Environment-specific noise suppression for improved speech intelligibility by cochlear implant users.

Area of Science:

• Audiology
• Biomedical Engineering
• Signal Processing

Background:

• Cochlear implant (CI) users often struggle with speech intelligibility in noisy environments.
• Existing noise-suppression algorithms have shown limited success due to the variability of real-world noise.
• A one-size-fits-all approach to noise reduction is challenging for diverse listening situations.

Purpose of the Study:

• To develop and evaluate environment-specific noise suppression algorithms for CI users.
• To enhance speech intelligibility in complex auditory environments.
• To investigate the efficacy of selecting signal-to-noise ratio (SNR)-based audio channels for stimulation.

Main Methods:

• Developed an algorithm that classifies audio channels as target-dominated (SNR ≥ 0 dB) or masker-dominated (SNR < 0 dB) using binary classifiers.
• Trained classifiers on data from specific noisy environments.
• Selected only target-dominated channels for stimulation in CI listeners.
• Tested the algorithm with sentences in three real-world maskers at varying SNRs.

Main Results:

• Demonstrated substantial improvements in speech intelligibility for CI users.
• Achieved an improvement of nearly 44 percentage points at 5 dB SNR.
• Showcased significant gains in environments with complex, real-world maskers.

Conclusions:

• Environment-specific noise reduction algorithms offer a promising approach to improve CI user hearing.
• The selective stimulation of target-dominated channels can restore speech intelligibility to near-quiet levels.
• This tailored strategy addresses the limitations of universal noise-suppression algorithms.