Auditory training strategies to improve speech intelligibility in hearing-impaired listeners

Hearing loss causes substantial challenges in understanding speech. Hearing aids often improve speech perception in quiet but difficulties in understanding speech in noise remain one of the most common problems reported by hearing-aid users. Therefore, other ways of improving the speech-in-noise perception are of interest to both research and the hearing-aid industry. Auditory training is one of possible strategies to complement the usage of hearing aids by enhancing skills relevant for speech perception in noisy environment. The experiments described in this thesis investigated the efficacy of two selected auditory training methods (based on speech and musical stimuli) in a group of experienced hearing-aid users. The first method, phoneme-in-noise training, aimed to retrain the usage of speech cues in the presence of background noise. The second method, training motor synchronization with the musical beat, was meant to enhance predictive mechanisms that contribute to speech perception. The training outcomes were assessed using metrics of speech intelligibility in noise (both methods) and listening effort (in the case of phoneme-in-noise training) and were related to the results obtained in an active control group.
In the first chapter, the influence of phoneme-in-noise training on speech intelligibility is investigated. The training effects were assessed using logatomes from Danish nonsense word corpus (DANOK) as well as a hearing in noise test (HINT). The training improved identification scores in noise for vowels and consonants located in the middle of the logatomes but not for the onset consonants. The observed improvements were still present three months after the training. No significant impact on speech recognition thresholds (SRTs) measured with the HINT was found.
The second chapter describes the effect of phoneme-in-noise training on listening effort required to understand the HINT sentences, which was measured using pupillometry. Changes in pupil dilations were recorded when the participants were listening to sentences presented at two signal-to-noise ratios (SNRs): corresponding to 50% (SNR50) and 80% (SNR80) speech intelligibility (based on the pre-intervention performance). For both test conditions, the analysis of the pupil response patterns suggested that listening effort increased in both participant groups, but the change was more pronounced in the group that completed the training. As opposed to SRTs (reported in the first chapter), the speech recognition scores measured at fixed SNR revealed improved speech intelligibility in noise in the more challenging condition (SNR50). The outcomes demonstrated that pupillometry was sensitive to the effect of training in both conditions, also when the HINT did not reveal any significant performance improvement (SNR80). The elevation of listening effort exerted during the HINT might have been driven by increased motivation (in both groups) as well as changes in the cognitive resources allocation mechanisms (only in the group that completed the training).
The third chapter presents the design and implementation of a novel musicbased training approach to support speech understanding in noise in hearing-aid users. The training program aimed at improving motor synchronization with the beat of music and was administered at the participants’ homes using a mobile application. Modest improvements in synchronized tapping consistency throughout the training were found, but they did not result in significantly better performance in the outcome measures of beat perception or speech intelligibility in noise. Longer duration and higher intensity of training aswell as more individually tailored task demands might be needed to obtain significant benefits using this method.
The results presented in this thesis can inform future development and refinement of speech- and music-based auditory training approaches as well as the selection of methods used to assess the benefits of these interventions.