Improving speech understanding for face-to-face communication in noise when wearing hearing protectors

Active hearing protection devices (eHPDs) are commercially available that enhance communication at low noise levels by amplifying both speech and noise. A study has been performed to develop algorithms for improving face-to-face speech communication in both low- and high-level noises when wearing eHPDs. The (mixed) environmental noise and speech at frequencies from 200 to 6000 Hz are divided into 16 or 24 contiguous subbands corresponding to 1.5 and 1 times the bandwidth of auditory filters, respectively. Signals are processed in the time domain with the overall group delay intended to maintain synchronization between speech and lip movements. The temporal modulation in each subband is band-limited to 2–16 Hz using zero time delay, moving detrend, high-pass, and moving average low-pass filters. This modulation is compared to the normalized unfiltered modulation in the subband to provide an estimate of the speech signal-to-noise ratio (eSNR). A threshold is selected for the eSNR above which time signals in subbands are combined to produce the instantaneous output and below which subband signals are attenuated. Subjects “wearing” a simulated HPD and listening to speech in industrial noise experience, on average, a 12% increase in intelligibility when using the algorithm. Work supported by NIOSH and the Alpha Foundation.