Chair: Dennis Morgan, AT&T Bell Laboratories (USA)
Steven L. Gay, AT&T Bell Laboratories (USA)
Sanjeev Tavathia, AT&T Bell Laboratories (USA)
This paper discusses a new adaptive filtering algorithm called fast affine projections (FAP). FAP's key features include LMS like complexity and memory requirements (low), and RLS like convergence (fast) for the important case where the excitation signal is speech. Another of FAP's important features is that it causes no delay in the input or output signals. In addition, the algorithm is easily regularized resulting in robust performance even for highly colored excitation signals. The combination of these features make FAP an excellent candidate for the adaptive filter in the acoustic echo cancellation problem. A simple low complexity numerical stablization method for the algorithm is also introduced.
Shigenobu Minami, Toshiba Corporation (JAPAN)
This paper proposes a new stereophonic echo canceler which can be realized by using single adaptive filter. Since most of conversations in telecommunication are single talking, Pseudo-Stereophonic Echo Canceler (PST-EC), which is a monaural echo canceler and has only been applied to pseudo-stereophonic speech sound produced by attenuating and delaying monaural sound, is applicable to stereophonic communication. To cope with double talking, stereophonic speech sound is classified into strongly correlated component (single talking) and less correlated component (double talking). Therefore, PST-EC and echo suppressors can be applied to the former and the latter component, respectively.
Christiane Antweiler, Aachen University of Technology (GERMANY)
Horst-Gunter Symanzik, Aachen University of Technology (GERMANY)
For many applications such as acoustic echo compensation, adaptive noise reduction or acoustic feedback control it is of great interest to simulate reproducibly a real, time variant room. One approach to describe the transient behavior of a room is the generation of a physical room model, e.g. [1]. To identify the variation of a room impulse response with time an alternative concept is presented, which uses the normalized least mean square (NLMS) algorithm excited by perfect sequences. The proposed general concept is capable to efficiently simulate the fluctuations of the room impulse response. The time variant simulation can be performed without the necessity to store large amounts of data by storing only the reaction of the unknown system instead of all sets of filter coefficients. The practical aspects of the new concept are pointed out for an acoustic echo control application.
A.N. Birkett, Carleton University (CANADA)
R.A. Goubran, Carleton University (CANADA)
One of the limitations of linear adaptive echo cancellers is nonlinearities which are generated mainly in the loudspeaker. The complete acoustic channel can be modelled as a nonlinear system convolved with a linear dispersive channel. Two new acoustic echo canceller models are developed to improve nonlinear performance. The first model consists of time-delay neural network (TDNN) and the second model consists of a memoryless neural network followed by an adaptive Normalized Least Mean Square (NLMS) structure. Simulations demonstrate that both neural network based structures improve the Echo Return Loss Enhancement (ERLE) performance. Experimental results using the TDNN improved the ERLE by 10 dB at low to medium loudspeaker volumes.
O. Tanrikula, Imperial College of Science, Technology and Medicine (UK)
B. Baykal, Imperial College of Science, Technology and Medicine (UK)
A.G. Constantinides, Imperial College of Science, Technology and Medicine (UK)
J.A. Chambers, Imperial College of Science, Technology and Medicine (UK)
P.A. Naylor, Imperial College of Science, Technology and Medicine (UK)
All-pass Polyphase Networks (APN) are particularly attractive for Acoustical Echo Cancellation (AEC) arranged in sub-bands. They provide lower inter-band aliasing, delay and computational complexity than their FIR counterparts. Moreover, APNs achieve higher Echo Return Loss Enhancement (ERLE) performance and faster convergence than full-band processing. In this paper, the finite precision implementation of APNs is addressed. A procedure is presented for re-optimising the all-pass coefficients for finite precision operation. Robust finite precision implementation of a prototype low-pass filter is discussed. The results of a set of AEC experiments are reported with full and 16-bit precision implementation.
R. Martin, Aachen University of Technology (GERMANY)
J. Altenhoner, Aachen University of Technology (GERMANY)
This contribution presents new adaptive algorithms for acoustic echo control and noise reduction which employ one, two, or possibly more microphone signals. The new algorithms accommodate high echo attenuation and lead to implementations with reduced complexity. These algorithms combine a conventional FIR echo canceller with a second NLMS-adapted FIR filter which attenuates residual echoes. The paper presents a one-microphone system with improved echo attenuation and a two- microphone system which attenuates acoustic echoes as well as ambient noise and near end speech reverberation. The algorithms can be interpreted as a frequency selective generalization of the well known voice controlled switch. This paper explains the algorithms and presents experimental results in real acoustic environments.
Peter Heitkamper, Technische Hochschule Darmstadt (GERMANY)
This contribution deals with the application of hands-free algorithms in wide-band telephone systems with limited computing resources. The basis is the combination of an acoustic echo canceller and an adaptive gain control method. The paper describes how the effect of the echo canceller can be optimized without increasing the computational expense. This is achieved by extending the length of the adaptive filter at the cost of a reduced affected frequency range. The study shows to which extent one can concentrate on the low frequency portion of the acoustic echoes in a wide-band application, if a suitable additional gain control method is available. The optimization is accomplished using a simple room model. Real- time measurements were carried out using an implementation of the complete hands-free system on a single DSP.
Akihiko Sugiyama, NEC Corporation (JAPAN)
Frederic Landais, NEC Corporation (JAPAN)
This paper proposes a new subband adaptive filtering algorithm with intersubband tap assignment. The number of taps for each subband adaptive filter is adaptively controlled based on a sum of squared coefficients to the impulse-response tail of the unknown system. The sum is weighted by the input reference signal power to reflect its power imbalance. For rapid convergence of intersubband tap assignment, a time-varying number is newly introduced as the tap-redistribution step- size. This variable is controlled based on repeated redistribution of the maximum number of taps to the same subband. For colored signals, the proposed algorithm reduces the convergence time for intersubband tap-assignment by more than 60% compared with a fixed tap-redistribution step-size. It is insensitive to parameter-selection and has good capability to track a path change. It is also shown that the proposed algorithm is effective for real speech signals.
Alex Stephenne, Universite du Quebec (CANADA)
Benoit Champagne, Universite du Quebec (CANADA)
Time delay estimation (TDE) between the signals received by two or more spatially separated microphones can be used as a means for the passive localization of the dominant talker in applications such as audio-conference. However, in a recent study, it has been shown that reverberation can have disastrous effects on TDE performance. In this paper, we develop and evaluate a new cepstral prefiltering technique which can be applied on the microphone signals before the actual TDE in order to obtain a more accurate estimate of the position of a source in a typical reverberant environment. The performance of a TDE system with and without cepstral prefiltering is investigated under controlled conditions via Monte-Carlo simulations. The results clearly demonstrate the beneficial effects of the new cepstral prefiltering technique on TDE performance (i.e., reduction of the bias, variance and of the number of anomalous estimates).
Suehiro Shimauchi, NTT Human Interface Laboratories (JAPAN)
Shoji Makino, NTT Human Interface Laboratories (JAPAN)
The effect that cross-correlation between stereo signals has on a stereo echo canceller is studied and a new stereo projection echo canceller is proposed that can identify the true echo path impulse responses. This echo canceller accelerates filter coefficient error convergence by adding the variations in the cross-correlation to stereo signals or by utilizing the fact that the cross-correlation between stereo signals varies slightly in actual teleconferencing situations. Computer simulations demonstrate that this echo canceller can reduce the filter coefficient error much faster than a conventional stereo normalized least mean squares (NLMS) echo canceller.