Loading...
Please wait, while we are loading the content...
Similar Documents
Acoustic echo and noise control: where do we come from - where do we go to?
| Content Provider | Semantic Scholar |
|---|---|
| Author | Hänsler, Eberhard |
| Copyright Year | 2001 |
| Abstract | In this paper we describe a short history of acoustic echo and noise control systems and of the International Workshop on Acoustic Echo and Noise Control. We also try to look into the future. 1. WHERE DO WE COME FROM? Acoustic echo and noise control units as used in handsfree communication systems are comprised of three subunits: A loss control circuit, an adaptive filter parallel to the loudspeaker–enclosure–microphone system (LEMS) and also an adaptive filter within the path of the output signal (see Fig.1). Their functions are obvious: The loss control circuit attenuates the input and/or output signal such that the communication loop remains stable. In addition, echoes caused by the acoustic transfer path from loudspeaker to microphone do not impede a communication between two users. The adaptive filter arranged in parallel to the LEMS is able to cancel echoes according to the degree to which it is matched to the LEMS. The filter within the path of the output signal is used to attenuate remaining echoes and background noise. In the early days of acoustic echo and noise control a socalled center clipper – a nonlinear circuit – took the place of this filter. Of these units, the loss control circuit has the longest history in hands-free communication systems. In its most simple form it reduces the usually full-duplex communication system to a half-duplex one by simply switching input and output lines on and off alternately. Besides preventing howling and suppressing echoes, any natural conversation was prevented, too. A device for hands-free telephone conversation using voice switching was presented as recently as in 1957 [1]. The introduction of a center clipper in 1974 [2] meant a noticeable improvement. Laboratory experiments applying an adaptive filter for acoustic echo suppression were reported in 1975 [3]. Figure 1: General structure of acoustic echo and noise control systems The series of biennial international workshops on acoustic echo and noise control started in 1989 in Berlin. Even at that time, hands-free telephone conversations were still made possible by the help of loss control circuits only. The reason for this fact was very simple: the lack of processing power at an affordable cost. The workshop series was continued in several European countries and made its way to the USA in 1999. The main interest at the first workshop was focused on adaptive algorithms for acoustic echo cancelling. On the one hand, it was quite clear that the comfort of hands-free communication units would have to be improved. On the other hand, the already foreseeable increase of available processing power would enable the implementation of high-order adaptive filters in the near future. First laboratory implementations of filters using the Least Mean Square Algorithm clearly indicated that this algorithm without any additional support could not solve the problem. Therefore, complexity reduction and stability improvement of the Recursive Least Square Algorithm was one of the main discussion issues. It seems worthwhile to mention that the class of Fast Newton Filters was introduced at the first workshop. In addition, an astonishing variety of topics that are still under discussion today were presented at the 1989 workshop: Echo cancelling in the frequency domain, subband systems, step-size control, adaptive antennas, noise reduction based on microphone arrays and post filtering, nonlinear techniques for echo cancelling, problems in measuring, etc. It seems strange that it took three more workshops before the problem of stereophonic echo cancelling appeared on the agenda in 1995. Knowing just the titles of the early workshops one may ask whether Ben Joseph Akiba with his ”We have seen it all, nothing ever changes” was right. The answer is clear: ”Definitely not!” Tremendous progress has been made on all those topics. This is documented in a multitude of papers. An attempt to reference even only a few of them would blow the scope of this contribution. Therefore, only one overview paper and two books published recently are cited here [4, 5, 6]. Each one contains a large number of references to all topics in acoustic echo and noise control. 2. WHERE ARE WE NOW? Powerful and affordable acoustic echo and noise control units are available now. Their performance is satisfactory, especially if compared to solutions in other voice processing areas like speech recognition or speech to text translation. The fact that echo and noise control systems have not yet entered the market on a large scale seems not to be a technical but a marketing problem: A customer who buys a high quality echo and noise suppression system pays for the comfort of his communication partner. If she/he decides to economize by using a very simple system the partner at the far end is typically too polite to tell her/him about the poor speech quality the system is producing. ”Be nice to your communication partner” might be an effective advertising slogan. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.iwaenc.org/proceedings/2001/main/data/haensler.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
