A frequency selective adaptive algorithm

A frequency selective adaptive algorithm (FSAA) is proposed. This frequency-domain algorithm is based on a weighted least squares (WLS) criterion, and is developed for short-time pole-zero selective spectral matching. First, we give the optimal solution for the parameters of the selective spectral envelope characterizing an N-point windowed sequence. A gradient algorithm, namely the FSAA, is then given for sample-by-sample and block-by-block updates of the envelope. The FSAA can be used in speech processing applications, such as efficient harmonic voice representation and vocal tract characterization. Results are given for the spectral matching (selective and non-selective) of the short-time speech spectrum. These are compared to those obtained using the classical adaptive predictor.