JP3365113B2 - Audio level control device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice level control device suitable for use in a voice recording device.

[0002]

2. Description of the Related Art Generally, a voice recording apparatus has a problem that if a large input voice that exceeds the recordable dynamic range of this apparatus is recorded as it is, the quality of reproduced voice is significantly deteriorated. When a voice is input, the gain of the voice amplification circuit is promptly reduced according to this input level, and when the input level becomes lower thereafter, the built-in voice level control device operates to restore the gain to the original value. ing.

In such a voice level control device, when the recording target is a telephone conversation or a conversation of a person in a conference or the like, the return speed for returning the gain to the original value is also set to a large value. By doing so, the voice to be recorded can always be recorded in a fixed size. However, if the recording target is a sound source in which, for example, a plosive sound such as applause and a low-level steady sound such as river murmuring are mixed, it is necessary to set the return speed to a large value as described above. There is a problem in that the sound of river murmuring becomes an unnatural sound pickup in which the fluffy level changes depending on the plosive sound.

On the other hand, if the return speed is set to a small value, such a problem will not occur, but, for example, the recording target is a river murmuring and a fireworks-like image is recorded during the recording. When the plosive sound occurs only once, the gain is immediately lowered to a level according to the plosive sound, and the sound of the river murmuring cannot be recorded in a sufficient volume. Moreover, since the recovery speed is slow, the plosive sound is generated. There is a problem that it takes time for the gain to return to the level before the occurrence.

Therefore, in the past, the analog voice level control device has adopted the control by the double time constant in order to solve the above problem of the return speed. This speeds up the level control tracking capability for a momentary loud volume, and slows tracking for a large energy sound that is steady or continues for a certain period of time. Two analog integrating circuits with different time constants are used in combination with the voice detection circuit for generating the.

[0006]

By the way, in the voice level control device using such a double time constant circuit, since only two integration time constants can be selected, it is not possible to finely follow the situation of the input voice. It is possible. Further, since the integration time constant is set to be several tens of seconds, it is necessary to improve the calculation accuracy when it is configured by a digital circuit.
It has the drawback of complicated processing.

[0007]

A voice level control apparatus according to the present invention comprises means for deriving absolute value data regarding a digital voice signal, and means for deriving average value data of a predetermined number of samples for the absolute value data. A means for deriving envelope detection data for the average value data, a means for setting a value of the predetermined number of samples, a means for deriving the envelope detection data for the absolute value data, and the absolute value data. Of the envelope detection data of the envelope detection data and the average value data, a means for setting the gain of the audio signal amplifying means based on the envelope detection data having the larger value. To do.

In this case, the means for setting the value of the predetermined number of samples sets the value of the predetermined number of samples to be higher as the level of the envelope detection data is higher based on the envelope detection data for the average value data. It is preferable to configure it, and when controlling the audio levels of multiple channels, use the largest absolute value of the absolute values of the digital audio signals of each of the multiple channels as absolute value data. It is desirable to do.

[0009]

The gain control is performed according to the level of the sound source, and the recovery time until the gain lowered by the voice of the large input returns to the original gain is adaptively controlled according to the level of the sound source. .

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A voice level control device according to the present invention will be described below.
Examples will be described with reference to the drawings. FIG. 1 is a diagram showing an audio signal processing stage provided with such an audio level control device, and this audio signal processing stage samples each analog input audio signal of channel 1 and channel 2 at 48 KHz for 20.8 μs. AD conversion circuits 31 and 32 for outputting digital audio signals of a cycle, and coefficient units 33 and 34 functioning as variable amplifiers for the digital audio signals A1 and A2 output from these circuits.
And a detector block surrounded by a dotted line for generating a coefficient k to be supplied to each of the coefficient units 33 and 34 on the basis of each output of the coefficient units 33 and 34.

Explaining the operation of generating the coefficient k in the detector block, the outputs A3 and A4 of the coefficient units 33 and 34 are first input to the absolute value conversion processing units 35 and 36, and their absolute values A7 and A6 are obtained. Taken out. These absolute values are input to the comparison gate unit 38,
Here, only the larger absolute value (A8) of the absolute values A7 and A6 is output. The absolute value signal A8 output from the comparison gate unit 38 is input to the averaging processing unit 39 and the envelope detection unit 40.

The processing in the envelope detecting section 40 will be described with reference to FIG. 2. (1) in FIG.
The input A8 to the envelope detection unit 40 is represented, and (2) represents the output A12 of the envelope detection unit. Here, the tracking characteristic of the output A12 with respect to the input A8 is set to about 20 ms for the rising characteristic and about 20 ms to 100 ms for the falling characteristic with respect to the input change of 40 dB corresponding to the dynamic range of the audio signal. A8 is
As shown in this figure, it is obtained as a signal that faithfully follows the input A8 with a small time constant.

Further, the input A8 is input to the averaging processing section 39, where it is subjected to averaging processing according to the sample number N set in the averaging sample number setting section 37. The number of samples N is set based on the output A11 obtained by envelope detection of the output A9 of the averaging processing unit 39 by the envelope detection unit 41. Next, these processing units 3
The processing in 9, 41 and 37 will be described with reference to FIG.

With respect to the input A8 shown in (1) of this figure, the averaging processing unit 39 produces an averaging output A9 according to the following arithmetic expression based on the sample number N given from the averaging sample number setting unit. Derive. A9 = (a1 + a2 + a3 + ... + aN) / N where a1, a2, a3, ..., aN represent N audio data of the input A8 that are continuously input. Further, the value of N is designed to be a value of about 320 when a normal level voice is input, and as a result, the averaged output A9 is shown in (2) of this figure. As described above, data having a period of about 6.6 ms (= 320 × 20.8 μs) is obtained.

Next, the averaged output A9 is input to the envelope detection section 41 to be converted into data of a period of 20.8 μs, and at the same time, predetermined rising and falling characteristics are given. The rising characteristic given here is set so as to be able to follow an input change of 40 dB for up to about 20 ms, like the rising characteristic in the envelope detection unit 40, and the falling characteristic is several seconds to several tens. Set it to an extremely large time constant such as seconds. By performing the above processing, the output A11 as shown in (3) of this figure is obtained as the output of the envelope detection unit 41 (note that the signal diagrams shown in these (1) to (3) are shown). Is a simplification, and the number of samples, the level of each sample data, etc. are not represented accurately).

Based on the level of the output A11 thus obtained, the number of samples to be given to the averaging processing unit 39 is set in the averaging sample number setting unit 37. In this case, the level of A11 is large. The larger the number of samples is set. Then, what characteristic the N is specifically set for the level of A11 may be determined according to the control characteristic of the required audio level control. For example, it is shown in (4) of this figure. like,,
Various characteristics such as, can be adopted.

The envelope detection output A11 generated as described above is input to the comparison gate unit 43. In the comparison gate section, the signal is compared with the envelope detection output A12 described above, and the signal with the higher level of both inputs A11 and A12 is taken out as the output A13. This output A13
Is input to the coefficient generator 42, and a coefficient k having a smaller value is generated as the level of A13 is higher. Then, the feedback control is executed by supplying the coefficient k to the coefficient units 33 and 34, and the digital audio signals A3 and A4 amplified to a desired level are obtained.

In this case, it is desirable that the amplification characteristic between the input and output of each coefficient unit is designed so as to have a characteristic as shown in FIG. It should be noted that, in this characteristic diagram, the output level of the input whose level is P or higher is suppressed to L, and it is necessary to set the output level L to a value close to the maximum dynamic range of the S / N. Desirable in terms of aspect. Further, in the case of an audio recording device in which the standard output level is set, it is desirable to set it to a value slightly exceeding the standard output level.

Although the embodiment described above records two channels of audio, the present invention can be applied to an apparatus for recording audio of any other number of channels. For example, the comparison gate unit 38 in FIG. 1 is not necessary in the case of a device that records only one channel of sound, and in a device that records a plurality of channels of sound, absolute value conversion processing for only that number of channels is performed. Units are provided and all the absolute value outputs thereof are input to the comparison gate unit 38.

Next, the control characteristics of the audio level control device shown in FIG. 1 will be described by taking a specific case as an example. (1) to (4) shown in FIG. 5 are signals A8, A12, A11, and A1 when a high-level plosive sound is generated only once while a low-level stationary sound is recorded.
3 shows each envelope waveform of No. 3 of FIG. As shown in this figure, the output A12 obtained by envelope detection of A8 in the processing unit 40 of FIG. 1 has substantially the same envelope waveform as that of A8, whereas A8 is a processing unit 39 in FIG. The output A11 obtained by the averaging and envelope detection in 41 and 41 has a waveform that rises slightly after the burst sound is generated and then gently falls.

As a result, the output A13 obtained by inputting A11 and A12 to the comparison gate 43 has an envelope waveform as shown in (4) of this figure. Therefore, the value of the coefficient k generated based on this A13 is rapidly reduced during the plosive sound generation period, the plosive sound is suppressed, and after the period elapses, it returns to almost the original value, so that it is at a low level. It will be possible to record enough steady sound again.

The waveforms shown in FIG. 6 are the signals A8, A12, A11, when a high-level plosive sound is repeatedly generated for a predetermined period while a low-level stationary sound is recorded.
And A13 represent envelope waveforms. In the case of this figure, the output A11 has a gradual fall characteristic, so that the level rises each time a plosive sound is generated, and has a waveform that gently falls after the elapse of the repeated plosive sound generation period. Will be things.

As a result, the output A13 obtained by inputting A11 and A12 to the comparison gate 43 has an envelope waveform as shown in (4) of this figure. That is, in the case where the plosive sound is repeatedly generated in this manner, the level of A13 in the period from the end of one plosive sound to the occurrence of the next plosive sound is gradually increased. The low level steady sound recording level is gradually maintained at a low level, and the steady sound recording level does not fluctuate as in the past.

Further, the higher the level of A11 is, the larger the number of samples for averaging when A11 is generated based on A8 is set. Therefore, the falling characteristic of the waveform of A11 is the level of A11. The higher the, the more gradual. That is, the smaller the number of times the plosive sound is repeated, the more quickly the waveform of A13 falls after the repeated occurrence period of the plosive sound is completed, and the lower level stationary sound is recorded accordingly.

In the detector block in the above-described audio level control device, all the internal processing units can be configured by hardware, but it is also possible to use a configuration using microcomputer control. Therefore, next, a configuration example in which the function of the detector block is realized by the microcomputer will be described with reference to FIGS. 7 to 12.

FIG. 7 shows the structure of the detector block when a microcomputer is used. The signals A3 and A4 are input to the microcomputer 55 via the interface 53, and the coefficient k derived in the microcomputer based on these inputs. Is output via the interface 53. In the RAM 56 in the microcomputer, a register Ra for storing the output A8, a register Rh for storing the output A12, and a register R for storing the output A9 are provided.
m, a register Ro for storing the output A11, a register Rn for storing the averaged sample number N, and a register Rc for storing the number of times of inputting audio data to the microcomputer 55, and the inside of the ROM 57. In addition to the normal microcomputer program, "A11 → N" for extracting the averaged sample number N according to the level of the output A11
A conversion table Tn and an “A13 → k” conversion table Tk for setting the coefficient k based on the level of the output A13 are provided.

Next, a flow of generating the coefficient k in the microcomputer 55 will be described with reference to FIG. In the figure, first, by executing steps S1 and S2, default values are stored in the registers Rn, Ro, and Rh, respectively, and the contents of the registers Rc and Rm are cleared.
Next, in step S3, the process waits until the voice data A3 and A4 are input. When the voice data is input, the absolute value data A8 having the larger value among the absolute value data of the input voice data A3 and A4 is registered in the register Ra. (Step S4).

In the next step S5, a value obtained by dividing the data A8 in the register Ra by the averaged sample number stored in the register Rn is added to the data stored in the register Rm, and the addition output is stored in the register Rm. While storing, the content of the register Rc is incremented by "1". Next, in step S6T, the envelope detection data A12 relating to the data A8 in the register Ra is derived and stored in the register Rh.
Then, in step S7, it is determined whether the value of the input count data in the register Rc is equal to the averaged sample count in the register Rn.

If the result of this determination is NO, the register R
Data A12 in h and data (A1 in register Ro
1) is compared, the coefficient k corresponding to the value of the larger data is read from the table Tk and output to the coefficient units 33 and 34 (steps S8 to S10), and then the process returns to step S3. When the determination result in step S7 is YES due to the repeated execution of the loop including steps S3 to S10, it means that the averaged data A9 for the number of averaged samples to be obtained is stored in the register Rm. Then, the envelope detection data A11 relating to this data A9 is derived, and the data in the register Ro is updated to this newly derived data A11 (steps S7 and S11).

Next, the number of averaging samples to be set corresponding to this data A11 is read from the table Tn and stored in the register Rn to prepare for the next averaging calculation process (step S12), and then the register The data A12 in Rh is compared with the data (A11) in the register Ro, and the coefficient k corresponding to the value of the larger data is read from the table Tk and output to the coefficient units 33 and 34 (steps S13 to S15). And then step S2
Return to.

The specific flow of step S4 in this figure is configured as shown in FIG. The specific flow of step S6 is configured as shown in FIG. Explaining the flow of FIG. 10, the absolute value data A8 (contents of the register Ra) extracted from the input voice data is the envelope detection data (register Rh) one sample before.
When the values are not equal, the value of the envelope detection data (A12) in the register Rh is not updated, but when the values are not equal, the value of the data A12 in the register Rh is updated.

That is, when the value of the data A8 in the register Ra is greater than the value of the data in the register Rh, the data A
The value obtained by multiplying 8 by α is stored in the register Rh (step S2
2) In the case of the opposite magnitude relationship, the numerical value obtained by subtracting the value β from the data in the register Rh is stored in the register Rh (step S23). Here, the value of α determines the rising characteristic of the envelope detection data, and the value of β determines the falling characteristic, and these values are such that the rising characteristic and the falling characteristic described in FIG. 2 are obtained. Is set to.

The concrete flow of envelope detection in step S11 is executed as shown in FIG. 11 as in the case of step S6. The value of γ in step S27 and the value of δ in step S28 in FIG.
As in the case of 0, the rising characteristic and the falling characteristic of the envelope detection are respectively determined, and these values are set so that the rising characteristic and the falling characteristic described in FIG. 3 can be obtained.

The arithmetic expressions for determining the envelope detection data values at the rising and falling edges used in the envelope detection flow of FIGS. 10 and 11 are step S22 and step S23, step S27 and step S28. In addition to what is shown in FIG.
2 and step S23, FIG.
Although various types can be considered as in (iii), in this embodiment, the arithmetic expressions shown in FIGS. 10 and 11 which have the best auditory quality in the experiment are adopted.

As described above, in the present invention, by adaptively controlling the followability according to the level of the sound source, a gain control which is natural to the listener when performing recording of a sound source having a wide dynamic range is performed. Moreover, when this is controlled by a microcomputer, it can be realized by a simple algorithm. In addition, as an audio recording device to which the present invention can be applied,
Various devices can be used regardless of whether analog recording or digital recording is used, but it is needless to say that the present invention can also be applied to a device for recording audio with an image such as a VTR or a video disc.

[0036]

When recording a sound source having a wide dynamic range, a gain control that is natural to the sense of hearing is performed. When configured using microcomputer control, it can be realized with a simple algorithm.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an exemplary embodiment of the present invention.

FIG. 2 is a diagram illustrating a process of an envelope detection unit 40 according to the same embodiment.

FIG. 3 is a diagram illustrating the processing of an averaging processing unit 39, an envelope detection unit 41, and an averaging sample number setting unit 37 in the same embodiment.

FIG. 4 is a diagram illustrating amplification characteristics in the example.

FIG. 5 is a diagram illustrating control characteristics when a plosive sound is generated only once in the embodiment.

FIG. 6 is a diagram illustrating control characteristics when a plosive sound is continuously generated a plurality of times in the embodiment.

FIG. 7 is an explanatory diagram of a case where the control unit of the embodiment is configured by a microcomputer.

FIG. 8 is a diagram showing an operation flow of the same microcomputer.

9 is a flow for realizing the processing of the absolute value conversion processing units 35 and 36 and the comparison gate unit 38 of FIG. 1 by the same microcomputer.

FIG. 10 is a flow for realizing the processing of the envelope detection unit 40 by the same microcomputer.

FIG. 11 is a flow for realizing the processing of the envelope detection unit 41 by the same microcomputer.

FIG. 12 is an example of another arithmetic expression that can be adopted when executing envelope detection processing by a microcomputer.

[Explanation of symbols]

33, 34 ... Coefficient unit, 35, 36 ... Absolute value processing unit, 38, 43 ... Comparison gate unit, 37 ... Averaging sample number setting unit, 39 ... Averaging processing unit, 40, 4
1 ... Envelope detection unit, 42 ... Coefficient generation unit,

Claims (3)

(57) [Claims] 1. A method of deriving (1) absolute value data relating to a digital audio signal, (2) a means of deriving average value data of a predetermined number of samples for the absolute value data, and (3) the average value. Means for deriving envelope detection data for data, (4) means for setting a value of the predetermined number of samples, (5) means for deriving envelope detection data for the absolute value data, (6) Of the envelope detection data for the absolute value data and the envelope detection data for the average value data, a means for setting the gain of the audio signal amplifying means based on the envelope detection data having the larger value. A voice level control device characterized by the above. 2. The means for setting the value of the predetermined number of samples comprises:
The audio level control device according to claim 1, wherein the value of the predetermined number of samples is set higher as the level of the envelope detection data is higher based on the envelope detection data for the average value data. 3. The audio level control according to claim 1, wherein the absolute value data regarding the digital audio signal is the absolute value having the largest value among the absolute values of the digital audio signals of a plurality of channels. apparatus.