RU2345477C1 - Automatic signal gain control method - Google Patents

Abstract

FIELD: physics, radio.

SUBSTANCE: invention refers to signal gain control devices in telecommunication equipment and specifically to VoIP voice gain control methods. Method contains the stages as followings: incomer measurements, simultaneous calculation of two, the first and second current gains based on measuring result, calculation of total gain by multiplication of the first and second current gains, and output gain control based on calculated total gain.

EFFECT: higher jump signal response speed, lower background noise, and smooth gain control.

12 cl, 6 dwg

Description

The technical field to which the invention relates.

The invention relates to devices for regulating the signal gain in telecommunication equipment and, in particular, to a method for adjusting the amplification of speech signals of subscribers of VoIP devices.

State of the art

Currently, many automatic gain control (AGC) systems have been developed that maintain the average value of the output value level.

So, for example, in patent US 5907622 A1, IPC H03G 3/20, publ. 05/25/1999 a sound frequency signal processing apparatus is disclosed, including: means for generating a desired volume and signaling a representation of a desired volume of said acoustic signal; interfering means to generate a representation of the noise signal of the background noise state in the listening space; variable gain means to receive an audio frequency signal and a gain control signal for amplifying said audio frequency signal in accordance with said gain control signal for generating an output signal; gain control means to generate a gain control signal in response to said desired volume signal and said noise signal, wherein a gain control signal representing a gain corresponding to an increasing value of a monotonic increase function of said background noise state; gain control means to generate a gain control signal in response to said gain control signal.

Another patent US 6731768 B1, IPC H04R 25/00, publ. 05/04/2004 reveals a hearing aid having a switchable automatic gain control. In one embodiment, the input signal is transmitted to a comparator, which controls the switch in an automatic gain control isolation circuit. While the instantaneous input level exceeds the threshold, the disconnect circuit allows automatic gain control. While the instantaneous input level does not exceed the threshold, the disconnect circuit is locked and automatic gain control supports setting the current gain. Since speech has a very high maximum peak value, this circuit will restore any necessary gain in the presence of speech (assuming an appropriate threshold), but will not restore the gain to background noise whose maximum peak value is below the threshold.

Application US 2002164039 A1, IPC H04R 3/00, publ. 11/07/2002 discloses a method for controlling the gain balance at a low frequency, including providing a microphone input with an audio signal to supply it to the input stage of the amplifier, calculating the dynamic instantaneous energy value of the audio signal in the first operating mode, implementing a predefined algorithm for the strength of the audio signal in the second mode using the energy value, the implementation of low-frequency gain in the communications device so that the low-frequency gain, in the first operational ohm mode, substantially approximates the low frequency gain in the second operating mode.

Patent US 5659582 A, IPC Н04В 1/06, publ. 08/19/1997, protecting an automatic gain control, including: means to detect the instantaneous power of a signal amplified or reduced in a controlled device; a control signal generator including: instant power filtering means, comprising: combining means producing an integrated power value, combining a value of a block representing said instantaneous power with respect to a predetermined threshold value; and smoothing means to reset the integrated power value to a predetermined value between a predetermined upper threshold and a predetermined lower threshold smaller than said upper threshold and control means to obtain a tendency for said instantaneous power to change based on the filtering result, repeatedly detecting that the integrated power value has reached an upper threshold or a lower threshold, wherein said control means controls the gain of ravlyaetsya device, whose power is transferred in accordance with the specified change trend obtained.

Currently, the main requirements for the AGC system of an IP phone are:

- The ability to reduce the level of acoustic echo arising from the design features of the telephone (the AGC system should not amplify the echo signal).

- The AGC system should reduce the level of the signal coming with overload (the subscriber screams into the phone).

- The AGC system should provide a value of the average level of the output signal at the level of - 19 dB, that is, increase the signal gain at a low level and reduce at a high level. Moreover, the change in the gain should be smooth, in order to avoid deterioration in the quality of communication due to the appearance of the effect of "ripple" of the signal (the case of a sharp change in gain).

- The AGC system should adjust the gain only if there is a speech signal, it should not amplify the background noise during pauses in speech.

Technical solutions, currently known from the prior art, completely do not meet the modern requirements for the AGC system of an IP phone, so the present invention was developed taking into account the above requirements.

SUMMARY OF THE INVENTION

The tasks are achieved using a method of automatically adjusting the signal gain, including measuring the level of the input signal, simultaneously calculating two, first and second current gains based on the measurement result, calculating the final gain by multiplying the first and second current gains and adjusting the output gain based on the resulting gain value.

Moreover, in one embodiment of the invention, the calculation of the first current gain is carried out by comparing the current level of the incoming signal with predetermined thresholds, if the level of the incoming signal exceeds the first preset threshold (limiter threshold), the first current gain is calculated by calculating the difference between the current level of the incoming the signal and the threshold, if the level of the incoming signal exceeds the second predetermined threshold (compressor threshold), but below the first predetermined threshold, and they use a signal compressor to calculate the first current gain by multiplying the current level of the incoming signal by the value of the slope of the compressor and, if the input signal is below a predetermined range (expander threshold), use a signal expander to calculate the first current gain by multiplying the current level of the input by the value of the slope of the expander.

In another embodiment, if the current input signal level is below the compressor threshold but above the expander threshold, the instantaneous gain is set to 0 dB.

In yet another embodiment of the invention, the calculation of the second current gain is carried out by calculating the difference between the current signal level and a predetermined signal level. In this case, the calculation of the second current gain is carried out for the input signal fragment, the vocalization of the input signal fragment is determined and the second current gain is calculated only if the input signal fragment is vocalized, and in the absence of the vocalization sign of the input signal fragment, the coefficient is used gain calculated for the previous voiced fragment of the input signal.

In another embodiment of the invention, the measurement of the level of the incoming signal is carried out in the logarithmic range.

In another embodiment of the invention, the calculation of the first and second current gain is carried out in the logarithmic range and translate into linear.

In another embodiment of the invention, the obtained first and second current amplification factors are smoothed to provide the required dynamic characteristics of both parts of the device.

In another embodiment of the invention, in the smoothing step, the gain is determined by the formula

g (n) = (1-k) g (n-1) + k f (n),

where k is equal to the coefficient of attack or release.

In yet another embodiment, all the steps of the method are performed on portions of the incoming signal.

In yet another embodiment, a predetermined range corresponds to a range of -10 to -30 dB.

In yet another embodiment, the step of calculating the instantaneous gain value based on the measurement result further includes the step of determining whether the input signal level matches a predetermined peak value, and if the input signal level value exceeds a predetermined peak value, the instantaneous gain value is calculated by inverting the excess value signal level of a predetermined peak value.

A brief description of the drawings.

Figure 1 General block diagram of the AGC.

Figure 2 Scheme of the level meter of the speech signal.

Figure 3 Graphs of the dependence of the output level and gain on the value of the input level.

Figure 4 The scheme of the meter peak value of the signal.

Figure 5 The algorithm of the gain control.

6 is a diagram of the smoothing unit.

DETAILED DESCRIPTION OF THE INVENTION

The block diagram of the AGC system is presented in figure 1. The input speech signal is input in fragments of 80 16-bit samples with a sampling frequency of 8 kHz (10 ms).

The measurement of the level of the speech signal in block 1 of the meter level of the speech signal. The measurement result, presented in the logarithmic range, is used in the gain controller 2 to calculate the instantaneous gain value. The obtained value of the instantaneous gain is processed in block 3 smoothing. The output unit 3 smoothing gives the current value of the gain in the linear range. Block 3 smoothing the gain sets the dynamic characteristics of the AGC system.

Speech level meter block

Block 1 of the level meter of the speech signal is presented in figure 2. Using the proposed scheme, the root-mean-square value of the signal energy is measured by squaring the input signal and averaging it when passing through a first-order low-pass filter. By changing the value of the TAV coefficient, the rate of change of the root-mean-square energy value at the meter output is adjusted, and the coefficient value must be less than 1. As the coefficient grows, the speed of the meter reaction to a change in the input signal increases. Thereby, the measurement of the average signal level value over a certain period of time is achieved.

At the filter output, the obtained value is used to calculate the logarithm of base 2. The use of the logarithmic scale when measuring the signal level is due to the physiological feature of the human hearing aid. A logarithmic measure of the volume level of a signal very well reflects the human perception of sound. The use of the base 2 logarithm is due to the simplicity of the integer calculation algorithm.

After calculating the logarithm of the root mean square value of the signal level, the obtained value is multiplied by a factor of 0.5, which in the linear scale corresponds to the square root extraction operation.

The meter calculates according to the following formula:

x _RMS (n) = (1-TAV) x _RMS (n-1) + TAV x ² (n).

Gain adjustment

The algorithm of the gain control 2 is shown in Fig.5

The dynamic range of the volume of the speech signal can be divided into three sub-bands:

- Very loud sound annoying the caller: speech turning into a scream, loud background noise. The threshold value for this range is 0-10 dB.

- Normal speech: -10 ÷ -30 dB.

- Background noise, acoustic echo: less than -30 dB.

Thus, the gain control 2 (KU) can be divided into two independent blocks: the KU regulator of the normal speech range and the signal regulator outside this range.

KU controller for level values outside the normal speech range

When the value of the input signal, the level of which lies in the first range (more than -10 dB), to increase comfort, it is necessary to reduce the signal level. For this purpose, a compressor and signal limiter are used. The compressor converts large changes in the input signal into small changes in the output, and the limiter limits the maximum output level.

In the case when the signal level value lies in the third range (less than -30 dB), a signal expander is used to reduce the level of background noise and signal echo, which increases the value of small level changes.

In both cases of using both the compressor and the expander, the input gain is less than 1, improving the quality of the speech signal by reducing the level of very loud sounds, reducing the echo level and reducing the background noise level. Graphical dependence of the level of the output signal and gain on the level of the input signal is presented in Fig.3.

The value of the current level is compared with the thresholds. If the level value corresponds to one of the ranges (expander, compressor), a new instantaneous gain is calculated by multiplying the current level by the value of the slope of the expander or compressor, respectively. If the level value is outside the ranges of the compressor or expander, then the gain value is set equal to 0 dB. For the limiter to work, a peak signal level is measured. The circuit of the meter peak signal level is presented in figure 4. If the peak signal level is higher than the limiter threshold (LT) in the logarithmic scale, in this case the gain in the logarithmic scale is calculated by inverting the value of the signal exceeding the limiter threshold level.

KU control at level values corresponding to the normal speech range

When the signal values are in the range of -10 ÷ -30 dB corresponding to the range of the normal speech signal, the AGC system must adjust the gain to achieve the required average speech level of about -19 dB.

The key point when adjusting the signal level in this range is that the AGC system should adjust the gain only in the presence of a speech signal, and in the absence of speech, the adjustment should stop. To determine the moments of speech in the input signal, many algorithms have been developed, the so-called VAD (Voice Activity Detector), which are often used in packet telephony. For example, VAD as recommended by ITU-T G.729B.

Using VAD according to G.729B recommendation determines the block algorithm of the regulator operation with a processing period of 10 ms. Using other algorithms, the processing period will be different.

In case of detecting a VAD voiced frame, a new instantaneous gain value is calculated, which is the difference between the current signal level and the desired (-19 dB).

Using the lower threshold at the input of the regulator of the normal speech range allows you to distance yourself from the acoustic echo at the input of the regulator. Since VAD is designed to analyze and decide on the presence of speech in the background of noise in a signal with a different level, including a level below -30 dB, therefore, VAD will give a sign of a voiced frame even in the presence of an acoustic echo signal in the processed signal.

The values of the instantaneous current gain from both parts of the controller are transferred from the logarithmic to the linear range by raising the two to the power of the instantaneous value of the gain. The obtained value of the gain is transmitted to the smoothing blocks to provide the necessary dynamic characteristics of the controller.

The use of two independent smoothing units for both parts of the regulator allows you to set different dynamic characteristics to meet specified requirements.

Gain smoothing unit 3

The block circuit 3 smoothing gain is presented in Fig.6. The dynamic characteristics of the control system are set using attack and release factors. When the gain is reduced, the attack coefficient is applied, and when increased, the release coefficient is applied.

The value of the new gain at the input of the block is compared with the previous value and after applying a small hysteresis loop, the coefficient currently used (attack or release) is determined. The use of a hysteresis loop guarantees the stability of the system at small deviations of the gain.

After selecting the current coefficient, the control signal is smoothed:

g (n) = (1-k) g (n-1) + k f (n),

where k is equal to the coefficient of attack or release.

The amplification factors obtained at the output of the smoothing blocks from both parts of the controller are multiplied, thereby ensuring the smoothness of the transient processes of the AGC system. The resulting common KU is used to scale the input signal.

The present invention allows to reduce the level of acoustic echo resulting from the design features of a VoIP phone and background noise due to the use of an expander. By using VAD, which is used as a signaling device for the presence of speech to activate the control process of the control unit, and not to turn off the path, a detuning from loud noises is achieved with voice-activated adjustment of the average signal level. The measuring / compressor / expander circuit allows you to adjust both very weak signals and very loud ones when adjusting the average signal level. The proposed invention allows to increase the response rate to a sharply changing signal level (a sharp transition from a quiet conversation to a very loud one), to reduce the level of background noise, to ensure smooth control of the gain of the speech signal (the effect of a pulsating sound in speech pauses is eliminated).

Claims (12)

1. A method of automatically adjusting the signal gain, comprising the following steps: measuring the level of the input signal, simultaneously calculating two, first and second current gains based on the measurement result, calculating the final gain by multiplying the first and second current gains, and adjusting the output gain based on the resulting gain value. 2. The method according to claim 1, characterized in that the calculation of the first current gain is carried out by comparing the current level of the incoming signal with predetermined thresholds, in this case, if the level of the incoming signal exceeds the first predetermined threshold (threshold limiter), the first current gain is calculated by calculating the difference between the current level of the incoming signal and the threshold, if the level of the incoming signal exceeds the second preset threshold (compressor threshold), but below the first preset threshold, using a signal compressor for calculating the first current gain by multiplying the current input signal level by the compressor slope value, and if the input signal level is below a predetermined range (expander threshold), a signal expander is used to calculate the first current gain by multiplying the current input level by expander slope value. 3. The method according to claim 2, characterized in that if the current level of the incoming signal is below the compressor threshold, but above the expander threshold, the first current gain is set to 0 dB. 4. The method according to claim 1, characterized in that the calculation of the second current gain is carried out by calculating the difference between the current signal level and a predetermined signal level. 5. The method according to claim 4, characterized in that the calculation of the second current gain is carried out for a fragment of the input signal, while determining the vocalization of the fragment of the input signal and the calculation of the second current gain is carried out only if there is a sign of vocalization of the fragment of the input signal, in this case the absence of a sign of vocalization of the fragment of the input signal uses the gain calculated for the previous voiced fragment of the input signal. 6. The method according to claim 1, characterized in that the measurement of the level of the incoming signal is carried out in the logarithmic range. 7. The method according to claim 1, characterized in that the calculation of the first and second current gain is carried out in the logarithmic range and translate into linear. 8. The method according to claim 1, characterized in that the obtained first and second current amplification factors are smoothed to provide the required dynamic characteristics of both parts of the device. 9. The method according to claim 1, characterized in that at the stage of smoothing, the gain is determined by the formula
g (n) = (lk) g (n-1) + k f (n),
where k is equal to the coefficient of attack or release. 10. The method according to claim 1, characterized in that all the actions of the method are carried out on portions of the incoming signal. 11. The method according to claim 2, characterized in that a predetermined range corresponds to a range from -10 to -30 dB. 12. The method according to claim 1, characterized in that the step of calculating the instantaneous value of the gain based on the measurement result further includes the step of determining whether the level of the input signal matches a predetermined peak value, and if the value of the level of the incoming signal exceeds a predetermined peak value, the instantaneous value of the gain calculated by inverting the value of the excess signal level of a predetermined peak value.

Images