Speech recognition in reverberant settings is challenging for people with hearing loss
Many Hearing Care Professionals will have heard their clients report about the challenges of listening in reverberant environments. At Eriksholm Research Centre we have heard this too, and this gave us the notion that people with hearing loss might find reverberation more harmful than their normal-hearing peers.
We decided to investigate this notion in terms of an MSc project*. The study comprised two experiments, both conducted in echo-free as well as in reverberant conditions, and with two groups of hearing-impaired and normal-hearing listeners, respectively.
In the first experiment, Speech Reception Thresholds (SRTs) were measured. This is the Signal-to-Noise Ratio (SNR) in decibel at which 50% of spoken sentences can be repeated correctly by the test subject. The higher the SRT, the harder it is for the test subject to understand speech in general.
The two main observations from this test were:
SRTs were about 4 dB higher (worse performance) in the reverberant room than in the echo-free room, and this difference was the same for the hearing-impaired and the normal-hearing listeners.
SRTs were in general about 7 dB lower (better performance) for the normal-hearing listeners compared to the hearing-impaired listeners.
The first observation suggests that hearing-impaired and normal-hearing listeners are penalized by reverberation by the same amount – contrary to the notion stated above. However, the second observation pinpoints that the listener groups were tested in very different circumstances in terms of the SNR: the mean SRT for the hearing-impaired listeners was about 3 dB (difficult but common), whereas it was -4 dB (extremely difficult and uncommon**) for the normal-hearing listeners.
In the second experiment, consonant recognition was measured. Both groups were tested in the same condition (no background noise). The main observations from this test were:
The normal-hearing listeners performed close to perfect (100% correct recognition) in both the echo-free and the reverberant room.
The hearing-impaired listeners scored on average 85% correct in the echo-free room and 80% (worse performance) in the reverberant room.
Combining the two sets of results explains the effect of reverberation on speech recognition: In typical listening scenarios (in terms of SNR) and for normal-hearing people, reverberation merely changes the listening conditions from easy to slightly challenging. In contrast, for people with hearing loss, typical listening scenarios – even without reverberation – are already difficult. Add reverberation, and now the conditions have turned to impossible. In this way, the same acoustical change (echo-free vs. reverberant) can have a markedly more harmful perceptual effects for hearing-impaired than for normal-hearing listeners – in good agreement with the reports from the clients.
This study once again shows that while understanding the acoustics is important, it is not enough – we need to understand the individual listener, from auditory periphery to the brain, in order to identify the challenges and invent solutions.
*Breitsprecher C (2011). Effects of reverberation on speech intelligibility in normal-hearing and hearing-impaired listeners. Eriksholm research report 2011-27-E.
**Smeds K, Wolters F, Rung M (2012). Realistic signal-to-noise ratios. International Hearing Aid Research Conference (IHCON), Lake Tahoe, CA, USA, poster presentation.