论文部分内容阅读
A novel parallel delayed least-mean-square (PDLMS) algorithm is proposed by introduc-ing the parallel processing method into delayed LMS (DLMS) algorithm. Compared with DLMS, the al-gorithm presented has the property of smaller delay, higher throughput rate and faster convergence speed, while it also exhibits some de-correlation effect for the correlated input sequence. These prop-erties make it more suitable for the cases of high order filter with high convergence speed. At the same time, it can be mapped onto the high-speed and/or high-pipelined hardware structure directly.
A novel parallel delayed least-mean-square (PDLMS) algorithm is suggested by introduc-ing the parallel processing method into delayed LMS (DLMS) algorithm. Compared with DLMS, the al-gorithm presented has the property of smaller delay, higher throughput rate and faster convergence speed, while it also exhibits some de-correlation effect for the correlated input sequence. These prop-erties make it more suitable for the cases of high order filter with high convergence speed. At the same time, it can be mapped onto the high-speed and / or high-pipelined hardware structure directly.