Improvement of Speech Intelligibility in Patients with Sensorineural Hearing Loss Using Noise Reduction Algorithms

Hong, Sung Hwa; Woo, Hyo Chang; Chung, Won Ho; Cho, Yang Sun; Ha, Byung Suk; Park, Kye Hoon; Kim, Dong Wook

Abstract

BACKGROUND
& OBJECTIVES: Speech intelligibility is significantly decreased in noisy environments in patients with sensorineural hearing loss (SNHL) compared to those without hearing loss. It is caused by increased hearing threshold, recruitment, and decreased frequency selectivity. The hearing threshold and recruitment could be grossly controlled by the hearing aid with a compression circuit. It has been well established that various techniques can be used to trigger the signal to increase the noise ratio (SNR), such as noise reduction algorithms, in order to compensate for decreased speech intelligibility in noisy environments. These algorithms are incorporated into some of the commercially available digital hearing aids. This study was designed to measure the effect of noise in speech intelligibility of patients with sensorineural hearing loss and the effect of reducing the background noise with a noise reduction algorithm.
MATERIALS AND METHODS
Twenty-one ears with sensorineural hearing loss were selected for this study. Speech reception thresholds (SRTs) and speech discrimination scores (SDSs) were checked with or without noise, and after removal of noise using a noise reduction algorithm.
RESULTS
SRTs were increased in noise compared with a quiet environment. By reducing noise using the algorithm we prepared, SRTs were decreased by 4.6 dB, 4.2 dB, and 4.9 dB at +6, 0, -6 dB SNRs, respectively, in comparison with those who didn't use the algorithm in noise. It was observed that SDSs were decreased by 24-50% in noise and patients partially recovered hearing loss when noise was reduced.
CONCLUSION
Speech intelligibility deteriorated in noise but partially improved with a noise reduction algorithm. A further study on the development of the noise reduction algorithm is needed to improve SNRs.