FR2501439A1 - DISCRIMINATOR OF ENERGY BAND - Google Patents

Abstract

THE INVENTION IS CONCERNED WITH DETERMINING THE SPECTRAL NATURE OF THE ENERGY CONTAINED IN A RECEIVED SIGNAL. AN ENERGY DISCRIMINATOR 103 DETERMINES WHETHER THE ENERGY PRESENTED IN A RECEIVED SIGNAL IS FULL BAND ENERGY OR PARTIAL BAND ENERGY, BY COMPARING AN AVERAGE VALUE OF THE RECEIVED SIGNAL TO AN MODIFIED ABSOLUTE VALUE OF THAT SIGNAL. WHEN THE MODIFIED ABSOLUTE VALUE EXCEEDS THE AVERAGE VALUE, THE SIGNAL RECEIVED CONTAINS FULL BAND ENERGY AND, OTHERWISE, CONTAINS PARTLY BAND ENERGY ONLY. THIS TECHNIQUE CAN BE USED IN AN ECHO 100 CANCELER TO IMPROVE ITS SIGNAL ADAPTATION CHARACTERISTICS. APPLICATION TO TELEPHONY.

Description

The present invention relates to an energy band discriminator

  for use in an echo canceller which includes an adjustable signal processing circuit connected to a first transmission path for generating an echo estimation signal, a combination network connected to an echo canceller,

  second transmission path to combine a pre-signal

  feel in the second way with the estimation signal

  echo to generate an error signal, a first cir-

  cooked which reacts to the error signal by setting the processing circuit, and a second circuit which applies the signal

  of error to the adjustable processing circuit.

  Echoes commonly appear because of imperfect coupling of incoming signals in junctions

  4 wires - 2 wires in telecommunications systems.

  The echoes typically result from an adaptation

  imperfect impedance of the 2-wire installation in the 4-wire - 2-wire junction, so that the incoming signal is partially reflected on an outgoing path,

to the source of incoming signals.

  Self-adaptive echo cancellers were used to attenuate the echoes by generating an estimate of the reflected signal, or echo, and subtracting it from the signal

  outgoing. The echo estimate is updated under the

  the outgoing signal to form a better approximation of the echo to be canceled. In the prior art, it is forbidden to update the echo estimate when speech signals from the near end are emitted or when no significant energy from the far end

  is not received. However, when authorizing the updating of

  echo timing when significant energy from the far end is received, be it speech, noise, single frequency signals, multi-frequency signals

discrete or analog signals.

  It has been determined that by allowing the canceller to update the echo estimate during intervals during which the far-end signal includes

  energy that occupies only part of a frequency band

  interesting, such as for example a monofre-

2501 4 3 9

  Thus, a signal having a plurality of discrete frequencies or a similar signal (hereinafter referred to as partial band energy) results in an undesirable condition of the telecommunication circuit comprising the canceller. Specifically, the canceller includes a self-adaptive processing circuit that can be set to a large number of transfer functions to generate the echo estimate that is the best approximation of the echo. Allowing the processing circuit to set the transfer function when partial band energy is received raises the following problem: although the transfer function that is being achieved is optimized for the frequency components of the partial band energy,

  it may not be optimal for frequency components.

  quence remaining in the interesting frequency band,

  for example the voice band. In fact, the function of trans-

  fert on which the processing circuit is set for

  frequencies other than those contained in the energy

  A partial bandwidth can be significantly different from the desired optimal setting, which would be obtained in the case of tuning to a full-band signal, ie speech or Gaussian noise. Therefore, a path says

  low attenuation is established for the frequencies

  these other than partial band energy. This

  low attenuation back can lead to oscilla-

  in the telecommunications circuit. These oscilla-

  tions are extremely undesirable and should be avoided.

  The problem of low attenuation back and others

  problems of the echo canceller structures of the prior art.

  This is because the canceller is allowed to adjust the echo estimate during the intervals during

  partial band energy from the far end of the

is received.

  The problem is solved according to the invention which, in a particular embodiment, comprises an energy band discriminator which interconnects the combination network with the first circuit to discriminate.

  between full-band and band-wave energy

  in a signal received in the first path of trans-

  mission, and for generating a control signal indicative of this, the energy band discriminator comprising a first filter circuit for generating a first signal representative of an average value of the received signal, a second filter circuit for generating a second signal representative of an absolute value of the received signal, and a control circuit for comparing the first and second signals and generating a first state of the control signal when the second signal is greater than the first signal;

  signal, the control signal being applied to the second circuit

  fired to allow the application of the error signal to the adjustable signal processing circuit during the intervals in which the first state of the

command is generated.

  The invention will be better understood when reading

  the following description of an embodiment, and

  with reference to the accompanying drawings in which: Figure 1 shows in simplified block diagram form an echo canceller which includes an embodiment of the invention; Figure 2 shows in simplified form the details of the energy discriminator used in Figure 1; FIG. 3 represents details of the circuit of

  command that is used in the discriminator of the

  re 2; Figure 4 is a state diagram useful for

  description of the operation of the discriminator of FIG.

  re 2 and the control circuit of Figure 3; Figure 5 shows details of another

  version of the control circuit used in discriminating

  of Figure 2; and Figure 6 shows in simplified form

  details of the filter that is used in the control circuit

of Figure 5.

  The echo canceller 100 including a mode of realignment

  The invention is represented in the form of a diagram.

  simplified synoptic in Figure 1. However, contrary-

  echo canceller structures of the prior art, such as those described in U.S. Patents 3,499,999 and

  3,500,000, and in an article entitled "Bell's Echo-

  Killer Chip ", IEEE Spectrum, October 1980, pages 34-37, the echo canceller 100 includes an energy discriminator 103 for controllably authorizing the update of an echo signal estimate, in accordance with FIG. to an aspect of the invention, when a far end signal received

  on a first transmission path, includes some

  class of signals including what is called energy

  complete band. In other words, the updating of the echo signal estimate is forbidden when the signal of the far end has a noticeable level.

  of energy that is only partial band. Generally speaking

  in an embodiment of the invention, an average absolute value of the received signal is compared to a modified absolute value of the received signal, and if the modified absolute value is greater than the average, the received signal is considered to contain full band energy. Yes

  this is the case, updating or adapting the estimate

  echo signal is allowed. Otherwise, the update of the echo estimate is prohibited. This allows the echo canceller to adapt to a transfer function only when the received signal contains full band energy and prohibits the update of

  the transfer function when only

  partial is received, which would be likely to give

  low return attenuation for other

  posers in the relevant frequency band

  health, for example the voice frequency band. As a result

  This avoids parasitic oscillations and other

  problems in the transmission network.

  - Briefly, the canceller 100 includes an adjustable signal processing circuit which includes a system

  closed-loop error reduction method which is self-adapting

  tif, since it automatically follows the variation

  signal in an outgoing path. More specifically, the cancellation

250 1 4 3 9

  their 100 uses an echo estimator 101 which includes a transversal filter structure for synthesizing a

  linear approximation of the echo, that is, an estimate of

echo.

  For this purpose, the incoming signal of the far end X (K) is usually applied by a person speaking at the far end, and via

  of a first transmission path, for example the conductive

  102, at a first input of the echo canceller 100 and in the latter at an input of the echo estimator 101,

  at an input of the energy discriminator 103 and at a first

  first input of the speech detector 104. The signal from the

  remote end, X (K), can be for example a digitally sampled speech signal, in which K is

  an integer identifying the sampling interval

  ge. The signal from the far end X (K) is also

  applied by the driver 105, possibly through the

  of a conversion circuit, for example a

  digital-to-analog converter, not shown, at a first

  first input of the hybrid circuit 106. It is usually desirable that the input signal applied to the hybrid circuit 106 from the driver 105 is transmitted to

  a close listener via the bidirectional path

  107. However, because of a mismatch of

  pedance in the hybrid circuit 106, which results from

  characteristic of the fact that the balancing impedance 108 is not exactly adapted to the impedance of the path

  bidirectional 107, a part of the input signal of the circuit

  cooked hybrid appears on the outgoing driver 109 and

  it is reflected back to the signal source of the

  far away, in the form of an echo. An exit from the

  baked hybrid 106 applies the echo to a second input of the canceller 100, by a driver 109, and in the canceller, the echo is applied to a second input of the detector of

  speech 104 and at a first entrance of the network of combin-

  110. A conversion device, for example a

  analog-digital converter, not shown, can also

  be inserted in driver 109. A second

2501 43 9

  input signal of the combination network 110 consists of an estimate of the echo generated by the echo estimator 101. The echo estimate from an output of the echo estimator 101 is applied by the driver 111 to the second input of the combining network 110. The combining network 110 generates an error signal E (K)

  which corresponds to the algebraic difference between the estimate

  echo and the output signal of the hybrid circuit 106, including the unwanted echo. The error signal E (K) is applied by a second transmission path, for example the conductor 112, to the source of the far end and to a controlled switching gate 113. The gate 113 is controlled so as to be enabled or disabled by an output signal from the AND gate 114. A first state of the output signal of the AND gate 114, for example a state

  logic 1, validates the gate 113 so as to apply the

  error signal E (K) to the estimator 101, while a second state of the output of the AND gate 114, for example a logic state 0, prohibits the gate 113 from applying the signal

E (K) to the estimator 101.

  In the prior art, the door 113 was

  to prohibit the application of the error signal

  E (K) to the estimator 101 in the absence of a noticeable energy

  from the far end, in the presence of

  from the near end, or where a rela-

  determined between the error signal.E (K), the signal

  from the far end X (K) and an indi-

  the presence of speech signals from the

  as described in U.S. Patent 4,129,753.

  As indicated above, the signal from the far end

  X (K) could contain speech, noise,

  how often among a certain number of frequen-

  these individual, multi-frequency signals

  cretes, or similar signals. Thus, in the structures

  earlier, the error signal E (K) was not

  that in the case of lack of detection of an energy

  noticeable from the far end or in the case of detecting

  the speech of the near end. On the other hand, the

  error code E (K) was applied to the estimator 101

  intervals during which a noticeable energy

  ble of the far end was detected in the signal

  X (K). This energy could be band energy

  which is a single-frequency signal, signals

  at multiple discrete frequencies or analog signals.

  Therefore, the estimator 101 could adapt or be adjusted in any other way during the intervals in

  which only partial band energy was received.

  As indicated above, such a setting leads to results

  unwanted states. More particularly, the transfer function on which the estimator 101 can adjust for the frequency components of the partial-band signal was likely to lead to a low return attenuation for the other frequency components in the frequency band of interest. This can, in turn, give

  parasitic oscillations in the tele-

  communications. In accordance with one aspect of the invention, parasitic oscillations and other problems that result from allowing adjustment of the estimator 101 in the presence of partial band energy by using the energy discriminator are avoided. 103 to determine if the far-end signal X (K) contains only

  partial band energy or full band energy

  you. If it is determined that the signal X (K) does not correspond to full band energy, for example speech or

  noise, or in other words if X (K) corresponds to energy

  partial band, such as a single-ended signal

  In the case of signals with multiple discrete frequencies or similar signals, the discriminator 103 generates an output signal which invalidates the AND gate 114. On the contrary,

  when full-band energy is detected, discri-

  The timer 103 generates an output signal that validates the AND gate 114. The AND gate 114 in turn generates a control signal to prevent the gate 113 from applying the signal E (K) to the estimator 101. More precisely, a first state

  control signal from the gate 114, for example

  a logical state 1, validates the gate 113, while a

2 501439

  second state of the control signal, for example a state

  logic 0, invalidates the gate 113. Therefore, the estimate

  echo generated by the estimator 101 remains constant during the intervals in which only partialband energy is present, and an inde-

  sirable of the transfer function of the canceller.

  The estimator 101 comprises a so-called tap delay line formed by delay elements -1 to 115-N so as to obtain desired delays on the taps which correspond to convenient Nyquist intervals. Delayed versions X (K-1) to X (K-N) of the incoming signal of the far end X (K) are thus

  generated on the corresponding sockets. The signal to

  that setting position, ie X (K-1) to X (K-N) as well

  that X (K) is set under the action of the error signal E (K).

  More precisely, the signals X (K) to X (K-N) are weighted individually under the action of E (K), via a corresponding respective network among the control networks 116-0 to 116-N. Each of the control arrays 116-0 to 116-N comprises multipliers 117 and 118 and a feedback loop 119. The feedback loop 119 sets the weight of the pick at a desired value.

  which will be apparent to those skilled in the art and who is

  in the above references. The weighted versions of X (K) from the setting networks 116-0 to 116-N are summed by the summation network 120 to generate the

  echo estimation signal which is an approximation

  the echo to be canceled. The echo estimate is applied

  driver 111 at the second network entrance

combination 110.

  Figure 2 shows in schematic form

  Simplified synoptic one embodiment of discrimi-

  energy generator 103 that can be used in accordance with

  to an aspect of the invention, to determine whether energy

  Noticeable present in the received signal X (K) is full band energy and thus not only partial band. In this example, we should not consider

  limiting the scope of the invention, the frequency band

0 1 4 3 9

  What is interesting is the tele-

  about 300 Hz to 4000 Hz.

  plete, consists for example in speech, in Gaussian noise,

  etc., ie signals with frequency components

  which extend over the entire frequency band. The partial band energy consists, for example, of single-frequency signals, of signals with several discrete frequencies,

  etc., ie signals with frequency components

  quence in relatively narrow parts of the band

interesting frequency.

  Therefore, the received signal X (K) is applied to the rectifier 202 by the splitter amplifier 201. Any of the many full-wave rectifiers known in the art can be employed for this purpose. If X (K) was a representative digital signal, for example, of a law sample li, we would use a linear y-law law converter, not shown, after the rectifier 202. In this example, we assume that X (K )

is an analog signal.

  The MAG rectified version of X (K) is applied to a first filter 203 and a second filter 204. The filters 203 and 204 are used to determine prescribed characteristics of the received signal X (K), in order to distinguish if X (K) ) includes full band energy or only partial band energy. In this example, the filter 203 is used to determine an average value of MAG, while the filter 204 is used to determine a modified absolute value of MAG. For this purpose, the filter 203 is a low-pass filter having a first predetermined time constant, while the filter 204 has a second fixed time constant. Since in this example, the filter 204 generates the modified absolute value MOD MAG of

  MAG according to a certain criterion, the second con-

  time is zero and the filter 204 is essential-

  an attenuator. In this example, MOD MAG is less than

  9 dB at MAG, that is to say we have

MOD MAG = MAG-9 dB.

  The filter 203 generates practically the average

250,143 9

  rotating of MAG and it has a short time constant which

  is as an example of the order of 8 to 16 ms. More precise

  the filter 203 is a resistance-capacitor

  (RC), not shown, having an expo-

  nential determined to generate a version of MAG

  corresponding to a history with exponential representation

  tielle (HRE). Note that one can also use

  other filter characteristics to get the histori-

  only with exponential representation (HRE) of MAG. Various configurations and techniques can be used to generate the short-term rotating average of the MAG signal. As noted above, one technique is to determine the historical exponential representation (HRE) of the

  signal. The average calculation of the HRE type is particularly

  useful in command or detection situations in which recent behavior is

  process, and it is described in the IRE Trans-

  sactions on Automatic Control, Vol. AC-5, January 1960,

  pages 11-17. The average HRE of a continuous signal is determined

  nude by weighting more strongly the signal that appeared recently

  than the signal appeared less recently. The relative weighting

  of a continuous signal is, for example, an expo-

  nentielle. The signal HRE and the signal MOD I4AG are applied to the control circuit 205 to generate an ADAPT signal according to certain criteria. In this example, the signal ADAPT is used to control the validation and the invalidation of the AND gate 114 (FIG. 1), and thus the authorization and the prohibition of the update of the echo estimation that generates the echo estimator 101 (Figure 1). More precisely, when the signal ADAPT is in a first state, for example a logic state 1, the signal X (K) contains full-band energy, and when the signal ADAPT is in a second state, for example a state logic 0, the signal X (K) contains energy at

partial band.

  Figure 3 shows details of a type of cir-

  205. The HRE signal is thus applied to

2 50 1 4 3 9

- 11

  a first input of comparators 301 and 302. The signal MOD MAG is applied to a second input of the comparator 302, while a signal TH is applied to a second input of the comparator 301. The comparator 301 is used to detect whether the received signal X (K) includes significant energy from the far end. So, if HRE

  exceeds a predetermined threshold TH, it is assumed that X (K) con-

  holds significant energy. In this example, TH, -50 dBmO. An output signal of the comparator 301 is applied to a cutter: tap 303. The timer 303 is erased.

  to determine whether the notable energy of the far end

  is present for at least one predetermined interval

  mined T1. In this example, timer 303 establishes a

  wait interval of T1, 24 ms. This is intended to

  to erroneously generate ADAPT = 1 during the initial interval of the received signal X (K), when the output signal of the filter 203 (FIG. 2) is in a transient state. An output signal of the timer 303 is applied to a first input of an AND gate 304. The AND gate 304 is thus disabled until HRE is greater than TH.

during a T1 interval.

  Comparator 302 compares MOD MAG to HRE.

  When MOD MAG is greater than HRE, comparator 302 generates an output signal in logic state 1. The output signal of comparator 302 is applied to a second

  the AND gate 304. The AND gate 304 is thus invoked

  the idea until MOD MAG is higher than HRE.

  The output signal from the AND gate 304 is applied to a coeffector: Torer CTR 305. The timer 305 responds to a logic state 1 from the AND gate 304 by immediately generating an output signal ADAPT, 1, and generating the signal of output ADAPT = 1 during a second predetermined additional interval T2, at the moment of a transition from logic state 1 to logic state O of the output signal of AND gate 304. The interval T2 is a so-called interval of extension and, in this example, it adds a duration of 24 ms to the output signal in logical state 1 of the AND gate 304. This generates ADAPT = 1 for an interval long enough for the canceller 100

  update the echo estimate that is generated.

  The state diagram shown in Figure 4

  summarizes the operation of the energy discriminator 103.

  We simply have ADAPT = 0 until HRE> TH during T1, and MOD MAG> HRE. When all the above conditions are satisfied, X (K) comprises full-band energy and ADAPT = 1 for an interval at least equal to

the interval T2.

  It can be seen that ADAPT = 0 during the inter-

  valleys in which we have: HRE> TH but MOD MAG <HRE.

  When this happens, the energy is partial band and

  the update of the echo estimate is prohibited.

  Figure 5 shows details of another type of control circuit 205. HRE (K) is applied here to

  a first input of digital comparators 501 and 502.

  MOD MAG (K) is applied to a second input of the comparator

502 while the threshold signal TH is applied to

  a second input of the comparator 501. The comparator

  controller 501 for detecting whether the received signal X (K) has significant energy from the far end. Thus, if HRE (K) exceeds a predetermined threshold TH, it is assumed that X (K) contains a notable energy. In this example, TH

  is equal to 16, over a total linear range of 4079.5.

  The output of the comparator 501 is applied to a timer 503. The timer 503 is used to determine whether the noticeable energy of the far end

  is present for at least a first predetermined interval

  completed T1. In this example, timer 503 sets a wait interval of T1 = 24 ms. This is done by counting 192 frames at 8 kHz, to generate HC (K) '= 1, and

  HC (K) = 0 otherwise. This is intended to prevent

  expensive to inadvertently generate ADAPT (K) = 1 during the inter-

  initial value of the received signal X (K), when transients

  res may be present. The output signal HC (K) of the timer 503 is applied to a first input of an AND gate 504. The gate 504 is thus disabled.

  until HRE (K) is greater than TH,

250 1 4 3 9

T1 valley.

  The comparator 502 compares MOD MAG (K) to HRE (K) by proceeding sample by sample. When MOD MAG (K) is greater than HRE (K), comparator 502 generates an output signal at logic state 1. For speech, ie for full band energy, MOD MAG (K) must be greater than HRE (K) approximately once per fundamental period. The output of comparator 502 is applied to a second input of AND gate 504. Thus, when AND gate 504 is enabled by HC (K) = 1, it applies to the digital filter 505 a configuration d (K) of logical states 1 and O which is representative of the result

  the comparison between HRE (K) and MOD MAG (K).

  The low-pass filter 505 is used, in accordance with

  to an aspect of the invention, so as to lower the comparison threshold between HRE and X '(K), which improves the detection performance when full band energy is received. This is possible because it is possible for the decefined circles to be wrong for the comparison between HRE and MOD MAG, without affecting the decision to generate ADAPT (K) = 1, because of the function of the filter. The filter 505 generates a digital output signal f (K) which is applied to an input of the digital comparator 506. The details of the filter 505 are shown in FIG.

described below.

  In combination with a threshold selector 507, the comparator 506 establishes a hysteresis, according to one aspect of the invention, in the decision to generate the

  first and second states of the ADAPT control signal (K).

  More precisely, the threshold selector 507 reacts with a first state of ADAPT (K), that is to say ADAPT (K) = 1, in

  applying a first predetermined threshold THI to a second

  input of the comparator 506, and it responds to a second state of ADAPT (K), i.e. ADAPT (K) = O, by applying a second predetermined threshold TH2 to the second input of the comparator 506. The values Threshold values are chosen in relation to the factor F of d (K) in the filter 505, as described below. In one example, we choose F

250 1 4 3 9

  equal to 512 and we choose TH1 equal to 4F = 2048, while TH2 is chosen equal to 2F = 1024. We thus see that we establish

  a hysteresis in the generation of ADAPT (K). More precise-

  Because TH1 equals 4F = 2048, F (K) must exceed this higher value before ADAPT = i is generated. This makes it possible to tolerate certain errors in the comparison between HRE and MOD MAG, because of transients and similar phenomena, without prematurely generating ADAPT 1 and to allow updating of the echo estimation on an incorrect signal. Moreover, since TH2 is chosen equal to 2F = 1024, once ADAPT = 1 is generated, it is held until f (K) falls

  below the TH1 lower threshold. This establishes a hysteria

  resists in the generation of ADAPT = 1. Therefore, once the ADAPT condition, 1 has been generated, it remains

  during a significantly longer interval than with the use

  extension timer. As a result

  the condition ADAPT = 1 is maintained longer, without returning to the ADAPT = O condition, so that the

  update of the estimated echo is less often prohibited.

  FIG. 6 represents in simplified form the details of the digital filter 505. For the clarity of the

  description, the time signals have not been represented.

  In this example, it is assumed a circulation of the bits in series, although one can also achieve the filter by using a circulation of the bits in parallel. The digital filter 505 is a digital low-pass filter and it is validated by the logic state 1 of the signal HC (K), so as to filter the signal d (K) according to the relation: f (K + 1) = (1-) f (K) + P (k) (1)

  in which = 1/512 and K is the generated sample

  rant. When HC (K) is in logic state 0, we have: f (K + 1) = f (K) (2) The output signal d (K) of the AND gate 504

  (Figure 5) is thus applied to an input of the multiplier

  401, while a framing factor F is applied to a second input to generate a framed version Fd (K) of d (K). The framing factor F is a chosen number

  so that f (K) is an integer and retains a pre-

  desired decision. In experimental practice, the framing function is performed by an approximate delay of d (K) until a desired value is obtained, for example F = 512. The signal Fd (K) is applied to a first

  input of an adder 402 while a signal representing

  (1-P) f-K) is applied to a second input. The output signal of the adder 402 is a current sample f (K), and then the next output sample is f (K + 1). The signal f (K) is applied to a shift register 403. When the shift register 403 is enabled by HC (K), 1, it generates Pf (K) on one output and f (K) on another output. We choose the number of floors in the

  shift register 403 to define P, in this example

  P, 1/512. When HC (K) = 0, the shift register

  403 is invalidated. The signal Pf (K) is applied by an invert

  405 to a first input of an adder 404, tan-

  say that the signal f (K) is applied to a second input.

  The adder 404 generates a signal representative of

  (1 - () f (K) which is applied to the second input of the addi-

402.

  We have just described the invention considering

  its use in an echo canceller, but it can also

  be used with other adaptive filters, or in any application in which the received energy is to be classified in the partial band type or

the full band type.

  It goes without saying that many modifications can be made to the device described and shown,

  without departing from the scope of the invention.

250 1439

Claims (13)

  1. Usable energy band discriminator   in an echo canceller which comprises: a processing circuit   adjustable signal system (101) which is connected to a first transmission path (105) so as to generate an echo estimation signal, a combination network (110) which is connected to a second transmission path (109). ) to combine the signal present in the second path with the echo estimation signal, to generate an error signal (E (K)), a first circuit (117) which responds to the error signal ( E (K)) so as to adjust the processing circuit, and a second circuit (113) for applying the error signal to the adjustable processing circuit, characterized in that said energy band discriminator interconnects the control circuit. combination (110) and the first circuit (117) to discriminate between the full-band energy and the partial-band energy in a signal received in the   first transmission path (105), and so generously   send a control signal indicative of this; and discri-   energy band miner includes a first circuit of   filter (203) for generating a first signal representative of   of an average value of the received signal, a second circuit   filter bake (204) for generating a second signal representative of an absolute value of the received signal, and a control circuit (205) for comparing the first and second signals and generating a first state of the control signal when the second signal is greater than the first   signal, this control signal being applied to the second circuit   cooked (113) so as to allow the application of the signal   (E (K)) to the signal processing circuit   during the intervals during which the first   state of the control signal is generated.   2. energy band discriminator according to claim 1, characterized in that the second filter circuit (204) comprises means for modifying the absolute value of the received signal according to a determined criterion. 250 1 4 3 9   3. energy band discriminator according to claim 2, characterized in that the second filter circuit (204) comprises an attenuator for generating the modified absolute value according to a given relation to the absolute value of the received signal. An energy band discriminator according to claim 1, characterized in that the first filter circuit (203) comprises means for determining   a short-term rotating average value of the received signal.   An energy band discriminator according to claim 4, characterized in that the first filter circuit (203) comprises low pass filtering means   having a predetermined time constant.   The energy band discriminator according to claim 4, characterized in that the first filter circuit (203) comprises means for determining the average value corresponding to the history with   exponential representation of the received signal.   7. energy band discriminator according to claim 6, characterized in that the second filter circuit (204) comprises means for attenuating by a predetermined value the absolute value of the received signal. 8. Energy band discriminator according to the   Claim 7, characterized in that the communication circuit   (205) further comprises the means for gen-   the first state of the control signal for at least a predetermined interval.   9. energy band discriminator according to claim 8, characterized in that it further comprises means for prohibiting the generation of the first state of the control signal until the first signal   has an absolute value that exceeds a threshold level   determined during a predetermined interval.   10. Energy band discriminator according to the   claim 1, characterized in that the control circuit   of (205) further comprises a filter circuit (505) which has a characteristic determined to generate the signal of 250 1,439   command. An energy band discriminator according to claim 10, characterized in that the filter circuit (505) includes a low pass filter.   12. Energy band discriminator according to the   claim 10, characterized in that the communication circuit   Means (205) further comprises a comparator (506) which receives a filter output signal (505) and a predetermined threshold value, for generating a first state of the control signal indicating that full band energy is not received, when the value of the amplitude of the output signal of the filter is lower than a first threshold value, and a second state of the control signal indicating that full band energy is received, when the value of the amplitude of the output signal of the filter is greater or   equal to a second threshold value.   13. Energy band discriminator according to claim 12, characterized in that the first value   threshold is greater than the second threshold value.