Development and implementation of echo cancellation techniques
MetadataShow full item record
In telecommunication, the echo canceler using an adaptive filter is the most effective technique for controlling echoes. Because the precision of speech detection significantly affects how the echo path is modeled, this thesis presents the structures of two echo cancelers and two speech detectors. Of the two echo cancelers, one is based on echo path loss (EPL) of 6 dB and the other uses two different echo path models. The speech detectors using a three variable-length windows technique are implemented as the far-end speech detector. Double-talk detectors based on EPL and echo return loss enhancement (ERLE) are also developed and evaluated. Different adaptation algorithms on the adaptive filter are discussed. The normalized partial block least-mean-square (NPBLMS) algorithm is discussed and compared with the normalized least-mean-square (NLMS) algorithm. The results show that the former algorithm achieves a better performance. The center clipper is also introduced and implemented to compress the residual echo. The overall programs have been implemented on the digital signal processor TMS320C54x, where the partial maximum algorithm is also developed for the echo canceler in order to save more computation on DSP program.