Christina Breining

A robust fuzzy logic-based step-gain control for adaptive filters in acoustic echo cancellation

Step-gain control is crucial for acoustic echo cancelers. Most double-talk detectors and step-gain estimators become unreliable in certain situations. In this contribution, we combine a step-gain estimation method with a double-talk detector by fuzzy logic. This control is robust and efficient while requiring only very little additional complexity.

Delay-free low-cost step-gain estimation for adaptive filters in acoustic echo cancellation

Abstract Step-gain control for acoustic echo cancellation filters is extremely important for the speech quality enhancement achieved by the adaptive algorithm. A very useful step-gain control method was proposed as early as 1982 in Yamamoto and Kitayama (Trans. IECE Japan 65 (1) (January 1982) 1–8), but it introduces additional delay in the signal path. To avoid this in time-critical transmission systems like wired telephony, a second parallel filter was suggested so that the delay could be eliminated form the signal transmission path. Unfortunately, this structure obviously requires double filtering effort. As another principle disadvantage of both realizations, the adaptation tends to freeze, i.e. the filter cannot track the time-variant room impulse response. In this contribution, we present a modified step-gain estimation method. Although it only requires one filter, it neither introduces a delay, nor does it exhibit the freezing behavior. This method therefore provides a complete step-gain control system which can efficiently be implemented in low-cost real-time applications.

Freisprechen : Die Jagd nach akustischen Echos

A microphone located in the same enclosure with the loudspeaker of an electro acoustic system picks up the loudspeaker output signal and its echoes. This may lead to instabilities. In addition, hearing delayed echoes of the own speech impedes a conversation. This paper deals with the hands-free telephone problem. The problem is explained and existing solutions are pointed out. Finally we focus onto two subproblems. Their solutions promise considerable improvements.