JP2012195772A - Audio signal processing device, control method thereof, and computer program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sound processing technique capable of reducing even noise produced at an initial stage of wind noise occurrence.SOLUTION: An audio signal processing device comprises: sound collection means; storage means for storing the audio signals collected by the sound collection means; noise detection means for detecting noise included in the audio signals stored in the storage means; noise reduction means for reducing the noise detected by the noise detection means from the audio signals; and setting means for setting a time constant which controls noise reduction characteristics of the noise reduction means according to a level of the noise detected by the noise detection means. The setting means starts setting of the time constant in the noise reduction means in response to detection of the noise detected by the noise detection means. The noise reduction means reads the audio signal from the storage means and reduces the noise when the setting of the time constant by the setting means according to the noise level is completed.

Description

  The present invention relates to an audio signal processing device, a control method thereof, and a computer program.

  As an audio signal processing device, there is a recording device that collects and records audio with a microphone. The recording function for recording audio is not only used for recording devices such as recording devices, but also for devices capable of generating movies by simultaneously acquiring audio and video, such as imaging devices. It is installed.

  Devices equipped with such a recording function have been equipped with a function of reducing wind noise (wind noise) superimposed on an audio signal. As the reduction technique, Patent Document 1 discloses a technique for attenuating wind noise superimposed on sound collected by a microphone with a high-pass filter, and a technique for changing the time constant of the high-pass filter based on the magnitude of noise. Yes. In Patent Document 2, a thermocouple is used to detect wind, the level of wind noise is estimated according to the output of the thermocouple, and a high-pass filter used for noise reduction is selected.

JP-A-5-176221 JP-A-5-328480

  However, the above-described proposed technique requires time until wind noise is detected, and noise at the initial stage of wind noise generation cannot be reduced with high accuracy.

  Therefore, the present invention provides an audio processing technique that can reduce noise from the beginning of wind noise generation.

In order to achieve the above object, an audio signal processing device according to the present invention provides:
Voice collection means;
Storage means for storing the voice signal collected by the voice sampling means;
Noise detection means for detecting noise contained in the audio signal stored in the storage means;
Noise reduction means for reducing noise detected by the noise detection means from the voice signal;
A setting means for setting a time constant for controlling a noise reduction characteristic of the noise reduction means according to a level of the noise detected by the noise detection means,
The setting means starts setting the time constant to the noise reduction means in response to detection of the noise by the noise detection means,
The noise reduction unit reads the audio signal from the storage unit and performs the noise reduction when the setting of the time constant according to the noise level by the setting unit is completed.

  According to the present invention, wind noise can be reduced with high accuracy from the initial generation of wind noise.

The block diagram of the audio | voice signal processing apparatus corresponding to Embodiment 1 of invention. The figure explaining the characteristic of the noise reduction part corresponding to embodiment of invention. The figure explaining the other characteristic of the noise reduction part corresponding to embodiment of invention. The figure which shows typically the structure of the noise reduction part corresponding to embodiment of invention with an analog circuit. The timing chart of the noise reduction process corresponding to embodiment of invention. The figure explaining the effect of the noise reduction process corresponding to embodiment of invention. The flowchart of the noise detection process corresponding to embodiment of invention. The flowchart of the noise reduction process corresponding to embodiment of invention. The block diagram of the audio | voice signal processing apparatus corresponding to Embodiment 2 of invention. The flowchart of the noise reduction process corresponding to Embodiment 2 of invention. The block diagram which shows another aspect of the audio | voice signal processing apparatus corresponding to Embodiment 2 of invention.

  Embodiments of the invention will be described below with reference to the accompanying drawings.

[Embodiment 1]
FIG. 1 is a block diagram showing a configuration example of an audio signal processing apparatus corresponding to Embodiment 1 of the invention, and shows an example of a recording apparatus having a stereo microphone. Note that the audio signal processing device may be an imaging device including a main configuration of the recording device. In that case, the imaging device may be any device that can record audio and video in synchronization, and includes, for example, a digital still camera, a video camera, a mobile phone, a laptop computer, and the like having a video shooting function. It is not limited to.

  The microphones 11R and 11L are audio detection units that detect audio and are paired as stereo microphones. Analog audio signals from the microphones 11R and 11L are output to the gain changing unit 12. The gain changing unit 12 increases the amplification factor of the outputs of the microphones 11R and 11L to improve the S / N when the subject sound is small. On the other hand, when the subject sound is loud, the amplification factor is reduced to prevent saturation in the amplification circuit. In the case where it is desired to digitally process the audio signal, the gain changing unit 12 can include an analog / digital converter (ADC), and converts the analog audio signal from the microphone into a digital audio signal. Further, when converting into a digital signal, the gain is controlled so that a desired gain can be obtained.

  The storage unit 13 is a buffer that temporarily stores the audio signal input from the gain changing unit 12. The output from the storage unit 13 is input to the noise reduction unit 14. The noise reduction unit 14 includes a plurality of high-pass filter stages (14a, 14b) having the characteristics shown in FIGS. 2a and 2b. The number of filter stages is two in this embodiment, but may be three or more. In FIG. 1, although two microphones are provided on the left and right, there is only one output line for subsequent signal processing, but in practice, noise reduction signal processing is performed separately for each of the left and right microphones. Yes.

  FIG. 2a shows the frequency characteristics of the output sound from the microphone of the sound input to the microphone. The horizontal axis indicates the frequency, and the vertical axis indicates the gain at the time of output of the input audio signal. In FIG. 2a, it can be seen that the audio signal having a frequency lower than 10 Hz is attenuated (low gain). Here, human voice is also included, and the subject sound source has almost no low frequency of 10 Hz or less, so that the subject sound is not deteriorated by the noise reduction unit having the characteristics as shown in FIG. FIG. 2b also shows the frequency characteristics of the output sound of the microphone, but here shows a case where an audio signal having a frequency lower than 200 Hz is attenuated. The noise generated by the wind, such as wind noise, has many frequency components of 200 Hz or less. Therefore, the wind noise included in the audio signal can be reduced by using the characteristics shown in FIG. 2b.

  Hereinafter, a method of switching from the characteristic of FIG. 2a to the characteristic of FIG. 2b will be described. First, the case where the noise reduction unit is formed of an analog circuit as shown in FIG. 3 will be described.

  FIG. 3 shows a known analog high-pass filter, which includes a capacitor 222, a variable resistor 223, and a buffer amplifier 221. The control signal 224 is a control signal for controlling the resistance value of the variable resistor 223 and is given from, for example, the noise reduction control unit 19. 10 Hz in FIG. 2 a and 200 Hz in FIG. 2 b can be set by the impedance product of the capacitor 222 and the variable resistor 223. For example, the resistance value of the variable resistor 223 can be increased to obtain the characteristics shown in FIG. 2A, and the resistance value of the variable resistor 223 can be reduced by the control signal 224 to obtain the characteristics shown in FIG. When the high-pass filter is formed by a digital filter such as a known difference equation, the characteristics shown in FIGS. 2a and 2b can be realized by adjusting the coefficient multiplied by the input value and the previous output value. Further, by changing the coefficient during the operation of the high-pass filter, the characteristic shown in FIG. 2a can be changed to the characteristic shown in FIG. 2b.

  Returning to FIG. 1, the output from the first noise reduction unit 14a is recorded in the audio recording unit 17 via the second noise reduction unit 14b or via the order changing unit 15 and the gain restoration unit 16 described later. The The audio recording unit 17 is a recording medium that records an audio signal on a recording medium, and includes a solid memory, a DVD, a hard disk, a tape, and the like, but is not limited to a specific type of recording medium.

  The noise detection unit 18 detects wind noise from the audio signal input from the storage unit 13. In general, subject sound input to a stereo microphone has a strong correlation between left and right channels, whereas wind noise is often uncorrelated. Therefore, the noise detector 18 first compares the audio signals obtained by the microphones 11R and 11L. If there is no correlation between the two, it can be determined that a wind noise is generated, and the level of the wind noise can be determined based on the signal amplitude having no correlation. Specifically, if the amplitude is large, it can be determined that the wind noise is large. The correlation determination may be performed by paying attention to a part of the band. For example, the noise detection unit 18 may determine whether or not the wind noise is loud by determining the degree of correlation in a frequency band of 400 Hz or less.

  The noise reduction control unit 19 changes the characteristics of the noise reduction unit 14 in response to the timing and magnitude of the noise signal detected by the noise detection unit 18. This change is realized by changing the time constant that controls the noise reduction characteristics of the high-pass filter that constitutes the noise reduction unit 14. For example, the state shown in FIG. 2a is changed to the state shown in FIG. 2b.

  In the present embodiment, the filter characteristic shown in FIG. 2A is referred to as “high-pass filter characteristic having a large time constant”. On the other hand, the filter characteristic shown in FIG. 2B is referred to as “a high-pass filter characteristic with a small time constant”. The time constant can be determined according to the level of noise detected by the noise detection unit 18 (the magnitude of the amplitude of the uncorrelated waveform of the microphones 11R and 11L). For example, in the case of a wind noise whose noise level is less than a certain value, a high-pass filter characteristic that attenuates 100 Hz or less that does not reach FIG. 2b is used. On the other hand, in the case of a loud wind noise whose noise level is a certain value or higher, a high-pass filter characteristic that attenuates 200 Hz or less shown in FIG. In the case of a larger wind noise, a high-pass filter characteristic that attenuates 400 Hz or less may be used. In this way, the cut-off frequency of the high-pass filter can be changed according to the wind noise level.

  The noise reduction control unit 19 also controls the second noise reduction unit 14b and changes its time constant. The second noise reduction unit 14b can be a high-pass filter having the same configuration as the first noise reduction unit 14a. The reason why the first noise reduction unit 14a and the second noise reduction unit 14b are provided in the present embodiment is as follows. That is, the first noise reduction unit 14a changes its characteristics according to the magnitude of the wind noise. When the wind noise is loud, the high-pass filter is changed to a characteristic that attenuates 400 Hz or less. However, in the case of wind noise that is larger than that, the time constant is not further reduced, and for example, a high-pass filter characteristic that attenuates 600 Hz or less is not obtained. This is because if the frequency to be attenuated is increased, the sound signal of the subject is affected and deteriorates.

  Therefore, in the present embodiment, the change is made not by changing the high-pass filter time constant but by changing the order. That is, by directly connecting two high-pass filters, the low-frequency attenuation performance is improved without changing the frequency at which attenuation starts. The noise reduction control unit 19 outputs a switching signal to the order changing unit 15 and switches the internal switch of the order changing unit 15 to reconfigure the noise reducing unit 14. The order changing unit 15 switches the audio signal that has passed through only the first noise reducing unit 14a and the audio signal that has passed through the first noise reducing unit 14a and the second noise reducing unit 14b by the order changing unit 15 to the gain restoring unit 16. Output. In addition, the arrangement | positioning location of the order change part 15 does not need to be the output end of the 1st noise reduction part 14a and the 2nd noise reduction part 14b. For example, it may be connected to the output terminal of the first noise reduction unit 14a to switch between bypassing the second noise reduction unit 14b or passing the second noise reduction unit 14b. Moreover, you may make it switch whether the 1st noise reduction part 14a is bypassed instead of the 2nd noise reduction part 14b. In this case, the second noise reduction unit 14b is always used, and the first noise reduction unit 14a is used according to the detected noise level.

  The noise reduction control unit 19 switches the order change unit 15 and passes through the first noise reduction unit 14a and the second noise reduction unit 14b when the wind noise level becomes extremely high and cannot be dealt with by the noise reduction unit 14a alone. The obtained audio signal is input to the gain restoring unit 16. The gain restoring unit 16 plays a role of returning the gain variation changed by the gain changing unit 12. For example, normally, when the gain changing unit 12 gives a gain of 20 times, the gain of the gain restoring unit 16 is 1, and when the gain changing unit 12 is changed and the gain is given twice, the gain restoring unit The gain of 16 is 10. As a result, the audio signal output from the gain restoring unit 16 is commonly given a gain of 20 times.

  The reason why the gain changing unit 12 changes the gain of the audio signal in the range of 20 to 2 times is to prevent the wind noise level from becoming high and saturating the audio signal. For example, when the level of wind noise mixed in the audio signal input to the microphone 11 is large, the gain in the gain changing unit 12 is doubled, and when the level is small, the gain is 20 times. Therefore, the detection result of the noise level detected by the noise detection unit 18 is fed back to the gain changing unit 12 to adjust the gain. Therefore, when a gain of 10 times is given by the gain restoration unit 16, saturation occurs again. However, since the wind noise level is actually reduced by the first noise reduction unit 14a and the second noise reduction unit 14b, saturation is achieved. Will not occur.

  The overall control unit 20 controls the overall operation of the audio signal processing device. In the present embodiment, the main operation of each functional block for executing the noise reduction process will be described. The operation is executed under the control of the overall control unit 20.

  FIG. 4 is a timing chart of the block of FIG. 1, where the horizontal axis represents time and the vertical axis represents the operation of each element. First, the noise detection unit 18 reads out an audio signal temporarily stored in the storage unit 13 and performs noise detection processing. A waveform 41 is a detection result of the noise detection unit 18 and detects wind noise (wind pressure level) based on the correlation state of the pair of microphones 11R and 11L and the sound pressure level at that time. In the waveform 41, wind noise is generated at time t1.

  The noise detector 18 determines the generation of wind noise and the level thereof based on the amplitude value of the waveform 41. The level determination may be performed using a threshold for level determination set in the noise detection unit 18. When detecting the occurrence of wind noise, the noise detection unit 18 notifies the storage unit 13 of the occurrence of wind noise and the detected level, and stores it as a voice processing instruction in synchronization with the audio signal. The noise reduction control unit 19 determines the time constant and the order based on the noise level included in the audio processing instruction recorded in the audio signal, and the first noise reduction unit 14a and the second noise reduction unit 14b, and the order change unit 15 is controlled.

  In FIG. 4, the setting of the time constant 1 of the first noise reduction unit 14a is started at time t1. Therefore, after the wind noise is detected at time t1, the noise reduction processing is actually started at time t2, as shown by the waveform 44. This waveform 44 shows the timing at which the detected noise is processed. Between time t1 and t2, as shown by the waveform 42, the time constant 1 of the first noise reduction unit 14a changes in the direction of decreasing. Depending on the level of wind noise, for example, it can be changed to correspond to the time constant of FIGS. 2a to 2b. When the setting of the time constant 1 is completed at time t2, the first noise reduction unit 14a reads an audio signal from the storage unit 13. As described above, in the present embodiment, since the audio signal is temporarily stored in the storage unit 13, the audio signal can be delayed for the audio processing system, and the audio signal can be read out at the timing of processing. it can.

  In the example shown in FIG. 4, the noise reduction control unit 19 determines that there is no need to change the high-pass filter characteristics up to the characteristics shown in FIG. 2b at the detection level of the waveform 41 at time t1. In that case, the noise reduction control unit 19 ends the change of the time constant 1 when the time constant 1 is reduced to a filter characteristic that attenuates, for example, 100 Hz or less. As a method for determining the level, the wind noise level is divided into a plurality of stages according to a threshold value, one time constant is associated with each stage, and the time constant is determined according to the stage to which the detected wind noise level belongs. be able to. Alternatively, a time constant corresponding to the level of each detected wind noise may be determined based on a linear function that receives the amplitude value of the wind noise.

  The change of the time constant is set to end just at time t2, and when the process of reducing noise from the audio signal is started at time t2, the high-pass filter is already in a standby state with desired characteristics. The reason for changing the time constant gradually over time from time t1 to time t2 is as follows.

  First, care must be taken when switching the time constant of the high-pass filter because a change in the output phase occurs. For example, in the first-order high-pass filter that attenuates 200 Hz or less shown in FIG. 2B, the signal phase of 200 Hz changes by 45 degrees compared to the original signal. In contrast, a first-order high-pass filter that attenuates 10 Hz or less as shown in FIG. Therefore, when the characteristics are instantaneously switched from FIG. 2a to FIG. 2b, the continuity of the 200 Hz signal is lost, and unnecessary noise is generated at that timing. In order to avoid this, a predetermined time (for example, 100 ms) is spent, and the time constant is gradually changed from “large” to “small”. The predetermined time may be a fixed value when the time constant is changed. In that case, for example, when the time constant is continuously changed with the maximum change width (for example, 10 Hz to 200 Hz), it is possible to set the time so that the continuity of the signal is not impaired.

  Next, at time t3, as indicated by the waveform 41, the noise detector 18 detects the generation of a larger wind noise. The noise reduction control unit 19 sets the time constant 1 of the high-pass filter of the first noise reduction unit 14a to be even smaller, assuming that the characteristics of the high-pass filter set at time t2 are insufficient for wind noise reduction. Here, the characteristic of the high-pass filter is changed from time t3 to time t4 to the characteristic of attenuating 200 Hz or less shown in FIG. 2b. The time t4-t3 also corresponds to the predetermined time.

  Next, at time t5, as indicated by the waveform 41, the noise detector 18 detects the generation of a larger wind noise. In this case, the noise reduction control unit 19 may further reduce the noise by further reducing the time constant 1 of the high-pass filter that is the noise reduction unit 14a to increase the attenuation in the low frequency region of the voice. For example, the characteristic is changed to a characteristic that attenuates 400 Hz or less. However, if the frequency at which the reduction is started is changed to the high frequency side (for example, 200 Hz is changed to 400 Hz), even the sound of the subject is attenuated, resulting in an uncomfortable sound recording.

  Therefore, in the present embodiment, when the noise level to be reduced becomes higher than a predetermined level, the high-pass filter time constant 1 that is the first noise reduction unit 14a is not reduced, but a high-pass filter is further added in series. Correspond with. That is, the output of the high-pass filter that is the first noise reduction unit 14a is input to the high-pass filter that is the second noise reduction unit 14b. This reduces the noise level by increasing the order, not the time constant of the high-pass filter.

  In this case, since the frequency at which the reduction is started is not changed to the high frequency side (for example, 200 Hz is changed to 400 Hz), the influence on the subject sound is reduced. It is also conceivable to take measures to increase the order even when the noise level is low. However, increasing the order also has a considerable influence on the subject sound, so that when the noise level is small, that is, only the high-pass filter that is the noise reduction unit 14a is used, so that high-quality voice recording can be performed.

  The order of the noise reduction unit 14 is switched by switching the order changing unit 15 as shown in FIG. That is, the order changing unit 15 outputs the output of the high-pass filter that is the first noise reduction unit 14a, and the serial connection output signal of the high-pass filter that is the first noise reduction unit 14a and the high-pass filter that is the second noise reduction unit 14b. Both are entered. The order changing unit 15 switches paths in accordance with the order switching signal from the noise reduction control unit 19 as shown by the waveform 45. In the example of FIG. 4, at the time t5, the path of the first noise reduction unit 14a alone is switched to the path of the first noise reduction unit 14a and the second noise reduction unit 14b.

  The time constant 2 of the second noise reduction unit 14b is changed according to the switching of the order. The time constant 2 is changed over time as in the case of the first noise reduction unit 14a. In FIG. 4, the characteristic of the second noise reduction unit 14b is changed from a characteristic that attenuates 10 Hz or less to a characteristic that attenuates 200 Hz or less over time from time t5 to t6. As a result, at time t6 when the order is changed and the change of the time constant 2 of the second noise reduction unit 14b is completed, the characteristics are sufficient for reducing wind noise. Next, as shown in the waveform 41, the wind noise level decreases at time t7. Therefore, as shown in the waveform 43, the time constant 2 of the second noise reduction unit 14b is increased from time t8 to time t9 to obtain initial characteristics ( For example, the characteristic is attenuated to 10 Hz or less. Further, the time constant 1 of the first noise reduction unit 14a is increased from time t8 to t9 according to the noise level detected at time t7, and is changed to a characteristic that attenuates, for example, 50 Hz or less. When the time constant is increased, the time constant change timing is delayed until the processing of the delayed audio signal before the wind noise level is lowered. In the example of FIG. 4, even if a decrease in the level of wind noise is detected at time t7, the time constant change is actually started at time t8.

  In addition, after the change of the time constant is completed at time t9, the noise reduction control unit 19 controls the order change unit 15 to switch to the path of only the first noise reduction unit 14a as shown by the waveform 45. As shown in the waveform 41, the wind noise level further decreases at the time t10, so the time constant of the first noise reduction unit 14a is increased from the time t11 to the time t12 as shown in the waveform 42 to attenuate initial characteristics (for example, 10 Hz or less). Return to characteristics.

  In this way, it is possible to perform noise reduction based on an appropriate time constant from the beginning of the generation of wind noise by using the sound that temporarily holds the audio signal in the storage unit 13 and sequentially performs the necessary processing. . Note that a waveform 46 in FIG. 4 represents the state of the correction gain of the gain restoring unit 16. As described above, the amplification levels of the microphones 11R and 11L are automatically set according to the size of the subject (ALC: auto level control). This is because a small subject sound is recorded clearly and a large subject sound is recorded without saturation due to amplification. Therefore, when wind noise is generated, the amplification level of the microphone is reduced according to the level. As described above, since the wind noise is reduced by the noise reduction unit 14, if the amplification level of the microphone is small, the subject sound after the wind noise processing becomes small and unnatural.

  Therefore, complementation (gain restoration) of the amplification level changed after the wind noise processing is performed. In FIG. 1, the output of the order changing unit 15 is input to the gain restoring unit 16, and the gain is restored and output to the audio recording unit 17. The gain restoring unit 16 receives a gain change signal from the gain changing unit 12 that changes the amplification level of the microphone, and reports how much the gain has been changed at what timing. Therefore, the gain restoring unit 16 restores the gain according to the signal.

  A waveform 46 in FIG. 4 represents the gain restoration level, which is exactly synchronized with the change in the level of wind noise generation at each time. For example, if wind noise is generated at time t1, the amplification level of the microphone is reduced. Therefore, for the sound after noise processing, the amplification level is gradually increased from time t2 as shown in the waveform 46 in order to complement the amplification accordingly. ing. The reason why the gain is gradually changed is to prevent a sense of incongruity caused by abruptly changing the gain. At a later time, the gain is restored so as to compensate for the microphone amplification level that changes in accordance with the wind noise level.

  Next, the effects of the present invention will be described with reference to FIGS. In the present embodiment, the noise reduction unit 14 can wait by adjusting the characteristics prior to noise generation, and as a result, has a noise reduction effect immediately after noise generation in the audio signal. FIG. 5A shows a waveform in which wind noise is superimposed on the subject sound. The horizontal axis indicates time, and the vertical axis indicates the sound pressure level. Here, 51a and 53a are waveforms detected by the microphone 11R or 11L when there is no wind noise and only the subject sound, and are simulated by the spectrum of the male voice. Reference numeral 52a denotes a waveform in which wind noise is superimposed on the waveform 51a, and the waveform 51a is mixed with a waveform that is expressed in a pseudo manner from the spectrum of wind noise.

  FIG. 5B shows the result of processing the waveform of FIG. 5A by the noise reduction unit 14 having the characteristics shown in FIG. 2B, and the low-frequency component that is a wind noise is reduced (the low-frequency component is shown in the waveform 52b). There is no). However, when the waveforms 51b and 53b are compared with the waveforms 51a and 53a, the low frequency components are slightly deteriorated. That is, the subject sound is also affected by the noise reduction unit 14. Therefore, consider that the noise reduction unit 14 is operated only during a period in which wind noise is generated. FIG. 5C shows the result. The noise reduction processing is not performed on the waveforms 51c and 53c, and the noise reduction unit 14 is operated only in the section of the waveform 52c. Therefore, there is no difference between the waveforms 51a and 53a and the waveforms 51c and 53c. It seems that the wind noise in the waveform 52c is sufficiently reduced.

  FIG. 5 (d) shows the difference between FIG. 5 (b) and FIG. 5 (c). In the sections of the waveforms 51d and 53d in which no wind noise is generated, the difference between the waveforms 51b and 53b deteriorated by the noise reduction unit 14 and the waveforms 51c and 53c in which the noise reduction unit 14 is not operated is present. The degradation state of is shown.

  In the section of the waveform 52d where the wind noise is generated, since the noise reduction unit 14 is operated, the waveform is the same and the difference between the waveforms 52c and 52b should be zero. However, as shown in FIG. 5 (d), at the initial stage of the waveform 52d, the unprocessed portion of the wind noise remains in a circled area 54. This is because the effect cannot be obtained immediately even if the noise reduction unit 14 is operated.

  Therefore, for example, both a waveform in which the noise reduction unit 14 is always operated as shown in FIG. 5B and a waveform that has not been processed in the noise reduction unit 14 as shown in FIG. 5A are always prepared. A method of switching the waveform of FIG. 5B to the waveform of FIG. 5A only when a wind noise is generated is also conceivable. In this case, since the noise reduction unit 14 has already been operated before the generation of wind noise, the unprocessed waveform as in the range 54 is eliminated. However, when switching is performed, there is a problem that the joint becomes discontinuous. This is to connect the signal deformed by the noise reduction unit 14 (with attenuation or phase shift) and the original signal not transformed by the noise reduction unit 14. If this discontinuous part is heard as sound, the recording will be of poor quality.

  FIG. 5E shows processing corresponding to this embodiment. In the waveforms 51e and 53e, since the noise reduction unit 14 is not operated, the subject sound is not deteriorated as in the method of FIG. As for the waveform 52e in which the wind noise is generated, the low frequency component which is the wind noise is attenuated by the operation of the noise reduction unit 14.

  As described with reference to the timing chart of FIG. 4, the noise reduction unit 14 changes the frequency attenuation characteristic (decreases the time constant) prior to the generation of the wind noise, and at the time of the wind noise generation, the wind noise level is changed. The filter characteristics are suitable for reduction. That is, in the second half period 55 of the waveform 51e, the noise reduction unit 14 is gradually activated, and a noise reduction effect is produced. Since the time constant of the noise reduction unit 14 is thus reduced over time (period 55), the noise reduction unit is compared with the case of FIG. 5C in which the noise reduction unit 14 is suddenly operated. The continuity of the waveform before and after 14 operations is maintained.

  FIG. 5 (f) shows the difference between FIG. 5 (e) and FIG. 5 (b). In the sections of the waveforms 51f and 53f in which no wind noise is generated, the difference between the waveforms 51b and 53b deteriorated in the noise reduction unit and the waveforms 51e and 53e in which the noise reduction unit is not operated is present. Is shown. In the section of the waveform 52f where the wind noise is generated, since the noise reduction unit 14 is operated, the waveform is the same and the difference between the waveforms 52e and 52b should be zero. As shown in FIG. 5 (f), in this embodiment, wind noise can be reduced even in the early stage of wind noise generation in a circled range 56, and high-quality recording is possible.

  Next, with reference to FIG. 6, the process in the noise detection part 18 in this embodiment is demonstrated. This process starts when voice recording starts.

  The operation starts in synchronization with the sound being stored in the storage unit 13 of FIG. 1 by the microphones 11R and 11L. In S601, the noise detector 18 detects wind noise by comparing the microphones 11R and 11L. If wind noise is detected (“YES” in S601), the process proceeds to S602. While the wind noise is not detected (“NO” in S601), the generation of the wind noise is monitored. In subsequent S602, the noise detector 18 uses the amplitude (sound pressure) of the waveform with no correlation between the microphones 11R and 11L, the amplitude of the waveform after the subject sound is attenuated by the low-pass filter, or the like as the noise level. To measure.

  In subsequent S603, the noise detection unit 18 notifies the storage unit 13 of the timing and level at which the wind noise has been detected. The storage unit 13 records the wind noise processing instruction in synchronization with the audio signal stored in the storage unit 13. At this time, a “voice processing instruction” for instructing the processing of wind noise is recorded at the timing when the detection delay in the noise detection unit 18 is expected. Thereafter, the process returns to S601.

  Next, noise reduction processing, gain restoration, and voice recording processing in the present embodiment will be described with reference to FIG. This process starts after a predetermined time from the start of voice recording. This time delay allows for the detection delay of the noise detection unit 18 and the startup delay of the noise reduction unit 14. First, the noise reduction control unit 19 reads and processes the audio signal from the storage unit 13 with a delay that can reliably read the recording of the audio processing instruction recorded in the process of FIG. The first noise reduction unit 14a further reads and processes the audio signal from the storage unit 13 with a predetermined time delay necessary for changing the time constant. Alternatively, the overall control unit 20 may control the readout timing of the audio signal in each functional block.

  First, in S701, the noise reduction control unit 19 searches for the audio processing instruction recorded in S603 of FIG. 6 while scanning the audio signal recorded in the storage unit 13. If a voice processing instruction is found (“YES” in S701), the process proceeds to S702. If no voice processing instruction is found, monitoring is continued in S701.

  In subsequent S702, S703, and S704, the noise reduction control unit 19 determines the noise level recorded in the voice processing instruction and selects noise reduction according to the noise level. The noise level is determined using a threshold corresponding to each level. Level 1 uses threshold Th1, Level 2 uses threshold Th2, and Level 3 uses threshold Th3. First, in S702, it is determined whether or not the noise level is higher than level 3 (threshold Th3) or higher. If the noise level is higher than level 3 ("YES" in S702), the process proceeds to S709, otherwise (" NO ") Proceed to S703. In S703, it is determined whether or not the noise level is equal to or higher than the intermediate level 2 (threshold Th2). If the noise level is equal to or higher than level 2 ("YES" in S703), the process proceeds to S708, otherwise (S703 "NO") S704. Proceed to In S704, it is determined whether the noise level is equal to or higher than the lowest level 1 (threshold Th1). If the noise level is equal to or higher than Level 1 (“YES” in S704), the process proceeds to S705. If the noise level is less than the threshold value Th1 (“NO” in S704), it is determined that the level cannot be detected, the voice processing instruction is ignored, and the process returns to S701.

  In S705, the time constant 1 of the first noise reduction unit 14a is changed to a corresponding time constant. Here, since the noise level is low, the time constant is not reduced so much, for example, the time constant is changed to τ 1 corresponding to the filter characteristic for attenuating noise of 100 Hz or less. In this embodiment, the time constant is changed over a predetermined time (for example, 100 ms), but the change can be performed based on the control signal 24 from the noise reduction control unit 19.

  In subsequent S 706, the gain restoring unit 16 restores the gain of the speech waveform processed by the noise reducing unit 14. As described above, when wind noise is generated, the gain for converting the sound into an electrical signal by the gain changing unit 12 is automatically reduced. Therefore, after performing the wind noise reduction process, the gain change is complemented and the recording level is kept constant. In subsequent S707, the audio signal whose gain has been restored is recorded in the recording medium 21, and the process returns to S701.

  Next, the case where it is determined in S703 that the level is 2 or higher will be described. In S708, the noise reduction control unit 19 changes the time constant 1 of the first noise reduction unit 14a to a corresponding time constant. Since the noise level at this time is relatively high, the time constant is made sufficiently small, for example, changed to a time constant corresponding to a filter characteristic that attenuates noise of 200 Hz or less: τ2 over a predetermined time. Thereafter, the process proceeds to S706.

  Next, the case where it is determined in S702 that the level is 3 or higher will be described. In this case, not only the first noise reduction unit 14a but also the second noise reduction unit 14b are operated to reduce noise. Therefore, in S709, the noise reduction control unit 19 switches the order changing unit 15 and selects processing by a secondary high-pass filter in which the first noise reduction unit 14a and the second noise reduction unit 14b are connected in series.

  In subsequent S710, the time constants of the first noise reduction unit 14a and the second noise reduction unit 14b are changed after the order change unit 15 is actuated. Since the noise level at this time is the maximum level, the upper limit for changing the time constant is also the maximum level. For example, the time constant corresponding to the filter characteristic for attenuating noise of 200 Hz or less: τ2 is changed over a predetermined time. If the time constant 1 of the first noise reduction unit 14a has already been changed to the time constant τ2, only the time constant 2 of the second noise reduction unit 14b needs to be changed. Thereafter, the process proceeds to S706. The above processing continues until an instruction to end audio recording is given.

  As described above, in the present embodiment, the audio signal is temporarily stored in the storage unit 13 and the location where the wind noise is generated is detected and recorded in advance. In the actual noise reduction process, prior to the occurrence timing of the detected wind noise, the time constant of the high-pass filter that is the noise reduction unit is changed to a value corresponding to the detected noise level. Thereby, the characteristics of the high-pass filter can be stabilized so as to have sufficient wind noise reduction characteristics immediately before the wind noise generation location in the audio signal. As a result, the effects of wind noise detection delay and noise activation start delay on noise reduction processing are eliminated, and high-quality audio recording is possible while maintaining continuity of the signal before and after activation of the noise reduction unit. .

[Embodiment 2]
Hereinafter, Embodiment 2 will be described. In this embodiment, a configuration using a thermocouple as a noise detection mechanism will be described. FIG. 8 is a diagram illustrating a configuration example of an audio signal processing device corresponding to the second embodiment. In FIG. 8, a monaural microphone is used as a mechanism for collecting sound.

  The voice detection unit 80 includes a microphone unit 81, a microphone 82, a microphone support member 83, and a thermocouple 84, and is attached to a casing 85 of a voice signal processing device shown partially. The housing 85 is provided with a hole 85a, and the sound reaches the microphone 82 through the hole 85a. Since the microphone 82 is elastically supported by the microphone support member 83 formed of vibration-proof rubber with respect to the microphone unit 81, the vibration generated by the recording device itself is not recorded as sound.

  The thermocouple 84 is attached to the microphone unit 81. The thermocouple 84 is used for an anemometer or the like, and detects the wind speed based on a resistance change caused by the thermocouple 84 being cooled by wind. Since the wind flowing in from the hole 85a is convected in the microphone unit 81, the phase of the wind hitting the microphone and the wind hitting the thermocouple 84 may be shifted. Therefore, the thermocouple 84 needs to be arranged as close to the microphone 82 as possible. When the thermocouple 84 is arranged in the microphone unit 81, the thermocouple 84 is not exposed to the housing 85, and thus the reliability is improved.

  The audio signal from the microphone 82 is subjected to noise reduction processing and recorded on the recording medium 21 as in the first embodiment. The signal from the thermocouple 84 is input to the noise detector 18. The noise detector 18 amplifies the signal of the thermocouple 84 and detects the presence and level of wind speed using a known wind comparator or the like. The noise reduction control unit 19 controls the noise reduction unit 14 in accordance with the wind noise level from the noise detection unit 18 to perform wind noise reduction processing in the same manner as described in the first embodiment.

  In general, there is a delay in detecting wind speed by a thermocouple. Therefore, when trying to process wind noise in real time, it is difficult to reduce wind noise at the initial stage due to this detection delay. Therefore, in this embodiment, a delay correction unit 86 is newly provided between the noise detection unit 18 and the noise reduction control unit 19. The delay correction unit 86 is set with a detection delay (for example, 0.2 seconds) of the thermocouple 84. In S603 of FIG. 6, the noise reduction unit 14 is activated in anticipation of the detection delay and the time constant change time. The timing is determined, and an audio processing instruction is recorded in the audio signal stored in the storage unit 13. For this reason, the detection delay of wind noise is eliminated during recording of sound on the recording medium 21.

  The power supply state detection unit 87 detects the consumption state of the power supply 88, and when the power supply is lowered to a predetermined level, notifies the noise detection unit 18 and stops driving the thermocouple 84. This is because driving the thermocouple 84 further consumes power. The power supply state detection unit 87 also notifies the noise reduction control unit 19 at the same time so that the time constant for the noise reduction unit 14 is fixed to the time constant set at that time. If the power has been consumed since the start of recording, after the thermocouple 84 is driven once and the wind noise level is detected, the characteristics of the noise reduction units 14a and 14b and the state of the order changing unit 15 are determined. After fixing them, the driving of the thermocouple 84 is stopped. The power supply 88 supplies power for operating the audio signal processing apparatus, for example, a battery. In FIG. 8, the connection relationship with other than the power supply state detection means 87 is omitted, but power is supplied to each functional block. It is carried out. In FIG. 1, the description is omitted because it is not directly related to the operation of the invention.

  FIG. 9 is a flowchart for explaining the above, and this flow is started after the power of the audio signal processing apparatus is turned on. In S901, the power supply state detection unit 87 detects the power supply state. If a state in which the power consumption is consumed below a predetermined level is detected (“YES” in S901), the process proceeds to S902. If the power is not consumed below a predetermined level (“NO” in S901), the process returns to S901 to continue monitoring the power supply state. In this case, normal noise reduction processing is performed.

  In subsequent S902, it is determined whether or not noise detection has already been performed since the start of recording, and if it has been detected (“YES” in S902), the process proceeds to S907. If noise detection has not yet been performed (“NO” in S902), the process proceeds to S903. The presence or absence of this noise detection process may be determined based on whether or not the noise detection unit 18 has the previous detection value. Note that the detected value is reset when the power is turned on, and it is determined that no noise is detected when the power is turned on.

  In subsequent S903, the noise detection unit 18 drives the thermocouple 84, and in subsequent S904, the wind noise level is detected by the thermocouple 84. In S905, the noise reduction control unit 19 determines the time constants of the noise reduction units 14a and 14b according to the detected wind noise level, and determines the state of the order changing unit 15. At this time, if noise detection has already been performed and the time constant and order are set, the time constant and order already set are not changed. On the other hand, when noise detection is newly performed (that is, when “NO” is determined in S902), the time constant and the order are set in accordance with the noise level detected in S904. The set value is fixed thereafter. When changing the time constant and the order, for example, after changing the time constant of the noise reduction unit 14a, the order is changed by the order change unit 15, and then the time constant of the noise reduction unit 14b is changed. In subsequent S906, the noise detection unit 18 stops driving the thermocouple 84 to save power, and the process returns to S901.

  In S907, the noise detection unit 18 determines whether or not a certain time has elapsed since the previous noise detection. Here, the certain time can be set to 0.5 s, for example. If the predetermined time has elapsed (“YES” in S907), the process proceeds to S903. on the other hand. If the predetermined time has not elapsed (“NO” in S907), the process proceeds to S905.

  According to the above configuration, power consumption due to operation of the thermocouple 84 can be suppressed. This method contributes to power saving, but if the time constant and the order of the noise reduction unit 14 are fixedly set, there may be cases where it is not possible to cope with subsequent changes in wind noise. Therefore, in S907, it is determined whether or not a certain time has elapsed after the previous noise detection, and the thermocouple 84 can be driven intermittently. As a result, it is possible to cope with changes in wind noise and to save power by suppressing power consumption.

  As described above, the noise detection is performed using the thermocouple in the second embodiment, but the wind noise detection accuracy can be increased in combination with the wind noise detection based on the output correlation value of the pair of microphones as in the first embodiment. . For example, there is a method in which the wind noise generation timing is grasped using a pair of microphone output correlation values, and the level is detected by a thermocouple. Further, it may be determined that wind noise is generated only when wind noise is detected, and noise reduction processing may be performed.

  Moreover, although FIG. 8 is an embodiment of a monaural microphone, a thermocouple can also be used in a stereo microphone. For example, as shown in FIG. 10, the thermocouple 84 is provided only on one side of the stereo microphone, and the audio signal of each of the left and right channels may be processed by the output, or the thermocouple 84 is provided on both of the stereo microphones. Audio signal processing may be performed with the average value.

  In Embodiment 2 of the invention described above, wind noise can be detected with higher accuracy by providing a thermocouple for wind noise detection.

(Other examples)
The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media, and a computer (or CPU, MPU, or the like) of the system or apparatus reads the program. It is a process to be executed.

Claims (10)

Voice collection means;
Storage means for storing the voice signal collected by the voice sampling means;
Noise detection means for detecting noise contained in the audio signal stored in the storage means;
Noise reduction means for reducing noise detected by the noise detection means from the voice signal;
An audio signal processing apparatus comprising: a setting unit that sets a time constant for controlling a noise reduction characteristic of the noise reduction unit according to a level of the noise detected by the noise detection unit;
The setting means starts setting the time constant to the noise reduction means in response to detection of the noise by the noise detection means,
The noise reduction unit reads the audio signal from the storage unit and reduces the noise when the setting of the time constant according to the noise level by the setting unit is completed. Signal processing device. The noise reduction means is configured by connecting a first noise reduction means and a second noise reduction means in series,
The audio signal processing apparatus further includes a switching unit that switches between using either the first noise reduction unit and the second noise reduction unit, or using only one of them.
The setting means uses both the first noise reduction means and the second noise reduction means when the noise level exceeds a level corresponding to a predetermined time constant set in the noise reduction means. The audio signal processing apparatus according to claim 1, wherein the switching means is switched so as to do so.   When both the first noise reduction unit and the second noise reduction unit are used, the setting unit switches the switching unit to use both, and then sets the time constant of the second noise reduction unit. The audio signal processing apparatus according to claim 2, wherein the setting of the audio signal is started. The said noise detection means has a thermocouple provided in the vicinity of the said audio | voice collection means, The said noise is detected based on the signal detected by the said thermocouple, Any one of Claim 1 thru | or 3 characterized by the above-mentioned. The audio signal processing device according to claim 1. The noise detection means detects the noise by activating the thermocouple every predetermined time,
5. The audio signal according to claim 4, wherein the setting unit fixedly sets a time constant of the noise reduction unit during the predetermined time based on a noise level detected by the noise detection unit. Processing equipment. The audio signal processing device further includes power state detection means for detecting a power consumption state of the audio signal processing device,
When the power supply state detection unit determines that the power supply is consumed below a predetermined level, the noise detection unit turns the thermocouple only when it is determined that a time constant is not set in the noise reduction unit. Activate and detect the noise,
6. The audio signal processing apparatus according to claim 4, wherein the setting unit fixedly sets a time constant of the noise reduction unit based on a noise level detected by the noise detection unit. The audio sampling means is a stereo microphone, and the thermocouple is provided only in the vicinity of the microphone of one of the left and right channels,
The audio signal according to any one of claims 4 to 6, wherein the noise detection unit detects the noise further based on a comparison of audio signals of left and right channels collected by the stereo microphone. Processing equipment. The voice sampling means is a stereo microphone,
4. The audio signal processing according to claim 1, wherein the noise detection unit detects the noise based on a comparison of audio signals of left and right channels collected by the stereo microphone. 5. apparatus. Voice collection means;
Storage means for storing the voice signal collected by the voice sampling means;
Noise detection means for detecting noise contained in the audio signal stored in the storage means;
Noise reduction means for reducing noise detected by the noise detection means from the voice signal;
A control method for an audio signal processing apparatus comprising: a setting unit that sets a time constant for controlling a noise reduction characteristic of the noise reduction unit according to a level of the noise detected by the noise detection unit,
The storage means storing the voice signal collected by the voice sampling means;
The noise detecting means detecting noise included in the audio signal stored in the storage means;
The setting means starting the setting of the time constant to the noise reduction means in response to detection of the noise by the noise detection means;
The noise reduction means comprises a step of reading the audio signal from the storage means and reducing the noise when the setting of the time constant according to the noise level by the setting means is completed. A control method of an audio signal processing device.   The program for functioning a computer provided with an audio | voice collection means and a memory | storage means to memorize | store the audio | voice signal collected by the said audio | voice collection means as an audio | voice signal processing apparatus of any one of Claim 1 thru | or 8.