Improving speech intelligibility of severely hearing- impaired people by frequency-lowering technique

Abstract

Frequency lowering is an important audio signal processing method used in digital hearing aids. It helps patients with severe hearing loss in high frequencies by compressing speech into the patients’ residual hearing regions. In this study, a general frequency lowering algorithm was proposed, which could separately process the contents in the disabled high frequencies and those in the low frequencies as well; and two different lowering strategies (the overlapped and segmented compression strategy) based on it were implemented, compared with traditional proportional frequency compression algorithm. The respective performances of both strategies were then evaluated using a specially designed software system, which simulated the signal processing procedure of digital hearing aids. After fitting this system to the patients, the results of a detailed speech intelligibility test showed that both of the frequency-lowering strategies provided more benefits, especially in consonant recognition, as compared with the hearing aids used by the patients. Furthermore, the segmented compression strategy showed better ‘anti-noise’ quality and was thus preferred by the patients. Finally, the impact of noise and training to the benefits of frequency lowering algorithms was also discussed in this paper. Key words: Hearing aid, frequency lowering algorithm, speech intelligibility, compression strategy, software system development, noise.