Effect of frequency selectivity on the perception of spectrally/temporally interrupted speech.

Abstract

Compared to normal hearing (NH), hearing impaired (HI) listeners have shown significantly less masking release, even when the speech and noise were amplified to compensate for their hearing loss [J. Nelson (2004)]. According to Jin (2003), the reduced masking release of HI listeners was highly related to auditory filter bandwidths, which is an indicator of frequency selectivity. As a follow-up, the present study examined whether the reduced frequency selectivity might influence on spectrally-based perceptual strategies for speech recognition of NH and HI listeners in complex noise. We hypothesized that even with mild hearing loss, HI listeners would show significantly different spectral weight which would be significantly correlated with individual’s hearing loss. In this study, signal processing was carried out in the digital domain and involved splitting of the signal spectra into five frequency bands for both speech and noise (band 1: 100–250 Hz, band 2: 250–750 Hz, band 3: 750–1750 Hz, band 4: 1750–3750 Hz, and band 5: 3750–7750 Hz). The subject’s task was to identify the sentences heard in quiet, steady, and fluctuating noise in each band. The hypothesis will be examined and its implication on speech perception of HI listeners and signal processing strategy for the amplification system.