Echo Cancellation (779800), страница 3
Текст из файла (страница 3)
The echoes on a line consist of the near-endecho which loops back at the first or the near hybrid, and the far-end echowhich is the signal that loops back at a hybrid some distance away. Themain purpose of the echo canceller is to cancel the near-end echo. Since thedigital signal coming from a far-end may be attenuated by 40–50 dB, thenear echo on a high speed data transmission line can be as much as 40–50dB above the desired signal level. For reliable data communication the echocanceller must provide 50–60 dB attenuation of the echo signal so that thesignal power remains at 10 dB above the echo.14.5 Acoustic EchoAcoustic echo results from a feedback path set up between the speaker andthe microphone in a mobile phone, hands-free phone, teleconference orhearing aid system.
Acoustic echo is usually reflected from a multitude ofdifferent surfaces, such as walls, ceilings and floors, and travels throughAcoustic Echo407different paths. If the time delay is not too long then the acoustic echo maybe perceived as a soft reverberation, and may add to the artistic quality ofthe sound. Concert halls and church halls with desirable reverberationcharacteristics can enhance the quality of a musical performance. However,acoustic echo is a well-known problem with hands-free telephones,teleconference systems, public address systems, mobile phones, and hearingaids, and is due to acoustic feedback coupling of sound waves between theloudspeakers and microphones. Acoustic echo can result from acombination of direct acoustic coupling and multipath effect where thesound wave is reflected from various surfaces and then picked up by themicrophone.
In its worst case, acoustic feedback can result in howling if asignificant proportion of the sound energy transmitted by the loudspeaker isreceived back at the microphone and circulated in the feedback loop. Theoverall round gain of an acoustic feedback loop depends on the frequencyresponses of the electrical and the acoustic signal paths. The undesirableeffects of the electrical sections on the acoustic feedback can be reduced bydesigning systems that have a flat frequency response. The main problem isin the acoustic feedback path and the reverberating characteristics of theroom.
If the microphone–speaker–room system is excited at a frequencywhose loop gain is greater than unity then the signal is amplified each timeit circulates round the loop, and feedback howling results. In practice, thehowling is limited by the non-linearity of the electronic system.There are a number of methods for removing acoustic feedback. Onemethod for alleviating the effects of acoustic feedback and the roomreverberations is to place a frequency shifter (or a phase shifter) in theelectrical path of the feedback loop. Each time a signal travels round thefeedback loop it is shifted by a few hertz before being re-transmitted by theloudspeaker.
This method has some effect in reducing the howling but it isnot effective for removal of the overall echo of the acoustic feedback.Another approach is to reduce the feedback loop-gain at those frequencieswhere the acoustic feedback energy is concentrated.
This may be achievedby using adaptive notch filters to reduce the system gain at frequencieswhere acoustic oscillations occur. The drawback of this method is that inaddition to reducing the feedback the notch filters also result in distortion ofthe desired signal frequencies.The most effective method of acoustic feedback removal is the use ofan adaptive feedback cancellation system. Figure 14.8 illustrates a model ofan acoustic feedback environment, comprising a microphone, a loudspeakerand the reverberating space of a room.
The z-transfer function of a linearmodel of the acoustic feedback environment may be expressed as408Echo CancellationAcoustic feedbackpath A(z)yf(m)x(m)+Microphone-Loudspeakery(m)G(z)Figure 14.8 Configuration of a feedback model for a microphone–loudspeaker–room system.H(z) =G(z)1 − G(z )A(z )(14.6)where G(z) is the z-transfer function model for the microphone–loudspeakersystem and A(z) is the z-transfer function model of reverberations and multipath reflections of a room environment. Assuming that the microphone–loudspeaker combination has a flat frequency response with a gain of G,Equation (14.6) can be simplified toH ( z) =G1− G A( z )(14.7)Note that in Equation (14.6), owing to the reverberating character of theroom, the acoustic feedback path A(z) is itself a feedback system.
Thereverberating characteristics of the acoustic environment may be modelledby an all-pole linear predictive model, or alternatively a relatively long FIRmodel.The equivalent time-domain input/output relation for the linear filtermodel of Equation (14.7) is given by the following difference equation:P−1y(m) = G ∑ ak (m)y(m − k) + G x(m)k =0(14.8)409Acoustic Echo^x(m)Room wallsLoud speakerAcoustic feedbacksynthesizerAdaptationalgorithmmicro phone^y (m)f–+y(m) = x(m) + yf (m)Figure 14.9 Illustration of adaptive acoustic feedback cancellation in aconference room environment.where ak(m) are the coefficients of an all-pole linear feedback model of thereverberating room environment, G is the microphone–loudspeakeramplitude gain factor, and x(m) and y(m) are the time domain input andoutput signals of the microphone–loudspeaker system.Figure 14.9 is an illustration of an acoustic feedback cancellationsystem.
In an acoustic feedback environment, the total input signal to themicrophone is given as the sum of any new input to the microphone x(m)plus the unwanted acoustic feedback signal yf(m):y ( m) = x ( m) + y f ( m)(14.9)The most successful acoustic feedback control systems are based onadaptive estimation and cancellation of the feedback signal. As in a lineecho canceller, an adaptive acoustic feedback canceller attempts tosynthesise a replica of the acoustic feedback at its output asyˆ f (m) =P−1∑ aˆk (m)y(m − k )k =0(14.10)410Echo CancellationAcoustic feedbackpathInputmicrophoneHearing aidprocessorOutputspeakerAdaptionalgorithmAcoustic feedbacksynthesiserFigure 14.10 Configuration of an acoustic feedback canceller incorporated in ahearing aid system.The filter coefficients are adapted to minimise the energy of an error signaldefined as(14.11)e(m) = x(m) + y f (m) − yˆ f (m)The adaptation criterion is usually the minimum mean square error criterionand the adaptation algorithm is a variant of the LMS or the RLS method.The problem of acoustic echo cancellation is more complex than line echocancellation for a number of reasons.
First, acoustic echo is usually muchlonger (up to a second) than terrestrial telephone line echoes. In fact, thedelay of an acoustic echo is similar to or more than a line echo routed via ageostationary satellite system.The large delay of an acoustic echo path implies that impractically largefilters on the order of a few thousand coefficients may be required. Thestable and speedy adaptation of filters of such length presents a difficultproblem. Secondly, the characteristics of an acoustic echo path is more nonstationary compared with that of a telephone line echo. For example, theopening or closing of a door, or people moving in or out of a room, cansuddenly change the acoustic character of a conference room.
Thirdly,acoustic echoes are due to signals reflected back from a multitude ofdifferent paths, off the walls, the floor, the ceiling, the windows etc. Finally,the propagation and diffusion characteristics of the acoustic space of a roomis a non-linear process, and is not well approximated by a lumped FIR (orSub-Band Acoustic Echo Cancellation411IIR) linear filter.
In comparison, it is more reasonable to model thecharacteristics of a telephone line echo with a linear filter. In any case, foracoustic echo cancellation, the filter must have a large impulse response andshould be able to quickly track fast changes in echo path characteristics.An important application of acoustic feedback cancellation is in hearingaid systems. A hearing aid system can be modelled as a feedback system asshown in Figure 14.10. The maximum usable gain of a hearing aid system islimited by the acoustic feedback between the microphone and the speaker.Figure 14.10 illustrates the configuration of a feedback canceller in ahearing aid system.
The acoustic feedback synthesiser has the same input asthe acoustic feedback path. An adaptation algorithm adjusts the coefficientsof the synthesiser to cancel out the feedback signals picked up by themicrophone, before the microphone output is fed into the speaker.14.6 Sub-Band Acoustic Echo CancellationIn addition to the complex and varying nature of room acoustics, there aretwo main problems in acoustic echo cancellation. First, the echo delay isrelatively long, and therefore the FIR echo synthesiser must have a largenumber of coefficients, say 2000 or more. Secondly, the long impulseresponse of the FIR filter and the large eigenvalue spread of the speechsignals result in a slow, and uneven, rate of convergence of the adaptationprocess.A sub-band-based echo canceller alleviates the problems associatedwith the required filter length and the speed of convergence.
The sub-bandbased system is shown in Figure 14.11. The sub-band analyser splits theinput signal into N sub-bands. Assuming that the sub-bands have equalbandwidth, each sub-band occupies only 1/N of the baseband frequency, andcan therefore be decimated (down sampled) without loss of information. Forsimplicity, assume that all sub-bands are down-sampled by the same factorR.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
