A new structure for subband adaptive filtering

A subband adaptive filter structure (SAFS) has been proposed by Petraglia, Alves and Diniz (PAD). For high-order adaptive filters, it requires M/3 times (M=the number of subbands) less computations than the full band LMS algorithm. We present a modified algorithm for this SAFS in which the optimum step sizes are automatically determined instead of trial and error. The performance of the modified algorithm is compared with the PAD algorithm. A new structure (RPS) for SAP that has significantly reduced computational load (CL) is also proposed.