Background. The starting point of the present research project is the relatively recent demonstration of the crucial role played by temporal-envelope cues (that is the slowest fluctuations in amplitude of speech sounds, < 50 Hz) and temporal fine structure cues (that is the fastest fluctuations in amplitude of speech sounds, 0.5-5 kHz) in speech identification.
Our first research project conducted with normal-hearing listeners has demonstrated that (i) nearly perfect intelligibility can also be achieved with mainly temporal fine structure cues provided listeners are given substantial training (Gilbert & Lorenzi, 2006), and that (ii) temporal fine structure cues play a crucial role in sub-optimal listening conditions, that is in the presence of background noise, concurrent sounds, or for other forms of signal distortions (e.g., interruptions). More precisely, these studies demonstrate that masking release (the important release from masking occuring when speech items are presented in a background noise showing slow envelope fluctuations) is abolished when fine structure cues are degraded (Füllgrabe, Berthommier & Lorenzi, 2006). This result is important because it explains why listeners with cochlear damage and cochear implantees show poorer-than-normal speech intelligibility in noise or in reverberant conditions (fine structure cues being required to segregate speech from noise).
Three research projects conducted by the GRAEC group currently investigate the speech perception in sub-optimal listening conditions in hearing-impaired listeners, and the specific role of temporal fine structure cues.
Project #1. An important result obtained recently by our research group confirms and extends these results in hearing impaired patients with moderate cochlear hearing loss (Lorenzi, Gilbert, Carn, Garnier, & Moore, 2006). These results were obtained with speech items processed in order to degrade either envelope or fine structure information. These original speech stimuli (vowel-consonant-vowel stimuli) were processed by filtering them into 16 adjacent frequency bands. The signal in each band was then processed using the Hilbert transform so as to preserve either the envelope or the temporal fine structure. The band signals were then recombined and the stimuli were presented to subjects for identification. The results reveal that hearing-impaired listeners show (i) normal ability to perceive and use temporal envelope information in a speech identification task (this explains the quasi-normal intelligibility usually observed in quiet in listeners with moderate hearing loss), and (ii) a massively degraded ability to perceive and use temporal fine structure information in a speech identification task (this explains the poorer-than-normal intelligibility usually observed in noise in listeners with moderate hearing loss).
We currently extend this initial study by assessing the ability to identify speech on the basis of temporal fine structure cues only in different groups of adults or children with acquired or hereditary mild-to-severe sensorineural hearing loss.
Sites: ENT clinics, Trousseau (Pr Garabedian, Pr Desnoyelle, Dr Loundon, Dr Marlin), Robert Debré Paris (Pr van den Abbeele), Necker (Dr Couloigner), Beaujon (Pr Sterkers, Dr Bouccara) & Lariboisiere (Pr Tran Ba Huy, Dr de Waele) Hospitals, Laboratoires Entendre SAS Versailles & Reims (S Garnier, P Fleuriot, X Debruille, L Debruille).
Project #2. Taken together, these recent results reveal that the ability of the auditory system to encode and use temporal fine structure cues in incoming sounds is essential for speech identification, and that selective deficits in this ability may explain the presence, the degree and nature of speech intelligibility deficits observed in listeners with cochlear hearing loss. 'Temporal fine structure (TFS) speech' may provide an important tool in evaluating how well this goal is achieved and in the diagnosis of sensorineural hearing loss. These results suggest therefore that further work on temporal fine structure speech is warranted.
In previous studies conducted by our group, the Hilbert transform was used to extract temporal fine structure information. However, alternative processing schemes may be considered. Two additional processing schemes have been developed in collaboration with Pr S. Sheft at Parmly Hearing Institute (Loyola Univ. Chicago, USA). Those schemes represent the fine structure as a phase- or frequency- modulated tone at the audio channel center frequency, and minimize potential artifacts corresponding to temporal envelope reconstruction. We currently investigate speech perception in normal-hearing and hearing-impaired listeners using those new TFS-speech processing schemes.
Sites: ENT clinics, Beaujon Hospital (Pr Sterkers, Dr Bouccara), Laboratoires Entendre SAS Versailles & Reims (S Garnier, P Fleuriot, X Debruille, L Debruille).
Project #3. The lack of ability to use temporal fine structure cues probably also limits the ability of people with cochlear implants to understand speech when background sounds (and especially background sounds with amplitude fluctuations) are present. Improving the ability to use temporal fine structure cues should be a goal for designers of hearing aids and cochlear implants.
We currently investigate speech perception in steady and fluctuating noise in adults wearing a multichannel cochlear implant. Different speech processors are under investigation. Our current results replicate recent work showing that the improved ability to understand speech in fluctuating noise (re: steady noise) reported in normal-hearing adults is totally abolished in adults wearing a multichannel cochlear implant.
Sites: ENT clinics, StAntoine (Pr Meyer, Dr Poncet), Avicenne (Pr Frachet, Dr Poncet, Dr Ouayoun, Dr Harboun Cohen), Beaujon (Pr Sterkers, Dr Boucara) & Robert Debré-Reims (Pr Chays) Hospitals, Laboratoires Entendre SAS Reims (X Debruille, L Debruille).