New FIR adaptive filtering algorithm based on quantized gradients and least-squares convergence factors

A new FIR adaptive filtering algorithm, using quantized gradients with a variable convergence factor that minimizes an exponentially weighted sum square error, is proposed. The proposed algorithm converges about as fast as the optimum block adaptive shifting algorithm does. However, when effective reusing data length due to the exponential weighting is a power-of-two number, the proposed algorithm requires a much smaller number of multiplications and divisions.