JPH0311139B2 - - Google Patents

Abstract

In a digital speech detector, a first detector (601) produces an output signal if any sample of a voice channel signal has a magnitude (T) exceeding a fixed threshold (TF) above noise levels. A second detector (602) produces a second threshold (TL) which is adaptively adjusted to noise levels on the channel by being set to a level above the current sample magnitude (T) each time that this does not exceed the preceding sample's magnitude (TP). When the sample magnitude (T) rises above the second threshold (TL), and for each subsequent sample of successively increasing magnitude, the second detector (602) produces an output signal. A speech decision for the voice channel is reached if either detector (601, 602) produces its output signal. The output signals of the first and second detectors (601,602) are maintained for fixed and variable, respectively, hangover periods to maintain the speech decision during intersyllabic pauses in speech. The samples are supplied to the speech detector by an offset remover and averaging circuit.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a method for detecting the presence of an audio signal in an audio channel signal.

BACKGROUND OF THE INVENTION Methods for detecting the presence of a voice signal in a voice channel signal are used in various voice systems in which a voice path is secured in response to the detection of voice activity in a voice channel. ing. One such system can utilize the method of the present invention. Patent application corresponding to Canadian Patent Application No. 356965 "Relaxation of Signal/Noise Contrast in Digital Speech Interpolation Transmission System" filed on September 9, 1980 (Japanese Patent Application Laid-Open No.
No. 57-80845).

The voice detection method must be very sensitive to the presence of voice signals, but at the same time maintain low sensitivity to idle signals such as noise. In particular, problems arise in quickly and accurately distinguishing between low-level audio signals and noise. For example, in a DSI transmission system, the voice detection method must be able to detect low-level voice signals to avoid excessive clipping of the voice signal at the beginning of a transmission; DSI even at relatively high levels
should not respond to noise even at relatively high levels, as this would increase the activity of the transmission due to

Various forms of audio detection devices have been devised to more reliably discriminate between audio signals and noise. For example, published on April 15, 1975,
Fariello, US Pat. No. 3,878,337, discloses an arrangement in which a predetermined sequence of symbols of a series of samples of an audio channel signal is detected to provide an indication of a call. La- issued on June 7, 1977
Marche et al., U.S. Pat. No. 4,028,496, discloses an arrangement in which detection sensitivity and noise rejection are improved by integrating weighted differences between samples of a signal and its short-term operating average. . Furthermore,
U.S. Patent No. Vagliani et al., issued November 8, 1977.
No. 4057690 is for distinguishing between audio signals and noise.
An arrangement is disclosed in which segments of the envelope of an audio channel signal are compared to each other over different time domains.

However, these configurations do not fully meet the requirements of the voice detection device in the DSI transmission system to discriminate between voice and low level of noise, and to avoid clipping of the voice signal when vocalizing the voice, and therefore this There remains a need for improved speech detection apparatus and methods that meet such requirements.

It is therefore an object of the present invention to provide an improved method of detecting the presence of an audio signal in a sampled audio channel signal.

Means for Solving the Problems According to the invention, the problem as described above is achieved by: generating a first signal state whenever the absolute value of a sample of a signal exceeds a first threshold level; A method for detecting the presence of an audio signal in an audio channel means in which the absolute value of each sample is compared to the absolute value of the previous sample, and when the absolute value of a sample is not greater than the absolute value of the previous sample Whenever the absolute value of a sample is greater than the absolute value of the previous sample, change the level of the second threshold to a greater level depending on the absolute value of the current sample. generating a second signal state if the absolute value exceeds the current second threshold level; and if the first signal state or the second signal state is present, at least related to the current sample; The present invention is solved by providing a method comprising: indicating the presence of audio by

Operation The method of the present invention is carried out, for example, by an apparatus as shown in FIG. FIG. 1 shows a conceptual diagram of an example of an apparatus for carrying out the method of the present invention, and the details thereof can be found in the description of the embodiments of the present invention described with reference to FIGS. 2 to 5. It will become clear.

The device in FIG. 1 includes a level detector device 601;
A slope detector device 602 and an OR gate 603 are provided.

The absolute value T of the sample is input to the level detector device 601 .

A signal M is output from the level detection device 601, and when the signal is in the first signal state, M=1, and otherwise, M=0.

The level detection device 601 detects the absolute value T of the sample.
generates a first signal state (M=1) when TF exceeds a first threshold level TF. Since an audio signal has a higher level than noise, it is determined that it is an audio signal when the absolute value T of the sample exceeds the first threshold level TF.

The slope detection device 602 has an absolute value T of the sample.
is input.

The slope detection device 6021 outputs a signal K, and when the signal is in the second signal state, K=1; otherwise, K=0.

The slope detection device 602 calculates the absolute value T of each sample.
is compared to the absolute value TP of the previous sample.

Whenever the absolute value T of a sample is not greater than the absolute value TP of the previous sample, the second threshold TL is set to a greater level that depends on the absolute value T of the current sample.

Whenever the absolute value T of a sample is greater than the absolute value TP of the previous sample, if the absolute value T of the current sample is equal to the current second threshold level TL
, a second signal state (K=1) is generated.

When the sound is generated, the absolute value of the sample T
continues to increase in magnitude in successive samples. Therefore, if the absolute value T of the sample continues to increase, the second signal state is easily established, thereby detecting the occurrence of speech. If the absolute value of the sample increases or decreases, this is not the time when audio is occurring, making it difficult to reach the second signal state.

OR gate 603 receives signals M and K, and if at least one of them is 1,
A signal 1 is output, and the presence of audio is displayed according to this signal.

In this way, with this configuration, it is possible to quickly detect a voice signal when a voice is generated.

Speech indication for several samples following the last sample giving rise to the transmission indication so that the third signal indicating the presence of speech does not end during a short pause in the transmission during which a third signal indicating the presence of speech occurs between sybrals. so-called hangover to maintain
It is desirable that a period be provided. To this end, in a preferred embodiment of the invention, in response to a first signal, a fourth
and in response to the second signal generate a fifth signal in relation to a second number of consecutive samples starting with the current sample, the third signal being a fourth signal. or a fifth signal.

This second number of consecutive samples is such that the representation of the audio does not occur during long shelving periods in response to unwanted noise signals that led to the generation of the second signal. It is preferable that the reliability is varied depending on the reliability generated in relation to the . Therefore, in a preferred embodiment of the invention, the second
, wherein the second number is increased from a predetermined number to a maximum number for each sample in connection with which a second signal is generated; It is only reduced by a predetermined number for each sample whose size is not larger than the size of the previous sample.

In this way, the shelving period associated with the generation of the second signal is such that the successive increases in signal level in successive samples increase the reliability of the detection of the audio signal. The shelving period increases gradually. This first signal occurs only for relatively high signal levels at which the representation of the audio signal is very reliable.
The shelving period associated with the generation of the signal need not be changed.

The second threshold level is set to a relatively high value so that a second signal does not occur due to fluctuating signal levels, so that successive signal samples of magnitude less than the first threshold level are initially the second threshold level is set to a relatively high value so that the second signal is not generated, and then the second threshold level is set to a relatively high value so that the second signal is not generated, and then the second threshold level is set to a relatively high value so that the second signal is not generated. , a rise again to a value lower than the new relatively high second threshold level may occur. Preferably, this second signal also occurs in such a situation. Accordingly, in a preferred embodiment of the invention, it is further provided that whenever the magnitude of a sample exceeds the magnitude of a previous sample, a fifth signal is generated in relation to the previous signal, but a second
is not generated, and if the magnitude of the current sample does not exceed the second threshold level but does not exceed the third threshold level, then a second signal is generated in relation to the current sample, and this 2 is generated in relation to a previous sample, and whenever the current sample magnitude is not greater than the previous sample magnitude, set a third threshold level equal to the previous sample magnitude. do.

DC against unwanted noise signals and voice detection
In order to reduce the effect of deviations, each sample is preferably constituted by the average of a plurality of individual samples of the audio channel signal, and in a preferred embodiment of the invention, the DC deviation is producing each signal sample by subtracting from and averaging a number of individual samples. This averaging operation means that the audio judgment for each channel is
The above-mentioned patent application (Japanese Patent Application Laid-Open No.
57-80845))
This is particularly easily achieved in DSI systems.

Each step of the method of the invention can be implemented by individual components such as comparators, memories, gates, or by one or more programmed read-only memories (ROMs).

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be further understood by reference to the following description of preferred embodiments thereof, taken in conjunction with the drawings.

The voice detection device described with reference to FIGS. 2 and 4 is for use in a DSI transmission system of the type described in the patent application already mentioned (corresponding to Japanese Patent Application Laid-Open No. 57-80845). In , once the determination of a certain voice in each superframe is updated for each of a plurality of voice signal channels, the determination of a voice is updated for each of a plurality of voice signal channels individually in each of a plurality of frames forming this superframe. Includes samples. In this case, assume that in each superframe there are 27 frames, each containing 48 audio channel signals of 8 bits each.

In FIG. 2, which is shown in the form of a block diagram of a voice detection device, the voice detection device includes two parts, referred to in the text as a level detection device 601 and a slope detection device 602, which are combined at their output OR gate 603. Audio determination contents stored in a 48 channel determination storage device 604 are generated for each channel, and an audio determination output circuit 110 is coupled to the output side of this storage device. Detectors 601 and 602 are provided with a 7-bit average value generated by the circuitry described below with respect to FIG. It is enabled in the 14th frame of In its preferred embodiment, each detection device 601 and 602 consists of a ROM. This audio detection device is required to be able to detect low level audio signals in order to avoid excessive clipping of the audio signal when the audio is generated, but at the same time it is required to be able to detect low level audio signals in order to avoid excessive clipping of the audio signal when the audio is generated, but at the same time it is also required to They are required to be unresponsive even to relatively high levels of noise, as this may increase the noise level.
To meet such requirements, the voice detection device is configured to look for differences in noise and voice signal characteristics. These characteristics are: a) the speech signal typically has a higher level than the noise; and b) the noise is continuous, but the speech signal occurs in bursts, with the level of the signal increasing at the beginning of each burst. This is to increase the number of people. Detection device 601 with two voice detection devices
and 602 are included for this purpose.

Each detection device 601 and 602 has one of three states: a talking state, a shelved state, and a no talking state.
Classify each channel as being in For ease of verification, in Figures 3 and 4,
These states are indicated by an index M for the slope detector and an index K for the level detector, each index being 0 for the no talk condition and 1 for the talk condition. , and has a value of 2 for the shelved state. Thus, M=1 indicates that the level detection device indicates that the particular channel is carrying audio.

The shelved state is a temporary state in which a certain channel is considered to be immediately after the transmit state, and is provided to avoid clipping of the transmit after a pause between syllables in the transmit. For each detection operation, channels that were already indicated as being in the transmitting state, but for which audio is no longer detected in connection with them, are considered to be in the said shelved state, and an initial shelving count is set. . If it is still not detected in successive superframes, this shelving count is decremented until it reaches zero, at which time the channel is marked as idle. The initial shelving count is constant in the level sensing device, but can be changed in the slope sensing device, as explained further below.

Referring again to FIG. 2, the level detection device 601 consists of three parts: a comparator 605, a shelving and control device 606, and a decision storage device 607. Frame 14 in each superframe
In , for each channel, comparator 605 compares the average value T with a fixed threshold TF that is higher than the maximum noise level. The results of this comparison are provided to device 606. Device 606 determines the state of the channel according to the result of this comparison and the previous state of the channel as stored in memory 607 and stores in memory 607 the current state of the channel and the appropriate shelving count. If this channel is determined to be either in the transmit state or shelved, device 606 provides a logical 1 on output line 608.

The gradient detection device 602 includes a delay device 609
, a comparator 610 , a shelving, control and threshold generation device 611 , and a determination and threshold storage device 612
It consists of The delay device 609 is configured to delay the line 6 by one superframe with respect to the average value T.
13 to a comparator 610 with the previous average value TP. At frame 14 in each superframe, for each channel:
The comparator 610 compares the current average value T with the previous average value TP, the threshold value TL, and the threshold value TH,
This comparison result is provided to device 611. Thresholds TL and TH are various thresholds stored for individual channels in storage 612. The device 611 determines the state of the channel according to the comparison result and the previous state of the channel as stored in the storage device 612, generates new thresholds TL and TH as necessary, and generates new shelving values in the storage device 612. Store the current state of the channel along with the count and thresholds TL and TH. If the channel is determined to be in either the transmit or shelved state, the device 611 outputs the output line 614.
Give a theoretical value of 1 to the top.

Thus, if both the level detection device and the slope detection device do not indicate that the channel is in the idle state, i.e., M=0 and K=0, then the busy state on line 110 is determined. It will be seen that a determination exists for each channel, ie, the channel is considered to be carrying audio.

The operation of the call detection device will be further understood from the following description of FIGS. 3 and 4. In Figure 3, B, D and G are integers, H is the shelved count in the level detector, HM is the maximum value of H, C is the shelved count in the slope detector, and CM is the shelved count in the slope detector. is the maximum value of the value C; other symbols have the meanings already explained. In Figure 4, B=1, D=5, G=
4 and CM=31. Although Figures 3 and 4 each relate to one of the 48 channels, all channels are treated similarly. Figure 4 is 1
Showing the average value T for the channel as a book line,
Each point on this line indicates a value T in one superframe, and the resulting values M, H, TL,
K and C are shown. First, M=K=C=
Set to 0. Adjacent points on line 801 have reference number 8
02 to 834.

First, considering the action of the level detection device, for each of points 802 to 821, T>|
TF (Question 701 in Figure 3), and the value stored before M is 0 (Question 702 in Figure 3)
As a result, in FIG.
Then, M remains at 0 (ie, no call state). Looking at each of points 822 to 827,
T>TF, so that M is set to 1 (talking state) in block 704 of FIG. 3, regardless of the previously stored value M. For point 828, the result of question 701 is negative, so the value of M is questioned in block 702 of FIG. The previously stored value of M is 1, resulting in block 705 of FIG. 3, where M is set to 2 and H is set to HM=31. For each of points 829 to 834, the result of question 701 is negative and the value M stored before being questioned in block 702 is 2, so that in FIG. Ru. For these points, H≠0, so that H is decremented each time block 707 in FIG. 3 is entered and M remains unchanged. Again T is the value TF
, the question 706 has a positive result;
Therefore, this decrement state continues in successive superframes until H=0, which reaches block 708 where M is determined to be 0 (no call state).

Next, considering the operation of the gradient detection device, after reading out the value T in each superframe (block 709 in FIG. 3), this value is compared to the previous value.
It is compared with TP (Question 710 in Figure 3). If T>TP, as at points 803, 805 and 808 in FIG.
At step 11, a question is asked as to whether K=1 (sending state). For each of points 803, 805 and 808, the previous value K is 0;
As a result, the result of this question is negative. In a subsequent question 712, T is compared with a first threshold TL, and for each of points 803, 805, and 808, T>|TL, so that the question in the block is valid for whether K=0. . For each of these points, the result of this question is positive, so in block 714 the previous value C is increased by G=4 and K remains unchanged.

For each of points 804, 806, and 807, the result of question 710 is negative, so that in block 715 the threshold TL is set to BT+D, ie, T+5 in FIG. The previous value K is then queried in block 716 and since in each of these points the previous value K is 0, C is set to 0 in block 717 and K remains unchanged. In this way, all points 803 to 8
For 08, K=0 (no call state).
This threshold TL is adjusted accordingly during this period, so that this threshold is found to be approximately slightly higher than the level of noise present in a particular channel.

Regarding point 809, the question 710 has a positive result, the subsequent question 711 has a negative result, and the resulting question 712 has a positive result because T>TL at this time. Result Block 7 in Figure 3
In step 18, K is set to 1 (sending state).
Regarding each of points 810 to 813, question 71
0 and the resulting question 711 both have positive results. Thus, for each of points 809 to 813, C is increased by G=4 in block 719, and this gradual increase in C, such that the subsequent shelving period is the first transmission to reach point 809. Reflects increased reliability of status determination. In each case, C is compared to CM=1 at query 720, and for each of these points the result of this query is negative, so no further action is taken.

For point 814, T<TP, so that the first threshold TL is reset again to block 715. In this case, the previous value of K questioned in block 716 was 1, so that in block 721 the threshold TH of FIG. is set. Then, in block 722, G is decreased by 1 to 23. For point 815, T>TP, K≠1, T>TL and K≠0
As a result, the question T > TH? (block 723 in FIG. 3) is reached and the result is correct. Therefore, K is set to 1 in block 718 and C is incremented in block 719. This recognizes point 815 as containing a sending condition, and this recognition indicates that the lower level point 813 has already been identified as containing a sending condition, and therefore the relatively higher level point 815 also has a sending condition. It is based on the fact that it includes the state of speech.

Point 816 results in a determination of shelved status (K=2) in the same way as for point 814, and the threshold
TL and TH are reset and C is decreased by 1 to 26. At point 817, T>TP
As a result, the threshold TL is reset, leading to question 716, where it is found that K=2, so that in question 724, C is assessed, and since it is not zero, it is reduced in block 722.
For point 818, T>TP, K≠1, T>|
TL, K≠0 and T>|TH, so C
is queried in block 25, and since it is not 0, it is decreased by 1 in block 720, leaving K unchanged. Point 819 and points 820 to 82
5 is increased in block 719 for each of points 820 through 825. C is increased in block 720 and the question C>CM? has the exact result, so that for each of these points C
is set to CM=1 in block 727. At point 826, both thresholds TL and TH are reset as at points 814 and 818, resulting in a shelved determination (K=2), resulting in C being decreased by one. At the same time as for point 817, at each of points 827-834, the threshold TL is reset and C is decreased by one. If the point 801 does not cross the threshold TL or TH again, this decrease in C continues in successive superframes until C=0, when one of the questions 724 and 724 has a positive result, so that respectively In one of blocks 728 and 729, K is set to 0 (no call state).

Therefore, the level detection device 601 provides a reliable detection of the presence of voice each time the average value T exceeds a fixed threshold TF, and that for each such detection the determination of the presence of voice on the line 110 occurs within the superframe 32. It can be seen that the above determination is maintained over a fixed shelving period of 0 to maintain said determination during pauses between syllables in a send. On the other hand, the slope detection device 602 detects the start of a voice burst at a point 809 etc. with low reliability but very early, and determines the voice on the line 110 as much as possible.
Therefore, excessive clipping of the audio signal at the beginning of the audio burst is avoided. Because such determinations are unreliable, the shelving period of the slope sensing device is instead set immediately to a maximum value as in the level sensing device to avoid unduly increasing the activity of the DSI system's transmission operation. Therefore, it only increases gradually. For example, the average value of 8 at point 809 could alternatively be due to a noise transition instead of the start of a transmit speech, in which case line 801 would not be energized until after this point.
In this case, the gradient detection device reaches an inaccurate determination result K=1 (talking state) for point 809;
This decision will only be maintained for a short shelving period of 8 superframes so that the transmission activity of the DSI system increases slightly. In any case, as stated below, the number T is itself 1
is the average value obtained over the duration of two superframes, and the threshold TL is adjusted accordingly to be greater than or equal to the average noise level of the channel so that the slope detector is less sensitive to noise transition conditions. do.

FIG. 5 shows how the line 10 is
2 shows a DC deviation removal and averaging circuit that operates to produce a 7-bit deviation-free average value T for each channel on line 115 from the 8-bit individual signal samples of the channels given above. ing. This averaging circuit is a 12-bit adder 4
04 and a 48 channel x 12 bit storage device 405
, a buffer 406 having an erase input, and a 48 channel x 7 bit storage device 407 having a writable input. Each of these storage devices is further addressed to each channel via an arm bus, not shown.

The deviation remover is operative to produce a 7-bit absolute value signal for each channel on circuit 409 from which long-term DC deviations have been removed, and for this purpose the deviation remover is activated. reaches an equilibrium state in which, for each channel, the channel's 16-bit deviation value is stored in storage 403. In each frame, for each channel, the stored deviation value of this channel is loaded from the storage device 403 into the counter 402 and is available at the output of the counter. This deviation value is 8
The two most significant bits are sent to the subtractor 4 via line 410.
01, which subtracts the deviation value bit from the current sample of the channel to produce a 7-bit magnitude signal on line 409 and a sign bit on yet another output line 411. This line 411
is connected to the addition/subtraction control input U/D of counter 402 to increase or decrease the count of the counter by 1 according to the polarity of the sign bit on line 411. In this way, the counter 402 produces at its output a new modified 16-bit deviation value for the channel, and this new value is written into the storage device 403 in place of the previous deviation value for the channel. This sequence is repeated for subsequent channels in each frame. The equilibrium state obtained in the long run is that for each channel the line 41
It appears that the number of positive and negative sign bits generated in 1 is equal. The stored deviation value for each channel changes, but only its 8 most significant bits are subtracted from the channel information, and the 256 sign channels of one polarity actually change the channel of the subtracted deviation value by one step. required to do so.

The averaging circuit includes circuit 11 for each channel.
5 to produce an average value T of 7 bits. In fact, to simplify the configuration of this circuit, the average value T on circuit 115 is actually
This results in 27/32, which is part of the actual average value of the above signal. For each channel, this average value T is input to buffer 406 via line 414 and to storage 407 in the 15th % of each top %.
The new average value T is written to the storage device 407 and the buffer 406 is cleared.

For each channel in each frame of the superframe, the output of adder 404 is stored in storage 405. The output of this adder is line 4
is equal to the sum of the 7-bit absolute value signal for the particular channel present on line 412 and the 12-bit cumulative sum for the particular channel present on line 412. The cumulative sum for this channel is the previously stored sum for the channel stored in storage 405, and this sum is added to the cumulative sum of each superframe when buffer 406 is cleared to reduce the cumulative sum to zero, as described above. It is clocked through buffer 406 in each frame except the 15th frame.

In the 15th frame of each superframe, the 12-bit cumulative sum produced at the output of storage device 405 for each channel is the absolute value signal with the deviations for that channel removed in the previous 27 frames. is equal to the sum of Only the seven most significant bits of this sum are written to storage 407,
A division of said sum by a factor of 32 is carried out, so that the mean value T is 27/32 of the actual mean value. This small difference does not adversely affect the operation of the voice detection device.

Although a particular deviation removal and averaging circuit has been described, the speech detection apparatus of the present invention may be used in conjunction with other forms of such circuitry, or may be used in conjunction with any deviation removal and averaging circuit of the present invention. It is also possible to use it without circuitry. Similarly, the audio detection device can be used for applications other than those described herein and can be provided for any number of channel signals. Many other configurations may be provided without departing from the scope of the invention as set forth in the following claims.

Effects As described above, according to the present invention,
It is possible to provide a method that can quickly detect an audio signal when audio is generated.

[Brief explanation of the drawing]

FIG. 1 is a conceptual diagram of an example of an apparatus for carrying out the method of the present invention. FIG. 2 is a block diagram showing a voice detection device capable of implementing the method of the present invention, and FIG. 3 is a flowchart illustrating the operation of the voice detection device of FIG. 2. FIG. 4 is a level diagram showing the operation of the voice detection device of FIG. 2. FIG. 5 is a diagram showing a deviation removal and averaging circuit for removing deviations and providing averaged signal samples for the voice detection device of FIG. 2; 110,115,409-608,613...
Line, 401... 8-bit subtracter, 402... Reversible counter, 403... 16-bit storage device, 40
4...12-bit adder, 405...12-bit storage device, 406...buffer, 407...7-bit storage device, 601...level detection device, 602...
... Gradient detection device, 603 ... OR gate, 604
...48 channel judgment storage device, 605,610...
... Comparator, 606 ... Shelving and control device,
607...determination storage device, 609...delay device,
611... shelving control and threshold generation device, 612...
...Judgment and threshold storage device.

Claims (1)

Claims: 1. In a sampled audio channel means producing a first signal state (M=1) whenever the absolute value T of a sample of a signal exceeds a first threshold level TF. In the method of detecting the presence of an audio signal, the absolute value T of each sample is defined as the absolute value of the previous sample.
Compared to TP, whenever the absolute value T of a sample is not greater than the absolute value TP of the previous sample, change the level of the second threshold TL to a greater level depending on the absolute value T of the sample at that time. Whenever the absolute value T of a sample is greater than the absolute value TP of the previous sample, if the absolute value T of the current sample is equal to the then second threshold level TL;
If the first signal state (M=1) or the second signal state (K=1) is present, then at least A method comprising: indicating the presence of audio in relation to a current sample. 2. Whenever the absolute value T of a sample does not exceed a first threshold TF and the above first signal state (M=1) is generated in relation to a previous sample, the value of the subsequent sample starting with the current sample With respect to the first predetermined number H, a third signal state (M=2) is generated, such that the absolute value T of the sample is the absolute value of the previous sample.
not greater than TP and second signal state (K=1)
Whenever is generated corresponding to a previous sample,
A fourth signal state (K=
2) generating said third and fourth signal states (M=2, K=
2) A method according to claim 1, comprising generating a signal representing a signal of the audio present corresponding to step 2). 3. determining said second number C according to the absolute value of the previous sample, said second number C being determined for each sample in relation to which said second signal state (K=1) is generated; is increased by some predetermined number less than or equal to the maximum number (C = 32), up to the minimum number (C = 0) of each of the samples whose absolute value T is not greater than the absolute value TP of the previous sample. A method according to claim 2 for reducing. 4. The absolute value T of a sample exceeds the absolute value of the previous sample, and in relation to the previous sample a fourth signal state (K=2) is generated and a second signal state (K=1) is not generated, if its absolute value T exceeds a third threshold TH, then a second signal state (K=1) is established in relation to the current sample.
and the second signal state (K=1) is generated in relation to the previous sample, and the absolute value T of the current sample is
4. A method according to claim 2, further comprising setting a third threshold value TH equal to the absolute value TP of the previous sample whenever TH is not greater than the absolute value TP of the previous sample. 5. Claim 1, characterized in that each time the second threshold level TL is set, the threshold is set to be larger than the absolute value of the sample at that time by a predetermined amount. Method described.