JP4241296B2 - Howling suppression device and karaoke device - Google Patents

Description

  The present invention relates to a technique for suppressing howling that occurs in a device such as a karaoke apparatus.

  In an audio device that amplifies the signal input from the microphone and outputs it from the speaker, a closed loop is formed by re-inputting the output from the speaker to the microphone, which may cause unpleasant and unpleasant howling. . Since this howling occurs when a constant frequency component having a large peak continues for a long time in the input signal of the microphone, various techniques for suppressing howling have been proposed focusing on this feature. Yes. As such a technique, for example, when a frequency analysis is performed on a microphone input signal and the above characteristic appears in the analysis result, it is determined that howling has occurred, and the center frequency of the notch filter is set to a peak frequency, A technique for attenuating a howling frequency component by filtering an input signal of a microphone (see, for example, Patent Document 1).

Japanese Patent Application Laid-Open No. 7-143034 (refer to the description of FIG. 1, FIG. 3, paragraphs 0026 and 0030)

By the way, according to the above technique, when speech such as a lecture or conversation is input by a microphone, it is relatively accurate to determine whether or not howling has occurred, but speech such as singing is input. In some cases, the problem was pointed out that the judgment tends to be inaccurate for the following reasons. That is, depending on the way of singing, there are cases where the peak of the voice is large and the constant frequency component continues for a long time, and the result of analyzing the input signal of the microphone at this time is very close to the characteristics of howling.
The present invention has been made in view of the above-described circumstances, and an object thereof is to provide a howling suppression apparatus and a karaoke apparatus capable of accurately determining and suppressing howling.

  In order to achieve the above object, a howling suppression apparatus according to the present invention analyzes a signal input by a microphone, specifies a peak frequency, and a notification for notifying a target frequency of sound input to the microphone. Determining means for determining whether a relationship between the peak frequency specified by the specifying means and the target frequency notified by the notifying means satisfies a predetermined condition, and the predetermined condition is determined by the determining means. And suppressing means for suppressing and outputting at least a component of the frequency specified by the specifying means in the input signal when it is determined that the condition is satisfied. According to the present invention, when a howling feature appears in an input signal, it is easy and accurate to determine whether the cause is truly due to howling or not.

Te present invention smell, said a predetermined condition, the peak frequency is a frequency range including the target frequency, the frequency range contains a frequency an integral multiple of the target frequency, or the inverse integer multiple of the target frequency The case where it is not included in any of the frequency ranges including the frequency is considered. As these ranges, it is desirable to be defined by the pitch deviation from the target frequency.

Moreover, you may comprise as a following karaoke apparatus by applying the howling suppression apparatus which concerns on this invention. That is, as a karaoke device, an input signal from a microphone is analyzed, a specifying means for specifying the peak frequency, a performance sequencer for outputting performance information and main melody information according to the progress of the performance, and a musical sound based on the performance information A tone synthesizing unit that synthesizes a signal; a notification unit that converts the main melody information into a target frequency of a voice that is input to the microphone; and a notification that the frequency specified by the specifying unit is notified by the notification unit. A determination unit that determines whether or not a relationship with a target frequency satisfies a predetermined condition; and when the determination unit determines that the predetermined condition is satisfied, at least the specifying unit is specified among the input signals. and and a suppression means component suppressing to the output of the frequency, the a predetermined condition, the peak frequency A frequency range including the target frequency, a frequency range including a frequency that is an integer multiple of the target frequency, or a frequency range including a frequency that is an inverse integer multiple of the target frequency. It is good also as a structure.

  According to the present invention, it is possible to accurately determine and suppress howling.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a karaoke apparatus to which the howling suppression apparatus according to the first embodiment of the present invention is applied.
In this figure, the microphone 10 converts the voice of the singer into an analog singing signal. The frequency analyzer 20 includes an A / D converter 22, a discrete Fourier transformer (DFT) 24, an integrator 26 and an extractor 28. Among these, the A / D converter 22 samples the singing signal from the microphone 10 at a sampling frequency of 32 kHz, for example, quantizes it, and outputs the data. The DFT 24 performs Fourier transform on the data obtained by quantizing the singing signal and obtains its frequency spectrum. In this frequency spectrum, the integrator 26 integrates (averages) data indicating the size of each frequency region, for example, every 2 seconds, that is, every cycle of a frequency sufficiently lower than the sampling frequency of the singing signal. The extractor 28 extracts howling if there is an integration value of data indicating the size of each frequency range that is greater than or equal to the threshold th. On the other hand, if the frequency f1 is output, if there is no frequency f1 or more, a signal indicating that is output.
The singing signal from the microphone 10 is also supplied to the input end of the notch filter 50. The notch filter 50 is an analog filter whose frequency characteristics are set by a determiner 30 described later.

On the other hand, the storage unit 40 stores MIDI data for a plurality of songs. Here, the MIDI data for one song includes guide melody data indicating the main melody of the song in addition to the performance data for generating the karaoke accompaniment sound.
The performance sequencer 42 reads out and outputs the MIDI data of the music designated by the operation unit (not shown) from the storage unit 40 with the passage of time from the start of the performance. The sound source 44 generates and outputs a musical sound signal based on the performance data supplied from the performance sequencer 42.
Further, when receiving the note-on data from the guide melody data supplied from the performance sequencer 42, the converter 46 converts the pitch data included in the note-on data into frequency data and notifies the frequency data. Since the singer should sing at a pitch (pitch) along the main melody of the song, this frequency data indicates the target pitch of the singing in terms of frequency. Therefore, the frequency indicated by the frequency data is set as the target frequency f2.

When the peak frequency f1 is output from the extractor 28 and the difference between the peak frequency f1 and the target frequency f2 notified by the converter 46 exceeds, for example, 200 cents, the determiner 30 generates howling. It is determined that If the determiner 30 determines that howling has occurred, the determiner 30 controls the notch filter 50 so that the notch frequency fn becomes the peak frequency f1.
On the other hand, the determiner 30 determines that the peak frequency f1 is not output from the extractor 28, that is, if there is no integrated value of the data indicating the size of each frequency region that exceeds the threshold th, and the peak Even when the frequency f1 is output, when the difference between the peak frequency f1 and the target frequency f2 is within 200 cents, it is determined that no howling has occurred. If the determiner 30 determines that no howling has occurred, it does not change the frequency characteristic of the notch filter 50.
The mixer 52 mixes the singing signal filtered by the notch filter 50 and the musical sound signal from the sound source 44. The sound system 54 emits sound after appropriately amplifying the mixed signal.

  Next, the operation of the karaoke apparatus having such a configuration will be described. First, the MIDI data of the designated song is read from the storage unit 40 by the performance sequencer 42 as time passes. Among the MIDI data to be read, the performance data is supplied to the sound source 44, so that a musical tone signal that is an accompaniment sound of karaoke is output.

  When the accompaniment sound playback point reaches the singing section, the karaoke singer starts singing. The voice of the singer is converted into a singing signal by the microphone 10 and supplied to the A / D converter 22 and the notch filter 50, respectively. The singing signal from the microphone 10 is converted into digital data by the A / D converter, and the frequency spectrum is obtained by the DFT 24. Data indicating the size of each frequency region in the frequency spectrum is temporally averaged by integration by the integrator 26. Then, as shown in FIG. 2, the extractor 28 extracts any average value that is equal to or greater than the threshold th, and outputs the peak frequency as f1. The If there is no threshold value th or more, a message to that effect is output.

In the present embodiment, the reason why the data indicating the size of each frequency region of the frequency spectrum is temporally averaged is that the singing signal includes not only the voice of the singer but also the sound emission component by the sound system 54. This is because of this.
This point will be described in detail. The sound emission component of the sound system 54 may include a single sound having a high level as an accompaniment sound, such as a hitting sound or a bursting sound like a drum. It must be avoided that the peak frequency of such a sound is analyzed as the peak frequency f1 of the singer. On the other hand, in singing, the sound usually continues for a long time. Therefore, in the present embodiment, as described above, data indicating the size of each frequency region in the frequency spectrum is averaged over time by integration by the integrator 26. Even if the singing signal includes a single-shot sound with a high level, the temporal average value is small, so it does not exceed the threshold th. Therefore, in this embodiment, the peak frequency of such a sound is not analyzed as the peak frequency of the singer.

  Of the MIDI data read out by the performance sequencer 42, the guide melody data is supplied to the converter 46, and the converter 46 converts the guide melody data into frequency data indicating the target frequency f2 as described above. Output. Since the guide melody data is read together with the performance data as time passes, the target frequency f2 also changes as the performance progresses as shown in FIG.

  Here, even when sound or the like is input to the microphone 10, if the peak average value of the frequency spectrum is lower than the threshold th, that is, if it is clear that no howling has occurred, the extractor 28 Therefore, the peak frequency f1 is not applicable. In this case, as shown in FIG. 4A, the frequency characteristics of the notch filter 50 are in an initial state flattened over the entire frequency range, so that the singing signal of the microphone 10 is directly mixed with the musical sound signal, It is output to the system 54.

In addition, when a relatively large level signal is input to the microphone 10 and the peak average value of the frequency spectrum is equal to or greater than the threshold th, the determiner 30 is caused by singing or the like. In order to identify whether it is due to howling or not, the following determination is performed. That is, the determiner 30 determines whether or not the difference between the peak frequency f1 and the target frequency f2 is within 200 cents, that is, the peak frequency f1 is within a range of ± 200 cents around the target frequency f2 in that case. It is determined whether it is within.
At this time, as shown in FIG. 5A, when the peak frequency f1 is within a range R1 of ± 200 cents around the target frequency f2, the determiner 30 determines that the cause is not due to howling. It is determined that it is due to singing. This is based on the assumption that no matter how bad the singing is, the pitch does not deviate from the main melody by 200 cents. However, the present invention is not intended to be limited to this range. This range is of a nature determined by trial and error.

When the determiner 30 determines that the cause is due to singing, the frequency characteristic of the notch filter 50 is maintained in the initial state flattened over the entire frequency range, as shown in FIG. Therefore, the singing signal of the microphone 10 is directly mixed with the musical sound signal and output to the sound system 54.
On the other hand, as shown in FIG. 5B, when the peak frequency f1 is not in the range R1 of ± 200 cents around the target frequency f2, the determiner 30 determines that the cause is not due to singing and is true. It is determined that this is due to howling. If it is determined by the determiner 30 that howling has occurred, the notch filter 50 is set to the peak frequency f1 as the notch frequency fn. For this reason, the notch filter 50 has a frequency characteristic as shown in FIG. Therefore, the signal filtered by the notch filter 50, that is, the signal including the sound emitted by the sound system 54 as well as the sound of the singer, attenuates the component of the peak frequency f1 that causes the howling. Thus, since the sound signal is mixed and emitted, unpleasant howling can be suppressed.

  As described above, in the first embodiment, even when a relatively large level signal is input to the microphone 10 and the peak average value of the frequency spectrum is equal to or greater than the threshold value th, the signal is a component along the main melody. If so, howling suppression operation is not executed. For this reason, the atmosphere of karaoke is not impaired by unnatural filtering. Further, in the first embodiment, when a relatively large level signal is input to the microphone 10 and the peak average value of the frequency spectrum becomes equal to or greater than the threshold th, the peak frequency f1 and the target frequency When the difference from f2 exceeds 200 cents, it is determined for the first time that howling has occurred, so that the determination accuracy is improved.

Next, a second embodiment of the present invention will be described. In the second embodiment, the difference from the first embodiment is only the determination contents in the determiner 30. Therefore, the configuration is exactly the same as that in FIG. 1, and therefore, different determination contents will be described here.
As shown in FIG. 6, the determiner 30 in the second embodiment includes a range R1 in which the peak frequency f1 that is equal to or higher than the threshold th in the frequency spectrum is ± 200 cents around the target frequency f2, and the target frequency f2. Whether or not it is included in any one of a range R2 of ± 200 cents centered on a frequency that is 1/2 times the frequency of R2 or a range R3 of ± 200 cents centered on a frequency that is twice the target frequency f2. judge.
If the peak frequency f1 is included in any of the ranges R1, R2, and R3, the determiner 30 determines that no howling has occurred and maintains the frequency characteristics of the notch filter 50, while the range R1. If it is not included in any of the ranges R2, R3, it is determined that howling has occurred, and the frequency characteristic of the notch filter 50 is controlled with the peak frequency f1 as the notch frequency fn.

According to the second embodiment, even if the peak frequency f1 is not in the range R1, if it is in the ranges R2 and R3, it is determined that it is not howling. Therefore, in the accompaniment sound picked up by the microphone 10, the lower harmonic string of the main melody, upper It is possible to prevent the frequency characteristics of the notch filter 50 from being controlled insignificantly when the component of the overtone string becomes large or when the singer intentionally sings at a pitch one octave above or above. .
In the second embodiment, the frequency of the target frequency f2 is doubled, that is, not only the second harmonic of the guide melody but also the third harmonic, the fourth harmonic,... Similarly, not only the ½ harmonic, but also the lower harmonic series such as な く, ¼, etc., may be excluded from the determination of howling occurrence.

In the first and second embodiments, there is a possibility that howling occurs when a relatively large level signal is input to the microphone 10 and the peak average value of the frequency spectrum is equal to or greater than the threshold th. Although the configuration is assumed to be, a third embodiment using a different determination method will be described.
FIG. 7 is a diagram illustrating a waveform of a singing signal from the microphone 10 when howling occurs. It can be said that it is a sine wave with almost no distortion as shown in this figure. In FIG. 7, the signal waveform at the time of howling is extracted for a relatively long time. However, even if attention is paid for a longer time, the signal waveform is stable in both period (pitch) and amplitude. On the other hand, the signal waveform indicating the human voice is a relatively distorted waveform as shown in FIG. 8, and has a periodicity when focusing on time. However, although not shown, when attention is paid to a change over a longer time, it can be easily imagined that the randomness becomes high without being stabilized in both period and amplitude. Therefore, it is possible to determine whether or not there is a high possibility of howling occurring by determining whether the pitch and amplitude of the singing signal are not random and stable over a certain period of time. it can. In the third embodiment, such a determination is performed.

FIG. 9 is a diagram showing the configuration of the third embodiment, which is a part that replaces the frequency analyzer 20 in FIG. In this figure, a pulse conversion circuit 71 is a circuit that converts a singing signal into a pulse signal Pt. Specifically, in FIG. 10, the pulse conversion circuit 71 creates the positive envelope 72 and the negative envelope 73 of the singing signal, and sets the pulse signal Pt to the H level at the cross point between the positive envelope 72 and the singing signal. On the other hand, the pulse signal Pt is changed to the L level at the crossing point between the negative envelope 73 and the singing signal.
Each time the pulse signal Pt falls, the counter 74 outputs a count value P0 indicating a period in which the pulse signal Pt was at the H level immediately before. On the other hand, every time the pulse signal Pt rises, the counter 75 outputs a count value P1 indicating a period in which the pulse signal Pt was at the L level immediately before.
The adder 77 calculates the sum of the count value P0 and the count value P1 output following the count value P0, and calculates the positive pitch P (+) in the singing signal. Similarly, the adder 77 calculates the sum of the count value P1 and the count value P0 output immediately thereafter, and calculates the negative pitch P (−) in the singing signal.
On the other hand, the level detector 78 detects the maximum value (peak) and the minimum value (bottom) of the singing signal, and outputs an amplitude At which is the difference between them.

The index calculation / discrimination circuit 79 samples the positive pitch P (+), the negative pitch P (−), and the amplitude At for a predetermined period, respectively, and indicates an index (for example, variance) indicating their randomness. While calculating, it is discriminate | determined whether each of those parameter | index is less than a threshold value. The index calculator / discriminator 79 determines that there is a possibility of howling if each index is within the threshold value, and uses the positive pitch P (+) or the negative pitch P (−) as the frequency. While this is converted and output as the peak frequency f1, if at least one of the indices exceeds the threshold, it is determined that no howling has occurred and the peak frequency f1 is not output.
Therefore, when the peak frequency f1 is output from the index calculation / discrimination circuit 79 and the difference between the peak frequency f1 and the target frequency f2 exceeds 200 cents, the determiner 30 in FIG. 1 generates howling. The notch filter 50 is controlled so that the notch frequency fn becomes the peak frequency f1. On the other hand, if the peak frequency f1 is not output, the determination that no howling has occurred is confirmed. Do not change the frequency characteristics.
According to the third embodiment, similar to the first embodiment shown in FIG. 1, howling can be appropriately suppressed, and an operation such as DFT is unnecessary as compared with the first embodiment. Therefore, the configuration can be simplified.
In the third embodiment mobile phone, both the positive side pitch P (+) and the negative side pitch P (−) are used, but either one may be used. Also in the third embodiment, the second embodiment as shown in FIG. 6 may be applied as the determination contents of the determiner 30.

In each embodiment described above, the setting of the attenuation of the notch filter 50 is not particularly mentioned, but the attenuation may be set in conjunction with the microphone volume of the karaoke apparatus. Specifically, when detecting howling, when the initial attenuation is set to, for example, −6 dB in the notch filter 50 and the microphone volume is operated to increase by 2 dB, the attenuation of the notch filter 50 is also 2 dB. The configuration is reset so as to increase. According to this configuration, when an operation for increasing the microphone volume is performed, it is possible to prevent recurrence of howling due to the operation. In addition, when the microphone volume is operated to be reduced by 2 dB, the attenuation of the notch filter 50 is reset so as to be reduced by 2 dB. As a result, when the microphone volume is reduced, the sound quality can be prevented from deteriorating due to unnecessary excessive attenuation.
In the configuration in which the attenuation is changed, the Q value (peak value) in the frequency characteristic of the notch filter 50 may be changed according to the attenuation. This is because if the attenuation is changed while the Q value is kept constant, the frequency range from the center frequency to be attenuated becomes narrow when the attenuation is small, so that there is a possibility that howling cannot be sufficiently suppressed. Because there is. Therefore, when changing the degree of attenuation, it is preferable that the frequency range to be attenuated is secured with a certain width. Specifically, if the Q value is set so as to decrease as the attenuation decreases, the frequency range to be attenuated can be secured with a certain width.

  Further, in each of the above-described embodiments, the singing that is the erroneous determination factor of howling is estimated on the basis of the target frequency f2, but in addition to this, the harmonic series components (double, 3 and 3) of the peak frequency f1 by the extractor 28 are estimated. (Times, 4 times,...) That exceed a predetermined threshold may be determined as a singing component. This is based on the fact that the singing includes a harmonic component in addition to the fundamental frequency component. When howling occurs, only the peak frequency f1 component is generated, and no overtone component is generated, so that it is easy to separate them. Such a determination may be replaced with a singing determination based on the target frequency f2, but the harmonic component ratio varies depending on the voice quality of the singer, and there is a possibility of erroneous determination based on the frequency component of the accompaniment sound. It is preferable to determine by combining these determinations. Of course, only one of the determinations may be adopted, or the determination may be made based on the logical product or logical sum of the two determinations, and it goes without saying that various application patterns can be considered.

In each embodiment, since the notch frequency fn may be set at a plurality of locations in the notch filter 50, an upper limit may be provided for the set number. When the set number reaches the upper limit, for example, the configuration may be such that the notch frequency is deleted in order from the oldest set time.
In this way, when an upper limit is set for the set number, if a notch frequency is deleted, the notch frequency is not attenuated. If the upper limit number is small, there is a possibility that performance will be lowered. Therefore, when a new notch frequency is to be set in a state where the upper limit number has been reached, the frequency characteristic applied to the set notch frequency closest to the notch frequency is reset so that the Q value becomes low. The frequency range to be expanded is expanded to cope with the setting of a new notch frequency. Thereafter, when a new notch frequency is to be set, similarly, the frequency characteristics related to the closest set notch frequency may be reset so that the Q value becomes lower. In this way, even when there is an upper limit to the number of set notch frequencies, it is possible to avoid reducing the howling suppression performance.

By the way, the frequency characteristic of the notch filter 50 needs to be returned to the initial state at an appropriate timing. Of course, when the suppression of howling is disabled, this timing is not limited even when the power is turned off. good. However, since howling often depends on the surrounding environment, it is considered that the notch filter 50 converges to the same frequency characteristic every time if the device is fixedly installed.
For this reason, for example, if the device is fixedly installed, the frequency characteristic (notch frequency) of the notch filter 50 immediately before turning off the power (or immediately before disabling howling suppression) is stored in a nonvolatile manner, and the power After turning on (or after suppressing howling), the frequency characteristics of the notch filter 50 may be set to be the stored frequency characteristics.

  In each embodiment described above, the notch filter 50 is an analog filter, but may be a digital filter. When the notch filter is a digital filter, the output of the A / D converter 22 is supplied to the notch filter, and a D / A converter is inserted before the sound system 54 to reconvert it into an analog signal. It ’s fine. In this case, the output of the sound source 44 may be used as digital data, and the mixer 52 may add the digital data.

Further, in each embodiment, when the determiner 30 determines that howling has occurred, the notch filter 50 is used to prevent howling by setting the peak frequency f1 to the notch frequency fn. You may comprise so that it may control so that a gain may be reduced uniformly. According to this configuration, howling can be suppressed only by the control for reducing the gain. That is, in the present invention, when it is determined that howling has occurred, it is sufficient that the configuration attenuates at least the component of the peak frequency f1.
On the other hand, the notch filter 50 is arranged at the front stage of the mixer 52, but may be arranged at the rear stage of the mixer 52.

  Further, in each embodiment, the guide melody data is used only for obtaining the target frequency f2. However, the guide melody data can be used for various other purposes. For example, the guide melody data may be supplied to the sound source 44 together with the performance data and used as a guide for singing a singer as a melody when singing, or synthesized by a voice processing unit provided separately. You may use it for the function to hear as a song. Further, it may be used for the function of scoring the singing by detecting how much the pitch and volume of the singer are shifted with reference to note-on data and velocity data included in the guide melody. Of course, it is needless to say that a configuration capable of appropriately setting on / off of these functions is desirable.

It is a block diagram which shows the structure of the karaoke apparatus to which the howling suppression apparatus which concerns on 1st Embodiment of this invention is applied. It is a figure which shows extraction of the peak frequency in the karaoke apparatus. It is a figure which shows the output of the target frequency in the karaoke apparatus. It is a figure which shows the frequency characteristic of the notch filter in the karaoke apparatus. It is a figure which shows the relationship between the peak frequency and the target frequency in the karaoke apparatus. It is a figure which shows the relationship between the peak frequency and target frequency in the howling suppression apparatus which concerns on 2nd Embodiment of this invention. It is a wave form diagram which shows the song signal at the time of howling generation | occurrence | production. It is a wave form diagram which shows a human voice. It is a figure which shows the principal part structure of the howling suppression apparatus which concerns on 3rd Embodiment of this invention. It is a figure for demonstrating operation | movement of the howling suppression apparatus.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Microphone, 20 ... Frequency analyzer, 30 ... Determinator, 42 ... Performance sequencer, 44 ... Sound source, 46 ... Converter, 50 ... Notch filter, 52 ... Mixer, 54 ... Sound system.

Claims (2)

Specific means for analyzing the input signal from the microphone and identifying its peak frequency;
Notification means for notifying the target frequency of the sound input to the microphone;
A determination unit that determines whether or not a relationship between a peak frequency specified by the specifying unit and a target frequency notified by the notification unit satisfies a predetermined condition;
Suppression means for suppressing and outputting at least a frequency component specified by the specifying means of the input signal when the determining means determines that the predetermined condition is satisfied,
The predetermined condition is that the peak frequency includes a frequency range including the target frequency, a frequency range including a frequency that is an integer multiple of the target frequency, or a frequency range including a frequency that is the integer multiple of the target frequency. A howling suppression device characterized in that it is not included in any of the above. Specific means for analyzing the input signal from the microphone and identifying its peak frequency;
A performance sequencer that outputs performance information and main melody information as the performance progresses;
Musical tone synthesis means for synthesizing musical tone signals based on the performance information;
Notification means for converting the main melody information into a target frequency of the voice input to the microphone and notifying it;
Determining means for determining whether or not a relationship between the frequency specified by the specifying means and the target frequency notified by the notifying means satisfies a predetermined condition;
Suppression means for suppressing and outputting at least a frequency component specified by the specifying means of the input signal when the determining means determines that the predetermined condition is satisfied ,
The predetermined condition is that the peak frequency includes a frequency range including the target frequency, a frequency range including a frequency that is an integer multiple of the target frequency, or a frequency range including a frequency that is the integer multiple of the target frequency. A karaoke apparatus characterized by being not included in any of the above .

Images