A conferencing unit reduces or eliminates percussive noise in audio that it outputs to a far-end. The unit filters near-end audio into a plurality of bands and calculates instantaneous energies for each band. The energies in first and second ranges of the bands are summed, and a difference between the two sums is compared to a threshold value to determine whether speech is present in the near-end audio received. The first range is preferably a mid-band range of frequencies of 300 to 600-Hz, while the second range is preferably an outer-band range of 100-Hz to 300-Hz and 600-Hz to 14-kHz. Based on the determination, the conferencing unit adjusts an output level of the near-end audio to reduce or eliminate any percussive noise in output audio while there is not a substantial amount of speech in the near-end audio.