JPH02288755A - Method of voice control attenuation of telephone line - Google Patents

Abstract

PURPOSE: To obtain high stability and articulation by dividing the speech frequency band of an incoming signal into frequency band parts by analytic filter banks and selectively controlling the individual frequency bands. CONSTITUTION: The incoming speech frequency signal is supplied to the analytic polyphase filter banks APF1 and APF2 and divided into the individual frequency bands TF1 to TFn. Speech evaluation logic parts SA1 to SAn are provided corresponding to the respective frequency band parts TF1 to TFn, deviation from a specific voiceless level is compared to decide which of signals in a transmission and a reception direction has a larger speech activity, and the attenuation constant of the attenuator having the smaller speech activity is made large. Consequently, there is possibly the case that two-way transmission and reception become possible at the same time under specific conditions.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to voice control attenuation in telephone lines.
In particular, it relates to such a reputation control attenuation method in a hens-free device configured such that a signal in one transmission direction influences a signal in the opposite direction, and the attenuation value of a variable attenuator inserted in the signal path is changed by the &1Jtll signal. .

BACKGROUND OF THE INVENTION A circuit configuration for hands-free communication in which both transmission lines are disconnected is known from West German Patent Publication DE-OS 3606973. In this configuration, when transmission on one transmission path is promoted in accordance with the audio signal levels on both transmission paths, transmission on the other transmission path is suppressed. In this case, the bandwidths of both transmission lines are influenced by controllable filters.

That is, a transmission line transmitting a human level signal is made to have a wider bandwidth than a transmission line transmitting a low level signal. Such M III Lt 'Fi pressure tl
This is done by changing the frequency of the I3111 oscillator according to the octuplet signal level. The threshold frequency set corresponding to the silent state moves as the level increases, so that when the transmission level is high, only low frequency signals are transmitted in the reception direction, and when the reception level is high, the transmission (n However, this method does not avoid that the quality of the call is impaired as a result of the frequency change, at least during the transition period.

From West German Patent Publication DE-・O83528973,
Audio $1 for two signal paths transmitting in opposite directions
111 attenuation circuit configurations are known. In this case, a controllable attenuator is inserted into each transmission line. that time,
The signal in one signal path is fed to the Dllillable amplifier and the other signal path is used for the attenuation initial stage.

This causes weak signals to be more attenuated in each transmission path, while signal paths carrying strong signals receive little attenuation. In order to avoid this, the circuit described in West German Patent Publication DE=OS 3528973 further includes a delay circuit. However, such delay circuits do not operate unadjusted over the entire call w!Ii!I, and are often used in cases where the line already contains a large signal delay, such as a line using a submarine cable, for example. This causes problems.

West German magazine [Electronic and transmission technology magazine J (Archive fa
rE Iektronik und Ubertrag
ungtechnik) Volume 39 (1985) No. 2
From page 123 onwards, the issue contains a paper titled "°Filter Design 4 for Contemplative QMF and Polyphase Filter Banks". This paper describes a technique for dividing an audio frequency signal into individual frequency bands using an analysis filter bank.

The signal divided into such frequency bands is restored to the original signal using a synthesis filter bank operating on the same principle. The analysis polyphase filter bank can also be configured as a demultiplexer of frequency-multiplexed signal channels, in which case connections are made selectively between different individual frequencies. In addition, frequency division multiplexed incoming signals can be time-division multiplexed.
A polyphase filter bank may be used if the purpose is to convert an incoming signal into a frequency division multiplexed signal or to convert a time division multiplexed incoming signal into a frequency division multiplexed signal.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method for performing voice-controlled attenuation in a telephone transmission circuit without disturbing voice transmission. According to the present invention, the loss of eight-tone parts (particularly consonants), which occurs in conventional methods due to delays or long switching times, is avoided. Delay circuits, which were previously required to avoid too frequent attenuation switching in the passage channel, are not used. Furthermore, the method according to the invention can be adapted to be used in place of known echo suppression means if necessary. The audio frequency band of the signal to be transmitted is divided into multiple frequency band parts by an analysis filter bank, and an independently adjustable M table is provided for each frequency band part for each transmission direction, and each attenuator has its own prestige rating [ The setting command is given by a level scale ruled by υ1 by one pick, and the total frequency m'i is synthesized by a filter bank containing individual frequency band partial signals.
This is achieved by a device configured to return the iA signal and output it from an output terminal.

High stability and clarity can be achieved by selectively controlling individual frequency bands. Furthermore, if the 8-way state differs depending on the frequency band, it is possible to transmit on one frequency and receive on another frequency at the same time, thereby forming a pseudo double line. Furthermore, by dividing the frequency Wt range, the adjustable range of the attenuator provided separately for each frequency band can be made smaller compared to the conventional case where only one attenuator is provided over the entire frequency band. is possible.

EXAMPLES The present invention will be described below with reference to the drawings.

FIG. 1 is a 10-step diagram of a Banstree telephone device according to the present invention. In the figure, the portion surrounded by broken lines performs digital processing of the signal. A10 converter and [)/A associated with digital processing
The conversion insect or aliasing removal filter is not shown in order to avoid multiple drawings. Digital processing is performed by a dedicated processor (hardware) or by a general digital processor that executes software corresponding to the functionality block. Further, a hardware ab-o-set and a digital processor may be used together.

Microphone M is connected to hybrid G via a device described later. As is well known, this hybrid G couples the claw direction signal in vll to the external line Δ. The signal coming from the external line AL connects the hybrid circuit to the variable amplifier V.
guided by. Setting of this variable 7-amp V is performed by a level setter Ps so that A/D conversion, which will be described later, is optimally controlled. Thereby, the reception level can be uniformly reduced.

To compensate for mismatching in the hybrid circuit, an echo barrier is provided between the two transmission directions. This eliminates the portion of the signal that travels from the transmit path through the unmatched hybrid to the receive path. Acoustic echo suppressor 8F has a similar purpose. In this case, the portion of the signal that travels from the loudspeaker through the room to the microphone is at least partially filtered out. In particular, short acoustic echoes (direct sounds) can be effectively suppressed. Providing complete acoustic coupling requires correspondingly expensive construction. Both echo suppressors operate on the same principle. This is a so-called adaptive filter, which operates, for example, according to the known LMS algorithm (Sinday,
M 1m; Berkeley, Day 1 - Science of echoes in telephone networks.

Proceedinges op the Iiiiii, No. 6
Volume 2, No. 854. August 1980; 3ondhi.

M, M: Berklev, D, A, , S
ciencingEChO3On the Te! cp
honc Network, Process-din
g of the IEEE, Vo1, 62
．． Ne8. A11 (1°1980).

Also, instead of Onken Echo Suppress AF, a speaker (
, the frequency band of the signal supplied to the frequency shifting device? t
Approximately 5l-1zi'i using FE! You can shift it.

Thereby, 8W feedback can be more effectively prevented. A detailed explanation of this method is given in the literature (Improving Stability Against Acoustic Feedback Through Frequency Shifts, Acoustic Society Op América 36).
1964), pp. 1717-1724: 5 Chr.
'o'der, M, R,, I 1111) rO
Vellent OfA coustic -Fee
dback Stability by Fre
-quency 3hHting, Ac
oustic 3ociety oz8-cr
ica 36 (1964), P, 1717
-1724).

FIG. 2 shows a W control device ff5E that performs direction-specific IIIvIJ in the frequency band portion. The division into band parts and the synthesis after processing in each band part is done by an analysis polyphase filter bank (APFI, APF2) and a synthesis polyphase filter bank (SPF1, SPF2). These are specially configured digital filter banks designed to minimize 61% programming and memory requirements based on predetermined boundary conditions. , detailed examples can be found in the literature (Barry, Bee, V7T [
1 Iterge, A Unified Approach to Digital Polyphase Filter Banks, AEC Volume 37 (1983)
1/2 N: Vary, P, , Wac
kcrstreuthcr.

G,, A L! n1fied approach
to [)igiralP olyphase F
i er Banks, A E LJ Band3
7 (1983) Heft1/2. Filter Design for Ideal QMF and Polyphase Filter Banks 81, Vol. 39 (198
5th year) 2nd M pages 123-130: WaCkerStr
eu-ther, G, , On the
Design or F 1lters f
oridal QMF and Polyphase
Filter 3anks39 (1985),
He “s 2. S, 123-130>.

The audio frequency signal coming from the microphone M is supplied to the input terminal E1 of the analysis polyphase filter bank APF1, and the filter bank APFI separates the individual frequency bands rF1 to T F
divided into n. Synthetic polyphase filter SP that operates on the same operating principle for each frequency band partial molecule F.
A signal path leading to F1 is provided. Variable attenuators DG1.1 to DG, 1.0 are installed in each of these signal paths.
An analysis polyphase filter bank APF2 and a synthetic analysis polyphase filter bank SPF2 of the same construction are also provided for the reception path, and the signal entering the input terminal E2 is processed in exactly the same way as for the transmission signal described above. Frequency band fraction FI
~TFn process the same bandwidth, so they are shown in the same way in the drawings. Also in the receiving path, each frequency band component F is connected to a single variable attenuator DG2.1 to D.
G2,n is provided. Synthetic polyphase filter bank SPF
The signal reconstructed by 2 is guided to the transmitter output terminal of the v1 output device. In addition, each frequency band part molecule F1 to TFn
Call evaluation 111I logics SΔ1 to SAn are provided correspondingly. By comparing the deviation from a predetermined silence level, it is determined which signal in the transmit direction and in the receive direction has more speech activity. Depending on the result, each level balancer PW1 to PWn is set, and the attenuator with lower call activity, for example DG
The attenuation constant of 1,1 is considered to be large. After a subscriber's call activity during a normal call, the call activity changes in magnitude;
The setting of the level balancers PW1 to PWn is switched to reverse sense, and the corresponding attenuator DG is thereby set to υ11II. As a result, each frequency band partial molecule FI to TFn is separated and t/3911 is performed independently of each other. As a result, under a predetermined condition +1, simultaneous bidirectional transmission and reception may even become possible. This occurs when the voice conditions are active in one frequency band group in the transmit direction and active in the other frequency band group in the receive direction when both subscribers speak at the same time.

In order to prevent these effects from causing undesirable feedback, the controller SF scans the settings of a level balancer (not shown) to determine which transmission direction the frequency band molecules F1 to TFn are more active in communication. Based on the result of this determination, a Q reducer, such as DG2, is used to reduce the overall call activity.
．． 1-DG, 2. 8th third, where n is given a large attenuation value
The figure is a diagram for clearly explaining the principle of 7'j' @ill t[I. The signals in each band in the transmission and reception directions are first compressed. As the compression characteristics, a known linear PCM conversion table, or in some cases, the Mu or A law can be used. What about this in the sending direction? This is to equalize level differences caused by various changes in propagation attenuation in the receiving direction.

Subsequently, the 8-voice detector was compressed (ff) to evaluate the recovery behavior and the noise behavior. A cursive filter (IIR filter) is used to test the motion. By these means, the actual comparison value is always subtracted from the actual comparison value for the audio signal.

A cut-off delay circuit is installed following the audio detector.

This circuit consists of two counters, and its operation is influenced by the evaluation result and the judgment result of the voice detector. A short circuit (holding time) is made by one counter. The other counter registers the dropout rf after voice detection. This counter is set to a predetermined value by the voice detector during voice recognition and is reset to zero at the end of the call by the evaluation logic. The area of this counter provides a reference for allocating the amount of attenuation to each frequency band portion by level paran+JPW.

Evaluation 1 IIli logic evaluates the state of the speech detector in all frequency bands. For example, if a significant activity clearly appears in one direction, the entire band is switched in this direction by decrementing the counter from 1 to 1.

Other evaluations can also be considered, such as evaluations based on auditory physiology or reputation theory reasons.

In the level balan "jPW", the difference in the end-of-speech detection counter is used as a measure when determining the attenuation distribution.

Based on this difference, the more active direction in each band portion is recognized. When allocating attenuation values to each band part, attenuation in one direction is maximized corresponding to the maximum value of the negative difference value, and transmission in the other direction corresponding to the maximum value of the positive difference value is unattenuated. Let it be done. Further, at the intermediate value, the attenuation value is set so that the sum of the attenuation values in both directions is kept at the maximum value rI#. I+ for those who are active on calls! i] changes, the attenuation value changes smoothly accordingly.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a telephone transmission device for a hands-free communication device, for example, and Fig. 2 is a block diagram of a telephone transmission device for a hands-free communication device, and Fig. 2 is a
I tll notc lh Q) uI t[Circuit (7
) 70 ツ'l Figure, I 31J $, is a diagram showing the principle of direction control. E1, E2...Input port, APF1, APF2...
・Analysis filter back, SPF1. SPF2...Synthesis filter back, DingFl-TFn...Frequency band part, DG1, 1 to DG2. n-...Attenuator, 5AL1~5
ALn...Call evaluation logic, PW1~PWn...
Level balancer, 81.32...output port, PS/
...Level control circuit, ■...Amplifier, FE...Frequency shift device, ES, 8E...1 code blazer, G...
...Hybrid, 8 [...Line, M...Mic,
[...Speaker. Patent Applicant: Telephone Bara Lant Normal RAFT GMBH Procedures, City Council, Required Method) Relationship with the August 1989 Incident

Claims (1)

[Claims] 1. A hands-free communication device configured such that signal transmission in the opposite direction is influenced by a signal in one transmission direction, and a variable attenuator inserted in the signal path is changed by a control signal. This is a voice control attenuation method for the telephone line used, in which the voice band signal entering each input port (E1, E2) is divided into multiple frequency band parts (TF1 to TFn) by an analysis filter bank (APF1, APF2). Then, a variable attenuator (
DG1, 1-DE2. n) is provided for each transmission direction, and a call evaluation logic (
It is set by a setting command from a level balancer (PW1 to PWn) controlled by SAL1 to SALn), and the individual frequency band parts (TF1 to TFn) are synthesized by a synthesis filter bank (SPF1, SPF2) to obtain the entire frequency band. This is returned to the signal and sent to the transmitting side output port (S1, S2
). 2. According to claim 1, the noise in a silent state during a "pause" that occurs during a call is newly evaluated by noise motion evaluation, and the result is used to evaluate the level for the next call. the method of. 3. When setting the level balancer (PW1 to PWn), evaluate the call activity in both transmission directions, determine in which transmission direction the frequency band part (TF1 to TFn) is active in call activity, and respond according to the determination result. 2. The method according to claim 1, wherein the attenuation is increased for the whole band portion in the other transmission direction. 4. one or more of the other transmission directions such that the attenuator (DG) settings are varied to enhance speech clarity and stability based on a determination of speech activity in one or more frequency band portions (TF1 to TFn); Multiple frequency band parts (TF1~
2. Method according to claim 1, characterized in that TFn) is influenced. 5. The control device (SE) has an integrated signal processor;
2. The signal processor according to claim 1, wherein the functions of an attenuator (DG), a level balancer (PW), a speech evaluation logic (SAL) and a filter bank (APF, SPF) are performed digitally within the signal processor. Method. 6. Method according to claim 1, characterized in that an echo suppression circuit (ES) is connected between both transmission directions for compensating for hybrid and/or line mismatches. 7. In order to further adapt the level, a level control circuit (P
2. Method according to claim 1, characterized in that a level-setting variable amplifier (V) controlled by S) is provided. 8. Echo suppression circuit (ES) and level control circuit (PS)
8. Method according to claim 6, characterized in that ) is formed in the control device (SE). 9. The method according to claim 1, characterized in that a frequency shifting device (FE) is formed in the receiving path in the control device (SE), thereby suppressing howling due to acoustic feedback. 10. Device according to claim 9, characterized in that, instead of the frequency shifting device (FE), an acoustic echo suppression device (AE) is provided so as to automatically adapt to the acoustic characteristics of the external world.