A pole-zero adaptive algorithm is presented for selective spectral matching. This algorithm is capable of forming highly selective spectral envelopes for accurate fitting of frequency components. This adaptive technique can be employed to reduce the data rate in transform coding of speech, by efficiently 'packing' magnitude and phase information on a spectral envelope. Pole-zero algorithms have been developed and employed in system identification, as well as in spectral matching. In this work, a frequency-selective adaptive process is proposed for the spectral matching of the harmonic frequency components of speech. This process is based on the equation-error adaptive model.