A novel adaptation scheme in the NLMS algorithm for echo cancellation

Abstract

This paper presents a new normalized least mean square (NLMS) adaptation scheme for echo cancellation. The scheme combines the advantages of the conventional Geigel algorithm with some initiative ideas. The Geigel is used for rough detection of double talk or echo path change within a short delay period, while a new control mechanism is proposed for further accurate discrimination of double talk and echo path change and the corresponding real-time updating. Experiments show that stable performance is ensured during abrupt changes. Efficiency as well as effectiveness is increased.