摘要 |
An echo canceller is described with improved performance. The echo canceller is capable of maintaining audio quality with less consumption of processor cycles by initially segmenting coefficients corresponding to successive audio samples of a delay line into a predetermined number of segments. The coefficients are used in computing an echo estimate. Upon segmenting the coefficients, certain segments are assigned to a first segment group, where the coefficients of that group are updated more frequently than coefficients assigned to one or more of the remaining groups. In one embodiment, this grouping is achieved by periodically computing a cumulative magnitude of the coefficients of each segment for all coefficients of the segments and subsequently placing a lesser number of segments with the highest cumulative coefficient magnitudes into the first segment group. Of course, in lieu of cumulative magnitude calculations, the energy of the coefficients may be computed. In any event, the remainder of the segments may be placed into a second segment group or multiple groups that are updated less frequently than the first segment group.
|