Study of a Signal-processing Algorithm Aiming at Improving Speech-in-noise Intelligibility in Normal-hearing and Hearing-impaired Persons

Pilot Study of a Signal-processing Algorithm Aiming at Improving Speech-in-noise Intelligibility in Normal-hearing and Hearing-impaired Persons

Nearly half a billion people suffer from disabling hearing loss. The most common form of hearing loss in adults is age-related hearing loss (ARHL), which causes a reduced ability to understand speech in noisy environments. The ability of people with ARHL to communicate is therefore greatly impacted, limiting their social interactions and thus their quality of life. Yet, the wear of hearing aids - which is the current standard rehabilitation treatment in such cases - does not lead to optimal satisfactory outcomes when it comes to understanding speech in noisy environments.

The objective of this pilot study is to test a new signal-processing algorithm, based on artificial intelligence, that aims at enhancing the intelligibility of speech-in-noise signals. The efficiency of the algorithm is compared to a standard denoising algorithm commonly used in hearing aids. The primary outcome measure is the word-identification performance of the participants, using the FrMatrix test (Jansen et al., 2012). Two secondary outcome measures are investigated: listening effort (self-assessed using a Likert scale, and measured through response times), and subjective preference (assessed in a paired-comparison task).

The study is conducted in 20 normal-hearing subjects and in 40 older (age ≥ 55 years) hearing-impaired subjects.

Inclusion criteria for normal-hearing participants :

• consenting to and available for the study
• age ≥ 18 years and ≤ 25 years
• right-handed
• native French speaker
• pure-tone air-conduction audiometric thresholds at .5, 1, 2, and 4 kHz all ≤ 20 dB hearing loss (HL)

Exclusion criteria for normal-hearing participants :

• score at the Tinnitus Handicap Inventory > 56
• uncorrected visual impairment
• pregnancy
• subject placed under legal authority (guardianship, tutorship)

Inclusion criteria for hearing-impaired participants :

• consenting to and available for the study
• age ≥ 55 years
• native French speaker
• pure-tone air-conduction average hearing thresholds (PTA) for frequencies of .5, 1, 2, and 4 kHz ≥ 30 dB HL and ≤ 70 dB HL
• right-ear PTA for low frequencies (< 2 kHz) lower than the right-ear PTA for high frequencies (> 4 kHz). The difference between the two PTAs must be ≥ 20 dB
• the difference between the PTA for frequencies of .5, 1, 2, and 4 kHz (PTA.5-4kHz) in the right ear and the left-ear PTA.5-4kHz must be ≤ 10 dB

Exclusion criteria for hearing-impaired participants :

• score at the Tinnitus Handicap Inventory > 56
• history of a hearing impairment that was left uncorrected for more than 10 years
• uncorrected visual impairment
• congenital hearing impairment
• Ménière's disease
• auditory neuropathy
• mixed hearing loss
• fluctuating hearing impairment
• sudden hearing loss
• hearing impairment concomitant with other symptoms (e.g. vertigo)
• score at the Montreal Cognitive Assessment (MoCA) < 27
• subject placed under legal authority (guardianship, tutorship)