JPH03186006A - Automatic gain controller - Google Patents

Abstract

PURPOSE:To perform natural acoustic signal level adjustment by performing adaptive control over a threshold value for sound/mute classification by using a noise signal level at the time of muting. CONSTITUTION:When an input signal is a sound signal, a switch circuit 9 is connected to its ON side, and the output signal of a squaring circuit is inputted to a 1st averaging circuit 4 as sound data to calculate the sound level, which is used to generate a gain coefficient Ag by an AGC coefficient generating circuit 5. At this time, a 2nd averaging circuit 6 holds the output signal as it is. Further, when the input signal is a mute signal, the switch circuit 9 is placed on the OFF side and the output signal of the squaring circuit is inputted as mute data to the 2nd averaging circuit 6 to calculate the mute level. This value is inputted to a threshold control circuit 7 to adjust the threshold value. Consequently, the natural level adjustment is performed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an automatic gain control device for audio signal processing such as audio signal level adjustment in a telephone line.

[Conventional technology]

In recent years, there has been remarkable development in the communications field, and many new products have been created. For example, a TV conference system that connects multiple terminals and allows people who are far apart to hold a conference without having to move;
A prominent example is a voice mail system that allows voice messages to be left even when the person on the receiving end is not available, thereby saving as much time as possible.

The role these new inventions play in our busy urban societies is significant, and they are of great interest from all perspectives.

Recently, there has been a desire to realize more natural sound and lower usage fees, and new innovations in the transmission field are expected.

An automatic gain control device is one of the devices that help achieve this high quality audio. Automatic gain control (8ut, o-mati
c Ga1n Control (hereinafter abbreviated as AGC) refers to the function of keeping the output within a certain range even when the level range of the human signal is wide. It is also used in acoustic signal processing for adjustment and has greatly contributed to the development of the communications field. This AGC function has typically been performed by analog circuits using operational amplifiers. However, with the recent development of digital processing technology, it has become possible to perform this AGC function using a digital circuit.

Furthermore, many signal processing processor systems are now implementing digital AGO circuits, and AGC devices for more limited purposes are being developed.

FIG. 3 is an example of a system diagram showing an outline of the conventional digital type GG system disclosed in, for example, Japanese Patent Publication No. 19089/1989.

Analog human power A3 A in is converted into a digital signal X by an A/D converter (1). This digital signal X is then input to the AGC loop of the digital AGC system. Here, it enters the multiplier (2>) and is multiplied by the AGC coefficient Ag to become the AGC output signal y. Therefore, y=Ag-x. The digital output signal y is branched and sent to the square circuit (3). After passing through the averaging circuit (4), it enters the GG coefficient generation circuit (5).In this way, the GC coefficient A
g is determined each time the sample value is input manually and the digital output signal y goes around the AGC loop.

An example of a hardware implementation of the GG loop shown in FIG. 3 is as shown in FIG. 4.

The operation shown in FIG. 4 will be explained in detail below.

The input acoustic signal is converted into a digital signal by an A/D converter and then applied to this circuit. The digital input signal X is first input to a multiplier (2), where it is multiplied by a GG coefficient Ag to calculate an AGC output signal y. The digital output signal y is branched into the AGC loop, and first a squaring circuit (3I) is used to detect its level.
Furthermore, the adder (32) is manually inputted. However, the adder (
32) is input as a - (minus) input. A reference value Dr is separately applied to the adder (32), and when the digital output signal y exceeds Dr, it is used as a negative input, and conversely, when the digital output signal y is less than Dr, it is used as a positive input and is applied to the multiplier. (4I) and given a certain weight. Including this multiplier (41), the adder (42
) and the delay circuit T1 performs the averaging. Note that b for the multiplier (5) and 1 and 0 for the adder (52) are both related to weighting, and here the AGC coefficient A
Determine and generate g.

[Problem to be solved by the invention]

Some AGC loops in conventional digital AGC systems are constructed as shown in FIG. 4, and when this is attempted to be used for level adjustment of audio signals, the following problems arise. As mentioned earlier, in FIG. 4, the loop including the adder (42) and the delay circuit T constitutes an integrator, and when calculating the average value, the integrator adds all input values. However, since telephone lines are not always in a busy state, data is captured even in a silent state where there is no voice signal.

If this is averaged regularly, the average value will depend on the situation at the time of calculation and may become a value unrelated to the actual audio signal level. Therefore, a method has been devised in which a sound detection circuit is used to extract only sound data when calculating the sound signal level, an average value is obtained from the data, and a gain control value is determined by referring to this. Previous devices have the advantage of being simple in configuration; they set a threshold in advance to determine whether data is active, and if the human input signal is larger than this, it is averaged as active data. It has a calculation function that calculates the sound level from only the sound data, excluding it from averaging as silent data when it is small, and a processing function that holds the gain value given to the previous sample during silent periods. It is something that

However, with this method, if a change occurs in the input level, the threshold may no longer be appropriate for detecting speech, and accurate speech detection may not be performed. The following description is added to clarify what kind of results will be obtained when the threshold value is not set to an appropriate value for the input level. FIG. 2 shows the results of detecting the presence of sound using three types of thresholds of different sizes and calculating the average value only from samples determined to be presence of sound. For example, the waveform shown in (a) is the square value of a certain acoustic signal, which is the result of detecting sound presence using three types of thresholds of different sizes such as a, b, and c, and calculating the average. is (0), (c).

It becomes like (d).

In other words, since the threshold value a is too small for the human input signal shown in (a), normally the parts shown in (a), (i), and (tsu) are determined to be silent sections and the gain given to the previous sample is The value should be held, but this noise during silent periods is also taken in as voice data. As a result, the sound level is smaller than that in FIG. 2(d) using C as the threshold, and also fluctuates over a wide range. Furthermore, even though the threshold value is too large for the input signal shown in (a), all the input signals are determined to be silent noise, making it impossible to calculate the voice level. Therefore, as described above, in processing the silent section, the gain value given to the previous sample is held and given to these manual inputs. The dotted line d shown in (c) is the level calculated from the sample immediately before the silent section begins.

Out of a, b, and c, the voice presence level when C is detected is the most appropriate value for the input signal and is stable. As can be seen from these figures, if an inappropriate threshold is used, the calculated level will often differ from the original level, and the gain value calculated based on this will not be an appropriate value and will not be natural. It is not possible to make precise level adjustments.

This invention was made in order to solve the above-mentioned problems, and it sets a threshold value according to the size of input signal No. 3 so that the target acoustic data as an input signal undergoes natural level adjustment. To provide an automatic gain control device capable of automatically detecting the presence of sound by automatically changing it, calculating a stable level, and generating a gain commensurate with the human power level based on this.

[Means to solve the problem]

The automatic gain control device according to the present invention is equipped with a sound detection circuit and a threshold control circuit in a GG loop that calculates an input signal level and generates a gain coefficient value based on the input signal level. With this, when calculating the signal input level, the average value of only the signals that are valid from the sound detection circuit is calculated and used as the input level, and the gain value given to the previous sample is held during silent periods. This prevents the input level from fluctuating, and at the same time, applies the average value of the invalidated signal to the threshold control circuit to control the threshold.

(Function) The automatic gain control device of the present invention is based on the premise that the circuit for detecting signals such as voices extracts only data effective for calculating the sound level, and also sets a reference value for extracting the sound data. Since the threshold value is adjusted as needed according to the silence level averaged only from the noise data during breath sounds, the process of classifying the signal into sound and silence is highly reliable, and the calculated value is based on this. The sound level is always determined from only the sound of the input audio signal, and is not affected by noise in the silent state, so an output signal can be obtained that has undergone stable gain control commensurate with human power.

〔Example〕

FIG. 1 is a block diagram showing one embodiment of the present invention, (1), (2), (3), (4),
(5) is the same as the conventional device shown in FIG. Also, (6) is an averaging circuit similar to (4).

(7) is a threshold control circuit, (8) is a sound detection circuit, (9)
is a switch circuit that switches the circuit between when there is sound and when there is no sound.

Next, the operation will be explained in detail. The digital automatic gain control device according to the present invention, like the conventional digital automatic gain control device, uses an analog audio input signal Ai.
After converting n into a digital signal ,
The power is extracted by the square circuit (3).

In parallel with the processing in this loop, the input signal X is compared with the threshold value C output from the threshold value control circuit (7) to determine whether X is sound data or silent data.

The control operation of the threshold value control circuit will be described later.

When the human input signal is active in voice detection, the switch circuit (9) is connected to the ON side, and the output signal of the square circuit is connected to the first
The signal is inputted as sound data to the averaging circuit (4), and the sound level is calculated, and the result is used to generate a gain coefficient Ag in an AGC coefficient generation circuit. At this time, the second averaging circuit (6) holds the output signal as it is. Also, when the human input signal is silent, the switch circuit (9) is turned OFF.
It is connected to the FF side, and the output signal of the squaring circuit is input as silence data to a second averaging circuit (6) to calculate the silence level. This value is then input to the threshold control circuit (7) and the following processing is performed. In other words, if the silent data is smaller than the threshold, the threshold is considered to be higher than the human power level, so the threshold is lowered, and conversely, if it is larger, the threshold is considered to be smaller than the input level, and the value is increased. . The threshold value is controlled in this manner, and when a new value is generated, the output of the threshold value control circuit (7) is newly set as the threshold value C of the sound detection circuit (8). Furthermore, while a new sound level is not calculated, the output of the first averaging circuit (4) shall be held at the same value, and A
It is assumed that the GC coefficient generation circuit holds the gain coefficient Ag for the previous sound sample and continues to output that value.

Therefore, even if there are turning points or silences during the call,
By setting an appropriate threshold, it is possible to distinguish whether the manually generated signal is loud or silent, and in the circuit that calculates the sound level, the noise signal during silent periods is excluded from the averaging target and processed. In contrast, a circuit that calculates the silence level has the advantage of being able to process only silence data. Furthermore, what we would like to emphasize in the present invention is that the threshold value in the sound detector is controlled by the soundless level obtained by averaging the soundless data, so that the presence of sound can always be detected using a threshold that is adapted to the input level. . In other words, by changing the threshold, such as lowering the threshold when the silence level is smaller than the threshold multiplied by a certain ratio, and increasing the threshold when it is larger, it is possible to maintain noise even if the human signal level changes. This has the effect that it is possible to detect a voice with high reliability without determining it as a sound, or conversely, without determining a voice as noise, and to calculate an appropriate gain value.

In the above embodiment, a square circuit (
3) is adopted, but the same effect can be achieved by taking the absolute value instead.

Further, the automatic gain device of the present invention can also be performed by software processing centered on a processor.

In addition, in the above embodiment, the silence level is calculated from the value below the threshold and the threshold is controlled using this value, but instead of the silence level, the sound level is multiplied by a constant value smaller than 1. If the threshold value is controlled, the averaging circuit will be 1
It may also be possible to get away with it.

In addition, in the above embodiment, voice detection is performed using the input signal X of the AGC loop, but the signal y after automatic gain control is
Sound presence detection may be performed using the following.

〔Effect of the invention〕

As described above, according to the present invention, in a circuit that detects a sound signal such as voice, the threshold for classifying sound/non-speech is adaptively controlled using the noise signal level during silence, so that the original sound level can be adjusted. , it is now possible to calculate the silence level, and the gain calculated based on this can be made stable, resulting in the effect that more natural sound signal level adjustment can be performed.

[Brief explanation of drawings]

Figure 1 is a block diagram showing a schematic diagram of an embodiment of the present invention, Figures 2 (a) to (d) are examples to which the present invention is applied, and Figure 4 is a hardware diagram of a conventional example of the present invention. It is a block diagram shown in FIG. (2) is a multiplier, (3) is a squaring circuit, (4), (6
) is a flattened circuit, (9) is a switch circuit for switching circuits, and A. is an analog human signal, X is a digital human signal, and y is an output signal. In addition, in the figures, the same reference numerals indicate the same or corresponding parts.

Claims (1)

[Claims] In an automatic gain control device that automatically sets a signal gain coefficient according to the strength of the signal input level and outputs a signal at a constant level, the signal input level is compared with a preset threshold, and the effective interval of the signal is determined. A signal detection circuit that detects an invalid section, a calculation section that calculates the signal input level when detecting a valid section and an invalid section, and an automatic gain coefficient generator that sets a signal gain coefficient according to the input level of the signal in the valid section. An automatic gain control device comprising: a circuit; and a threshold control circuit that controls a threshold for determining the valid section and the invalid section according to the signal level of the invalid section.