Echoes and Noise with Seamless Acoustic Man-Machine Interfaces - The Challenge Persists

Within a general framework for seamless acoustic man-machine interfaces, common problems for different classes of applications are identified. Removing loudspeaker echoes, noise, and reverberation of local sources from microphone signals, and conditioning loudspeaker signals to eliminate reverberation and noise at a listener’s ears constitute problems of system identification and signal separation. The corresponding signal processing techniques concentrate on adaptive filtering and signal separation methods based on temporal and spatial information. Major challenges are posed by the time-variance of the acoustic environment, the nonstationarity of the signals involved, and most of all by the blindness of signal processing with regard to the signals at the human’s mouth and ears. 1. ACOUSTIC MAN-MACHINE INTERFACES As a general framework for a broad class of acoustic manmachine interfaces, we consider the setup according to Fig.1. The interface consists of a digital signal processing unit for broadcasting and recording with multiple loudspeakers and microphones, respectively. Signal processing may include mixing, so that the number of input channels to the broadcasting and the number of output signals of the recording will in general not be equal to the number of transducers. A possible feedback path between the recording part and the broadcasting part closes a loop which is of concern for some important applications. Obviously, the scenario can be generalized to multiple desired sources/listeners and multiple noise sources in the local environment, as well as it includes important special cases of single-channel signals or singlesensor electroacoustic transduction. At a local listener’s ears, the sum of the reverberated loudspeaker signals and local noise signals will be heard instead of the original loudspeaker signals. On the other hand, the multi-microphone interface captures reverberated versions of the desired source signal and local noise sources as Noise