JP2009296298A - Sound signal processing device and method - Google Patents

Abstract

An object of the present invention is to suppress an unnatural sound temporal fluctuation feeling such as a breathing phenomenon caused by a single band DRC or AGC for an audio signal.
A frequency analysis unit that performs frequency analysis of an input audio signal, and a loudness smoothing time constant that calculates time constants of a loudness smoothing unit and a gain smoothing unit based on the frequency analysis result, respectively. When the calculation unit 170 and the gain smoothing time constant calculation unit 180 are provided and the low frequency component is larger than the predetermined value and / or the ratio of the low frequency component to the total frequency component is larger than the predetermined value Therefore, by increasing the time constant, it is possible to obtain an audio signal that is easy to hear and suppresses the temporal fluctuation of unnatural sound.
[Selection] Figure 1

Description

  The present invention relates to an audio signal processing apparatus and method for controlling the dynamic range of an audio signal.

  As an audio signal processing method that appropriately suppresses the dynamic range of the audio signal to make the audio signal easy to hear, Dynamic Range Control / Compression (DRC) or Automatic Gain Control (Automatic Gain Control): AGC) is known.

  FIG. 6 is a block diagram showing a configuration of a conventional audio signal processing apparatus using DRC described in Non-Patent Document 1. In FIG. 6, 600 is a multiplication unit, 610 is a loudness measurement unit, 620 is a gain calculation unit, 630 is a gain smoothing unit, and 640 is a gain smoothing time constant calculation unit. The operation will be described below.

  The loudness measuring unit 610 measures and outputs the loudness (a measure of the volume of sound by hearing) of the input audio signal. The loudness measurement unit 610 calculates the mean square of the input audio signal after passing through the frequency weighting filter in units of blocks (256 samples at a sampling frequency of 48 kHz) and outputs it as loudness.

  The gain calculation unit 620 calculates a gain from the loudness based on a loudness versus gain function set in advance to control the dynamic range of the input audio signal.

The gain smoothing time constant calculation unit 640 calculates a time constant when the gain is time smoothed, and outputs the time constant as a gain smoothing time constant. The gain smoothing time constant calculation unit 640 calculates a fluctuation value of the loudness by subtracting a smoothing loudness (a value obtained by temporally smoothing the past loudness) from the current loudness measured by the loudness measurement unit 610. Next, the gain smoothing time constant calculation unit 640 selects and outputs one of the following four types of time constants based on the fluctuation value of the loudness. Note that the value of the time constant increases from 1) to 4).
1) High-speed attack time constant (when the loudness variation is positive and greater than the attack threshold)
2) Low-speed attack time constant (when the loudness fluctuation value is positive and less than the attack threshold)
3) Fast release time constant (if the loudness variation is negative and less than the release threshold)
4) Low speed release time constant (if the loudness variation is negative and greater than the release threshold)
The gain smoothing unit 630 performs time smoothing on the gain from the gain calculation unit 620 with the gain smoothing time constant from the gain smoothing time constant calculation unit 640, and calculates and outputs the smoothed gain.

  Multiplier 600 calculates and outputs an output audio signal by multiplying the input audio signal by the smoothed gain.

  In the apparatus of FIG. 6, the dynamic range of the input audio signal can be appropriately suppressed, and it can be processed into an audio signal that is easy to hear even in a non-quiet environment and output.

  Further, Patent Document 1 describes a conventional automatic gain control device that uses AGC to output a signal having a constant input signal level.

Non-Patent Document 1 and Patent Document 1 describe single-band processing that performs DRC processing or AGC processing collectively for all frequency bands. Is divided into a plurality of frequency bands (bands), and multiband processing for performing DRC processing or AGC processing on each frequency band is described. Here, the terms single-band and multi-band used in audio compressors and the like are used, but in Non-Patent Document 2, instead of single-channel and multi-band used in hearing aids and the like. The word channel is used.
Japanese Patent No. 4014429 Charles Q. Robinson and 1 other, "Dynamic Range Control via Metadata", 107th AES convention, September 1999, preprint (Preprint) 5028 Harvey Dillon, translated by Masafumi Nakagawa, “Hearing Aid Handbook”, Ishigaku Shuppan Publishing Co., Ltd., October 2004, Chapter 6, p. 155-181

  However, in the conventional single-band audio signal processing apparatus described in Non-Patent Document 1 and Patent Document 1, breathing phenomenon (the level of noise or background sound varies according to the variation of the signal, and breathing is performed. There is a problem that there is a sense of temporal fluctuation of unnatural sound such as a phenomenon that can be felt in Also, if you use a slow release time constant to suppress the breathing phenomenon, the time fluctuation will be suppressed, but the time that the gain of the audio signal will be suppressed becomes longer, and the amount of information such as the sound of the hitting sound will decrease. I had the problem of doing. In the audio signal device described in Non-Patent Document 1, the problem relating to the breath-pass phenomenon is improved by adaptively switching a plurality of time constants. However, further improvements were expected.

  On the other hand, in the multi-band processing described in Non-Patent Document 2, by performing independent processing for each band, the breath-joining phenomenon hardly occurs, but the spectrum shape becomes flat and the frequency balance between the bands changes. , Had a problem of.

  The present invention solves the above-described conventional problems, and provides an audio signal processing apparatus and method that suppresses a time-dependent fluctuation of an unnatural sound such as a breathing phenomenon caused by a single band DRC or AGC for an audio signal. The purpose is to do.

  In order to solve this problem, an audio signal processing device according to the present invention multiplies an input audio signal by a gain and calculates an output audio signal, and a loudness for measuring the loudness of the input audio signal or the output audio signal. A measurement unit; a loudness smoothing unit that performs time smoothing on the measured loudness with a predetermined time constant; a gain calculation unit that calculates the gain based on the smoothed loudness; and the input audio signal or A frequency analysis unit that performs frequency analysis of an output audio signal; and a loudness smoothing time constant calculation unit that calculates a time constant of the loudness smoothing unit based on an analysis result of the frequency analysis unit. By performing time smoothing of the loudness with the time constant calculated based on the frequency analysis result, it is possible to reduce the unnatural sound temporal fluctuation feeling such as the breath tie phenomenon. It can be obtained.

  Further, the loudness smoothing time constant calculation unit, when the low frequency component is larger than a predetermined value, and / or when the ratio of the low frequency component to the total frequency component is larger than a predetermined value, The calculation is performed to increase the time constant of the loudness smoothing unit.

  The gain further includes a gain smoothing unit that performs time smoothing on the gain with a predetermined time constant and calculates a smoothed gain, and the multiplication unit adds the smoothed gain to the input audio signal. It is characterized by multiplication.

  Furthermore, a gain smoothing time constant calculation unit that calculates a time constant of the gain smoothing unit based on the frequency analysis result is further provided.

  Further, the input audio signal is multiplied by a smoothed gain to calculate an output audio signal, a loudness measuring unit for measuring the loudness of the input audio signal or the output audio signal, and based on the measured loudness A gain calculating unit that calculates gain, a time smoothing of the calculated gain with a predetermined time constant, and a smoothing gain to calculate the gain, and the input audio signal or the output audio signal A frequency analysis unit that performs frequency analysis; and a gain smoothing time constant calculation unit that calculates a time constant of the gain smoothing unit based on an analysis result of the frequency analysis unit. is there.

  Further, the gain smoothing time constant calculation unit, when the low frequency component is larger than a predetermined value, and / or when the ratio of the low frequency component to the total frequency component is larger than a predetermined value, The gain smoothing unit is calculated so as to increase the time constant.

  Further, the low frequency component is a component of 30 Hz to 150 Hz or less.

  According to the present invention, for a sound above the minimum audible limit, a low frequency sound has a large change in sound volume with respect to the same signal level change compared to a medium high frequency sound. When the low frequency component is large and / or when the ratio of the low frequency component to the total frequency component is large, the loudness and the time constant of a relatively large value are utilized. By performing time smoothing of the gain, it is possible to obtain an audio signal that can be easily heard while suppressing the unnatural sound of single band DRC or AGC over time. Since it is single band gain control, the frequency balance is not changed by gain control.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing a configuration of an audio signal processing device according to Embodiment 1 of the present invention. In FIG. 1, 100 is a superposed block dividing unit that divides an input audio signal into blocks to be superimposed, 110 is a multiplying unit that multiplies the output of the block dividing unit 100 and the output of the gain smoothing unit 160, and 120 is the output of 110. A superposition addition synthesis unit that synthesizes the output audio signal by superposition addition, 130 is a loudness measurement unit that measures the loudness of the input audio signal, and 140 is an output of the loudness smoothing time constant calculation unit 170 using the output of the loudness smoothing time constant calculation unit 170. A loudness smoothing unit 150 that smoothes the output time, 150 a gain calculation unit that calculates a gain from the output of the loudness smoothing unit 140, and 160 an output of the gain smoothing time constant calculation unit 180 that uses the output of the gain smoothing time constant calculation unit 180. A gain smoothing unit 170 that temporally smoothes the output, 170 is an output of the loudness smoothing unit 140 and an output of the frequency analysis unit 190 A loudness smoothing time constant calculation unit 180 for calculating the time constant of the loudness smoothing unit 140, and a gain 180 for calculating the time constant of the gain smoothing unit 160 from the output of the loudness smoothing unit 140 and the output of the frequency analysis unit 190. A smoothing time constant calculation unit 190 is a frequency analysis unit that performs frequency analysis of an input audio signal and outputs an analysis result.

  The audio signal processing apparatus in FIG. 1 performs DRC processing in units of single band blocks. The operation will be described below.

  The superposition block division unit 100 divides the input audio signal into blocks to be superposed and outputs them. That is, as shown in (Equation 1), by dividing the input speech signal x (n) by the window function w (n) that is superimposed by 50%, the divided speech signal y (n, t) is calculated. Here, n is a sample time number, t is a block number, N is a block length, and N / 2 is a block shift length. The value of N is set to, for example, 64 to 4096 when the sampling frequency Fs is 48 kHz.

  As the window function w (n), for example, a Hanning window shown in (Expression 2) is used.

  Multiplier 110 multiplies the divided audio signal from superimposed block divider 100 by the smoothed gain Gs (t) from gain smoother 160 and outputs a gain-controlled divided audio signal.

  The superposition addition synthesis unit 120 synthesizes and outputs the output audio signal z (n) by superimposing and adding the gain-controlled divided audio signal from the multiplication unit 110 as shown in (Equation 3).

  As described above, cross-fade superposition addition is performed, and divided audio signals in units of blocks can be smoothly connected.

  The loudness measuring unit 130 measures and outputs the loudness of the input audio signal in units of blocks. Various methods have been proposed as a loudness measurement method. In the first embodiment, a method using a weighting filter is used. That is, first, a signal v (n) obtained by passing a frequency weighting filter is created for the input audio signal x (n). Next, as shown in (Equation 4), the signal v (n) that has passed through the weighting filter is multiplied by the window function w (n) and divided into blocks. By calculating, the loudness L (t) of the block number t is measured. As the frequency weighting filter, for example, the recommendation BS.1 regarding standardization of ITU-R loudness measurement. A filter having the characteristics described in 1770 is used.

  The loudness smoothing unit 140 first receives the loudness L (t) of the block number t from the loudness measuring unit 130 and the smoothed loudness Ls (t−1) of the previous block number (t−1). When L (t)> Ls (t−1), the attack state is determined. Otherwise, the release state is determined. The result is used as attack / release information, and the loudness smoothing time constant calculation unit is determined. 170 and gain smoothing time constant calculation section 180. Next, the loudness smoothing unit 140 uses the loudness smoothing time constant Tl (t) calculated by the loudness smoothing time constant calculating unit 170 based on the attack / release information, and the loudness L (t (t) from the loudness measuring unit 130. ) Is calculated, and the smoothed loudness Ls (t) is calculated and output. The time smoothing of the loudness is performed according to (Equation 5). In (Equation 5), the value of A is calculated according to (Equation 6).

  As shown in (Equation 7), the gain calculating unit 150 uses the preset loudness versus gain DRC function F () to obtain the smoothed loudness Ls (t) from the loudness smoothing unit 140. Based on this, a gain G (t) for each block is calculated.

FIG. 2 is a schematic diagram showing an example of the DRC function of the loudness versus gain. In FIG. 2, the horizontal axis represents the loudness represented by the logarithmic axis, and the vertical axis represents the gain represented by the dB axis (20 log ₁₀ G (t)). The DRC function in FIG. 2 is classified into the following three areas.
1) When the loudness is equal to or less than the first threshold L1, amplification is performed with G (t)> 1.
2) When the loudness is larger than the first threshold and equal to or smaller than the second threshold L2, G (t) = 1 (that is, 0 dB) is not changed.
3) If loudness is greater than the second threshold, attenuate with G (t) <1.

  The DRC function shown in FIG. 2 amplifies a small sound and attenuates a loud sound. Therefore, the dynamic range can be suppressed and an audio signal that can be easily heard even in a quiet environment can be obtained.

  The gain smoothing unit 160 performs time smoothing of the gain G (t) from the gain calculation unit 150 with the gain smoothing time constant Tg (t) calculated by the gain smoothing time constant calculation unit 180, and smoothes the smoothing. The calculated gain Gs (t) is calculated and output. The time smoothing of the gain is performed according to (Equation 8). In (Equation 8), the value of B is calculated according to (Equation 9).

  The frequency analysis unit 190 performs frequency analysis of the input audio signal in units of blocks and outputs a frequency analysis result. First, the frequency analysis unit 190 performs a short-time discrete Fourier transform of (Equation 10) on the input speech signal x (n) to calculate a Fourier transform coefficient X (k, t). Here, w (n) is the window function, and k represents a Fourier transform coefficient number. Note that the short-time discrete Fourier transform can be efficiently performed using a fast Fourier transform.

  Next, according to (Equation 11), the Fourier transform coefficients are grouped for each frequency band, and the signal level P (m, t) of each frequency band is calculated. Here, m represents a frequency band number, and M represents the number of frequency bands.

  In (Equation 11), k1 (m) and k2 (m) represent the start Fourier transform coefficient number and the end Fourier transform coefficient number of the frequency band m, respectively. In the first embodiment, M = 2, and the low frequency component P (0, t) of 30 Hz to 150 Hz or less is grouped into the other middle / high frequency components P (1, t), and the analysis result is obtained. Output. When calculating the signal level P (m, t) of each frequency band, it may be calculated by multiplying the weighting coefficient c (k) as shown in (Equation 12) instead of (Equation 11).

  Alternatively, the signal level of each frequency band may be calculated using a band division filter bank instead of the short-time discrete Fourier transform.

  The loudness smoothing time constant calculation unit 170 is based on the attack / release information from the loudness smoothing unit 140 and the signal level of each frequency band from the frequency analysis unit 190, and the loudness smoothing used in the loudness smoothing unit 140. Calculate and output the time constant. The loudness smoothing time constant calculation unit 170 calculates time constants having different values for the attack state and the release state based on the attack / release information. In the attack state, it is usually calculated so as to have a small (fast) time constant as compared to the release state.

  In the first embodiment, when the low-frequency component P (0, t) is equal to or smaller than a predetermined value, a small (fast) time constant is calculated. Otherwise, all frequency components ( When the ratio of the low frequency component P (0, t) to P (0, t) + P (1, t)) is larger than a predetermined value, a large time constant is calculated. FIG. 3 is a schematic diagram showing an example of a function of a time constant with respect to the ratio of the low frequency components occupying the total frequency components. When the ratio is larger than a predetermined value R1, a large value is set according to the ratio. Calculate the time constant.

  FIG. 4 shows the standard of the loudness of a pure tone for an otologically normal person under free sound field listening conditions described in ISO 226: 2003 “Acoustics-Normal sound loudness sense curve” It is a feeling curve. The points on each curve all sound the same regardless of the frequency of the pure tone. The sound pressure level of each frequency for sounding the same magnitude is shown on the vertical axis. As can be seen from FIG. 4, for sounds above the minimum audible limit (Hearing threshold), the sound of the low frequency is louder than the sound of the middle and high frequencies with respect to the same signal level change. The change is large. That is, when the sound of the low frequency starts to be heard compared to the sound of the middle and high frequencies, the change in the sound level due to the change of the signal level is large.

  Therefore, the loudness smoothing time constant calculation unit 170 uses the knowledge of the psychoacoustics when the low frequency component is larger than a predetermined value and the ratio of the low frequency component to the total frequency component is predetermined. When the value is larger than the value, a change in the volume of the sound can be suppressed by calculating so as to increase the smoothing time constant.

  The gain smoothing time constant calculation unit 180 uses the gain smoothing unit 160 based on the attack / release information from the loudness smoothing unit 140 and the signal level of each frequency band from the frequency analysis unit 190. Calculate and output the time constant. The operation of the gain smoothing time constant calculation unit 180 is the same as the operation of the loudness smoothing time constant calculation unit 170, and a description thereof will be omitted.

  Note that the time constants calculated by the loudness smoothing time constant calculation unit 170 and the gain smoothing time constant calculation unit 180 may be the same value. In this case, either the loudness smoothing time constant calculation unit 170 and the gain smoothing time constant calculation unit 180 can be used together, and either one is sufficient.

  As described above, in the audio signal processing apparatus according to Embodiment 1, the frequency analysis unit 190 that performs frequency analysis of an input audio signal, and the time constants of loudness smoothing and gain smoothing based on the analysis result of the frequency analysis unit 190. Psychoacoustics that the loudness smoothing time constant calculating unit 170 and the gain smoothing time constant calculating unit 180 that respectively calculate the sound level have a large change in the sound level with respect to the change in the signal level at the low frequency. In the case where the low frequency component is larger than the predetermined value and the ratio of the low frequency component to the total frequency component is larger than the predetermined value, the time constant is increased. Since the loudness smoothing and the gain smoothing are performed, it is possible to obtain an audio signal that is easy to hear and suppresses a sense of sound fluctuation.

  In the audio signal processing apparatus according to the first embodiment, the loudness smoothing unit 140 and the loudness smoothing time constant calculating unit 170 for calculating the time constant thereof, and the gain smoothing unit 160 and the gain smoothing time constant for calculating the time constant thereof. Although the calculation unit 180 is included, the configuration may include only the former or the latter.

  Further, in the loudness smoothing time constant calculation unit 170 and the gain smoothing time constant calculation unit 180 according to the first embodiment, when the low frequency component is larger than a predetermined value, and the low frequency component that occupies all the frequency components When the ratio is larger than a predetermined value, the time constant is calculated to be large. However, when either one of the two cases is satisfied, the time constant is calculated to be large. Good.

  In addition, the loudness smoothing time constant calculation unit 170 and the gain smoothing time constant calculation unit 180 according to the first embodiment calculate the time constant based on the low frequency component. However, the same idea may be applied to calculate the time constant.

(Embodiment 2)
FIG. 5 is a block diagram showing the configuration of the audio signal processing apparatus according to Embodiment 2 of the present invention. The audio signal processing apparatus according to Embodiment 1 in FIG. 1 has a feed-forward type block unit DRC configuration, whereas the audio signal processing apparatus according to Embodiment 2 in FIG. 5 has a feedback type. This is a configuration of AGC in units of samples.

  In FIG. 5, 500 is a multiplication unit that multiplies the input audio signal and the output of the gain smoothing unit 540, 510 is a loudness measurement unit that measures the loudness of the output audio signal, and 520 is an output of the loudness smoothing time constant calculation unit 550. Is used to smooth the output of the loudness measuring unit 510, 530 is a gain calculating unit that calculates a gain from the output of the loudness smoothing unit 520, and 540 is the output of the gain smoothing time constant calculating unit 560. A gain smoothing unit 550 that smoothes the output of the gain calculation unit 530 using a loudness smoothing that calculates a time constant of the loudness smoothing unit 520 from the output of the loudness smoothing unit 520 and the output of the frequency analysis unit 570. The time constant calculation unit 560 calculates the gain smoothing unit 5 from the output of the loudness smoothing unit 520 and the output of the frequency analysis unit 570. Gain smoothing time constant calculation unit for calculating a time constant of 0, 570 performs a frequency analysis of the output audio signal, a frequency analysis section for outputting the analysis result. The audio signal processing apparatus in FIG. 5 performs AGC processing in units of single band samples. The operation will be described below.

  As shown in (Equation 13), the multiplication unit 500 performs the gain Gs (n) of the input audio signal x (n) (where n is a sample time number) smoothed from the gain smoothing unit 540. To calculate the gain-controlled output audio signal z (n).

  The loudness measuring unit 510 measures and outputs the loudness of the output audio signal from the multiplying unit 500. As the loudness measurement method, the method using the frequency weighting filter is used in the first embodiment, but in the second embodiment, the frequency weighting filter is not used (with a flat frequency characteristic), as shown in (Expression 14). Then, the loudness L (n) of the sample time number n is measured by calculating the root mean square of the output audio signals of the past N samples. The value of N is set to 64 to 4096 when the sampling frequency Fs is 48 kHz, for example.

  First, the loudness smoothing unit 520 first receives the loudness L (n) of the sample time number n from the loudness measuring unit 510 and the smoothed loudness Ls (n−1) of the previous sample time number (n−1). ), If L (n)> Ls (n−1), the attack state is determined, otherwise the release state is determined, and the result is used as attack / release information, and the loudness smoothing time constant is determined. It outputs to the calculation part 550 and the gain smoothing time constant calculation part 560. Next, the loudness smoothing unit 520 uses the loudness smoothing time constant Tl (n) calculated by the loudness smoothing time constant calculating unit 550 based on the attack / release information, and the loudness L (n ) Is smoothed, and the smoothed loudness Ls (n) is calculated and output. The loudness temporal smoothing is performed according to (Equation 15). In (Equation 15), the value of A is calculated according to (Equation 16).

  The gain calculation unit 530 calculates the gain G (n) in units of samples from the smoothed loudness Ls (n) from the loudness smoothing unit 520 and the preset target loudness LT as shown in (Equation 17). To do. When the gain value is larger than a preset upper limit value, the gain value is limited to the upper limit value. When the gain value is smaller than a preset lower limit value, the gain value is limited to the lower limit value.

  The gain smoothing unit 540 performs time smoothing of the gain G (n) from the gain calculation unit 530 with the gain smoothing time constant Tg (n) from the gain smoothing time constant calculation unit 560, and the smoothed gain Gs (n) is calculated and output. The time smoothing of the gain is performed according to (Equation 18). In (Equation 18), the value of B is calculated according to (Equation 19).

  First, the frequency analysis unit 570 receives the output audio signal z (n) as an input, and a signal Z (m, n) divided into M frequency bands by a band division filter bank (where m is a frequency band number). ) Is calculated. Next, according to (Equation 20), the sum of squares of the past N samples of each frequency band is calculated and output as a signal level P (m, n) of each frequency band.

  In the frequency analysis unit 570 of the second embodiment, as in the first embodiment, M = 2, and the low frequency component P (0, n) of 30 Hz to 150 Hz or less and the other middle / high frequency components P ( 1, n) is calculated and output as an analysis result.

  The loudness smoothing time constant calculation unit 550 determines the time constant used by the loudness smoothing unit 520 based on the attack / release information from the loudness smoothing unit 520 and the signal level of each frequency band from the frequency analysis unit 570. Calculate and output. In the loudness smoothing time constant calculation unit 550, as in the case of the loudness smoothing time constant calculation unit 170 of the first embodiment, when the low frequency component is larger than a predetermined value, the low frequency component that occupies all the frequencies When the ratio is larger than a predetermined value, the time constant is calculated to be increased.

  Similarly, the gain smoothing time constant calculation unit 560 is used by the gain smoothing unit 540 based on the attack / release information from the loudness smoothing unit 520 and the signal level of each frequency band from the frequency analysis unit 570. Calculate and output the time constant. The operation of the gain smoothing time constant calculation unit 560 is the same as the operation of the loudness smoothing time constant calculation unit 550, and a description thereof will be omitted.

  Note that the time constants calculated by the loudness smoothing time constant calculation unit 550 and the gain smoothing time constant calculation unit 560 may be the same value. In this case, the loudness smoothing time constant calculation unit 550 and the gain smoothing time constant calculation unit 560 can be used together, and either one of them may be used.

  As described above, in the audio signal processing device according to the second embodiment, the frequency analysis unit 570 that performs frequency analysis of the output audio signal, and the time constants of loudness smoothing and gain smoothing based on the analysis result of the frequency analysis unit 570 Psychoacoustics that a loudness smoothing time constant calculating unit 550 and a gain smoothing time constant calculating unit 560 that respectively calculate the sound level have a large change in sound volume with respect to a change in signal level at a low frequency. When the low frequency component is larger than a predetermined value and when the ratio of the low frequency component to the total frequency component is larger than the predetermined value, the time constant is increased and the time constant is increased. Since loudness smoothing and gain smoothing are performed, it is possible to obtain an audio signal that is easy to hear and suppresses a sense of sound fluctuation.

  In the audio signal processing apparatus according to the second embodiment, the loudness smoothing unit 520 and the loudness smoothing time constant calculating unit 550 for calculating the time constant thereof, and the gain smoothing unit 540 and the gain smoothing time constant for calculating the time constant thereof. Although the calculation unit 560 is included, the configuration may include only the former or the latter.

  Further, in the loudness smoothing time constant calculation unit 550 and the gain smoothing time constant calculation unit 560 of the second embodiment, the low frequency component that occupies all frequency components when the low frequency component is larger than a predetermined value. When the ratio is larger than a predetermined value, the time constant is calculated to be increased. However, when either one of the two cases is satisfied, the time constant may be increased.

  Further, in the loudness smoothing time constant calculation unit 550 and the gain smoothing time constant calculation unit 560 of the second embodiment, the time constant is calculated based on the low frequency component. A similar concept may be applied to calculate the time constant.

  In the first embodiment, processing is performed in units of blocks, but processing may be performed in units of samples as in the second embodiment.

  In the second embodiment, the processing is performed in units of samples. However, the processing may be performed in units of blocks as in the first embodiment.

  The audio signal processing apparatus according to the present invention may be configured by a computer that operates according to a program for causing a computer to execute processing in each block.

  As described above, the audio signal processing apparatus according to the present invention can suppress the sense of temporal variation of unnatural sound, control the dynamic range of the audio signal, and make an audio signal easy to hear. It is useful as an audio signal processing apparatus such as a television, radio, DVD, video camera, minicomputer, mobile phone, and the like.

The block diagram which shows the structure of the audio | voice signal processing apparatus in Embodiment 1 of this invention. Schematic diagram showing an example of a DRC function of loudness vs. gain used in gain calculation section 150 of Embodiment 1 of the present invention Schematic diagram showing an example of a function of a time constant with respect to a ratio of a low frequency component occupying all frequency components in Embodiments 1 and 2 of the present invention ISO 226: 2003 “Acoustic—Normal reference loudness isometric curve” shows a standard isometric curve of pure tone loudness for an otologically normal person under free-field audition conditions. Graph The block diagram which shows the structure of the audio | voice signal processing apparatus in Embodiment 2 of this invention. The block diagram which shows the structure of the conventional audio | voice signal processing apparatus described in the nonpatent literature 1.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Superimposition block division part 110,500 Multiplication part 120 Superposition addition composition part 130,510 Loudness measurement part 140,520 Loudness smoothing part 150,530 Gain calculation part 160,540 Gain smoothing part 170,550 Loudness smoothing time constant calculation Unit 180,560 Gain smoothing time constant calculation unit 190,570 Frequency analysis unit

Claims (10)

A multiplier for multiplying an input audio signal by a gain and calculating an output audio signal;
A loudness measuring unit for measuring the loudness of the input audio signal or the output audio signal;
A loudness smoothing unit that performs time smoothing on the measured loudness with a predetermined time constant;
A gain calculation unit for calculating the gain based on the smoothed loudness;
A frequency analysis unit for performing frequency analysis of the input audio signal or the output audio signal;
An audio signal processing apparatus comprising: a loudness smoothing time constant calculating unit that calculates a time constant of the loudness smoothing unit based on an analysis result of the frequency analyzing unit. The loudness smoothing time constant calculation unit calculates the loudness when the low-frequency component is larger than a predetermined value and / or when the ratio of the low-frequency component to the total frequency component is larger than a predetermined value. 2. The audio signal processing apparatus according to claim 1, wherein the calculation is performed so as to increase a time constant of the smoothing unit. Furthermore, the gain is smoothed with a predetermined time constant, and a gain smoothing unit that calculates the smoothed gain is provided.
The audio signal processing apparatus according to claim 1, wherein the multiplication unit multiplies the input audio signal by the smoothed gain. 4. The audio signal processing apparatus according to claim 3, further comprising a gain smoothing time constant calculating unit that calculates a time constant of the gain smoothing unit based on the frequency analysis result. A multiplier for multiplying the input audio signal by the smoothed gain and calculating an output audio signal;
A loudness measuring unit for measuring the loudness of the input audio signal or the output audio signal;
A gain calculation unit for calculating a gain based on the measured loudness;
A gain smoothing unit that performs time smoothing on the calculated gain with a predetermined time constant and calculates a smoothed gain;
A frequency analysis unit for performing frequency analysis of the input audio signal or the output audio signal;
An audio signal processing apparatus comprising: a gain smoothing time constant calculating unit that calculates a time constant of the gain smoothing unit based on an analysis result of the frequency analyzing unit. When the low frequency component is larger than a predetermined value and / or when the ratio of the low frequency component to the total frequency component is larger than a predetermined value, the gain smoothing time constant calculation unit calculates the gain. 6. The audio signal processing apparatus according to claim 5, wherein the calculation is performed so as to increase a time constant of the smoothing unit. The audio signal processing apparatus according to claim 2, wherein the low frequency component is a component of 30 Hz to 150 Hz or less. A multiplication step of multiplying an input audio signal by a gain and calculating an output audio signal;
A loudness measuring step for measuring a loudness of the input audio signal or the output audio signal;
A loudness smoothing step for performing time smoothing on the measured loudness with a predetermined time constant;
A gain calculating step of calculating the gain based on the smoothed loudness;
A frequency analysis step for performing frequency analysis of the input audio signal or the output audio signal;
An audio signal processing method comprising: a loudness smoothing time constant calculating step of calculating a time constant of the loudness smoothing based on an analysis result of the frequency analysis. A multiplication step of multiplying the input audio signal by a smoothed gain and calculating an output audio signal;
A loudness measuring step for measuring a loudness of the input audio signal or the output audio signal;
A gain calculating step of calculating a gain based on the measured loudness;
A gain smoothing step for performing time smoothing on the calculated gain with a predetermined time constant and calculating a smoothed gain;
A frequency analysis step for performing frequency analysis of the input audio signal or the output audio signal;
And a gain smoothing time constant calculating step of calculating a time constant of the gain smoothing based on an analysis result of the frequency analysis. The program for making a computer perform each step in the audio | voice signal processing method of Claim 8 or Claim 9.

Images