Echo Cancellation (779800), страница 2
Текст из файла (страница 2)
Therefore the only basis for discrimination of speechfrom echo is the signal level. As a result, the speech/echo classifier maywrongly classify and let through high-level echoes as speech, or attenuatelow-level speech as echo. For terrestrial circuits, echo suppressers have beenwell designed, with an acceptable level of false decisions and a goodperformance. The performance of an echo suppresser depends on the timedelay of the echo. In general, echo suppressers perform well when theround-trip delay of the echo is less than 100 ms. For a conversation routedvia a geostationary satellite the round-trip delay may be as much as 600 ms.Such long delays can change the pattern of conversation and result in asignificant increase in speech/echo classification errors.
When the delay islong, echo suppressers fail to perform satisfactorily, and this results inchoppy first syllables and artificial volume adjustment. A system that iseffective with both short and long time delays is the adaptive echo cancellerintroduced next.14.4 Adaptive Echo CancellationEcho cancellation was developed in the early 1960s by AT&T Bell Labs andlater by COMSAT TeleSystems. The first echo cancellation systems wereexperimentally implemented across satellite communication networks todemonstrate network performance for long-distance calls.Figure 14.5 illustrates the operation of an adaptive line echo canceller. Thespeech signal on the line from speaker A to speaker B is input to the 4/2wire hybrid B and to the echo canceller.
The echo canceller monitors thesignal on line from B to A and attempts to model and synthesis a replica ofthe echo of speaker A. This replica is used to subtract and cancel out theecho of speaker A on the line from B to A. The echo canceller is basicallyan adaptive linear filter. The coefficients of the filter are adapted so that theenergy of the signal on the line is minimised. The echo canceller can be aninfinite impulse response (IIR) or a finite impulse response (FIR) filter. Themain advantage of an IIR filter is that a long-delay echo can be synthesisedby a relatively small number of filter coefficients.
In practice, echocancellers are based on FIR filters. This is mainly due to the practicaldifficulties associated with the adaptation and stable operation of adaptiveIIR filters.402Echo CancellationFrom Speaker Ax A (m)AdaptivefilterHybridBechoxˆ A ( m)•To Speaker ASpeaker B+Echo cancellarechox B ( m) + x A ( m)Figure 14.5 Block diagram illustration of an adaptive echo cancellation system.Assuming that the signal on the line from speaker B to speaker A,yB(m), is composed of the speech of speaker B, xB(m), plus the echo ofspeaker A, x echoA ( m ) , we havey B (m) = x B (m) + x echoA ( m)(14.1)In practice, speech and echo signals are not simultaneously present on aphone line.
This, as pointed out shortly, can be used to simplify theadaptation process. Assuming that the echo synthesiser is an FIR filter, thefilter output estimate of the echo signal can be expressed asxˆ echoA (m) =P −1∑ wk ( m ) x A ( m − k )(14.2)k =0where wk(m) are the time-varying coefficients of an adaptive FIR filter andxˆ echoA ( m) is an estimate of the echo of speaker A on the line from speaker Bto speaker A. The residual echo signal, or the error signal, after echosubtraction is given bye(m) = y B (m) − xˆ echoA ( m)P −1= x B (m) + x echoA ( m ) − ∑ wk ( m ) x A ( m − k )k =0(14.3)403Adaptive Echo CancellationxA(m)xA(m) xA(m –1) xA(m–2)Echo/Speechclassifierw0w1…+AdaptationalgorithmHybridBwPw2FIR echosynthesis filterxA(m–P)Speaker B+^echoxA (m)echoe(m) = xB(m) + x~A (m)–+echoxB(m)+ xA(m)Figure 14.6 Illustration of an echo canceller using an adaptive FIR filter andincorporation a echo/speech classifier.For those time instants when speaker A is talking, and speaker B is listeningand silent, and only echo is present from line B to A, we haveechoe( m ) = ~x Aecho (m) = x echoA ( m) − xˆ A (m)P −1= x echoA ( m) − ∑ w k ( m ) x A ( m − k )(14.4)k =0x Aecho (m) is the residual echo.
An echo canceller using an adaptivewhere ~FIR filter is illustrated in Figure 14.6. The magnitude of the residual echodepends on the ability of the echo canceller to synthesise a replica of theecho, and this in turn depends on the adaptation algorithm discussed next.14.4.1 Echo Canceller Adaptation MethodsThe echo canceller coefficients wk(m) are adapted to minimise the energy ofthe echo signal on a telephone line, say from speaker B to speaker A.Assuming that the speech signals xA(m) and xB(m) are uncorrelated, theenergy on the telephone line from B to A is minimised when the echoechocanceller output xˆ echoA ( m) is equal to the echo x A ( m ) on the line. Theecho canceller coefficients may be adapted using one of the variants of therecursive least square error (RLS) or the least mean squared error (LMS)404Echo Cancellationadaptation methods.
One of the most widely used algorithms for adaptationof the coefficients of an echo canceller is the normalised least mean squareerror (NLMS) method. The time-update equation describing the adaptationof the filter coefficient vector isw (m) = w (m − 1) + µe( m )x (m) TA x A (m)x A ( m)(14.5)where xA(m)=[xA(m), ..., xA(m–P)] and w(m)=[w0(m), ..., wP–1(m)] are theinput signal vector and the coefficient vector of the echo canceller, and e(m)is the difference between the signal on the echo line and the output of theecho synthesiser. Note that the normalising quantity x (m) TA x A (m) is theenergy of the input speech to the adaptive filter. The scalar µ is theadaptation step size, and controls the speed of convergence, the steady-stateerror and the stability of the adaptation process.14.4.2 Convergence of Line Echo CancellerFor satisfactory performance, the echo canceller should have a fastconvergence rate, so that it can adequately track changes in the telephoneline and the signal characteristics.
The convergence of an echo canceller isaffected by the following factors:(a) Non-stationary characteristics of telephone line and speech. The echocharacteristics depend on the impedance mismatch between thesubscribers loop and the hybrids. Any changes in the connecting pathsaffect the echo characteristics and the convergence process. Also asexplained in Chapter 7, the non-stationary character and the eigenvaluespread of the input speech signal of an LMS adaptive filter affect theconvergence rates of the filter coefficients.(b) Simultaneous conversation. In a telephone conversation, usually thetalkers do not speak simultaneously, and hence speech and echo areseldom present on a line at the same time.
This observation simplifies theecho cancellation problem and substantially aids the correct functioningof adaptive echo cancellers. Problems arise during the periods when bothspeakers talk at the same time. This is because speech and its echo haveAdaptive Echo Cancellation405similar characteristics and occupy basically the same bandwidth.
Whenthe reference signal contains both echo and speech, the adaptationprocess can lose track, and the echo cancellation process can attempt tocancel out and distort the speech signal. One method of avoiding thisproblem is to use a speech activity detector, and freeze the adaptationprocess during periods when speech and echo are simultaneously presenton a line, as shown in Figure 14.6. In this system, the effect of aspeech/echo misclassification is that the echo may not be optimallycancelled out. This is more acceptable than is the case in echosuppressors, where the effect of a misclassification is the suppression andloss of a part of the speech.(c) The adaptation algorithm.
Most echo cancellers use variants of the LMSadaptation algorithm. The attractions of the LMS are its relatively lowmemory and computational requirements and its ease of implementationand monitoring. The main drawback of LMS is that it can be sensitive tothe eigenvalue spread of the input signal and is not particularly fast in itsconvergence rate. However, in practice, LMS adaptation has producedeffective line echo cancellation systems.
The recursive least square (RLS)error methods have a faster convergence rate and a better minimum meansquare error performance. With the increasing availability of low-costhigh-speed dedicated DSP processors, implementation of higherperformance and computationally intensive echo cancellers based onRLS are now feasible.14.4.3 Echo Cancellation for Digital Data TransmissionEcho cancellation becomes more complex with the increasing integration ofwireline telephone systems and mobile cellular systems, and the use ofdigital transmission methods such as asynchronous transfer mode (ATM)for integrated transmission of data, image and voice.
For example, in ATMbased systems, the voice transmission delay varies depending on the routetaken by the cells that carry the voice signals. This variable delay added tothe delay inherent in digital voice coding complicates the echo cancellationprocess.The 2-wire subscriber telephone lines that were originally intended tocarry relatively low-bandwidth voice signals are now used to providetelephone users with high-speed digital data links and digital services suchas video-on-demand and internet services using digital transmission406Echo CancellationTransmitterBTransmitterEchocancellarABAEchocancellernear-endechoReceiverReceiverfar-end echoFigure 14.7 Echo cancellation in digital modems using 2-wire subscriber's loop.methods such as the asynchronous digital subscriber line (ADSL).Traditionally, the bandwidth of the subscribers line is limited by low-passfilters at the core network to 3.4 kHz.
Within this bandwidth, voice-bandmodems can provide data rates of around 30 kilobits per second (kbps).However the copper wire itself has a much higher usable bandwidthextending into megahertz regions, although attenuation and interferenceincrease with both the frequency and the length of the wire. Using advancedsignal processing and modulation schemes methods such as ADSL canachieve a 10 megabits per second data rate over 240 MHz bandwidth ofsubscriber’s twisted wire line.Figure 14.7 shows a system for providing a full-duplex digital serviceover a 2-wire subscriber’s loop. To provide simultaneous transmission ofdata in both directions within the same bandwidth over the subscriber’s line,echo cancellation is needed.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
