a predetermined coefficient.

And calculating an intensity value sub-step for outputting the C (t) as a clip intensity value;
A clip noise detection method comprising:   The program for functioning a computer as a clip noise detection apparatus in any one of Claim 1 to 4.

Images