METHOD AND APPARATUS FOR NETWORK SPEECH ENHANCEMENT

An apparatus for enhancing intelligibility of a voice signal in a noisy environment includes a first noise estimator which estimates a far-end noise component from the far-end signal, and a second noise estimator which estimates a near-end noise component from the near-end signal. A noise reduction calculator determines a noise reduction gain from the estimated far-end noise, and an echo gain calculator determines an echo control gain. A master gain calculator combines both the echo control gain and the noise control gain into a master gain which is applied to the far-end signal. A comfort noise generator applies the pre-set minimum threshold of the master gain and constantly matches the spectrum of the far-end noise to synthesize a background noise for selectively mixing onto the far-end signal when echo is determined. Echo is determined based on a comparison of the near-end signal and the far-end signal spectra and by compensating the total gain applied. An automatic level controller adjusts the far-end signal responsive to the far-end signal level, and a noise compensator adjusts the far-end signal level responsive to the near-end noise, both adjustments being responsive to the identification of the far-end signal as either speech or noise.