International Hearing Aid Research Conference location:Tahoe City, California date:10-14 August, 2016
In clinical practice and research, behavioral tests are the golden standard to measure speech intelligibility. However, these methods require the active participation of the patient and an audiologist and are consequently not always suited to test special populations such as young children. Therefore, we aim to develop a more automatic and objective method to evaluate speech intelligibility, which could be a starting point for self-fitting hearing aids.
Recently, several studies have shown that cortical neural activity tracks the amplitude envelope of running speech. Given that the envelope is a primary cue for speech understanding, we hypothesize that the envelope should be well represented in the brain when speech is intelligible. This has been investigated in recent studies by applying a spatiotemporal decoder on EEG signals in order to reconstruct the stimulus envelope. The results show that the correlation between the actual and reconstructed envelope is higher when speech is presented at higher signal-to-noise ratios (SNRs). Taken together, we could use the reconstruction accuracy as a function of SNR as an objective correlate of speech intelligibility.
In normal hearing subjects, we found increased envelope entrainment with increasing SNR. As this trend is similar to a behavioral psychometric function, these results are very promising. In the current study, we explored the influence of hearing loss on our objective measure. Sentences of the Flemish Matrix speech material were presented at seven different SNRs to hearing impaired subjects while their EEG was recorded. The SNRs were chosen at fixed speech intelligibility levels for each subject.
Furthermore, research has shown that increased listening effort may cause increased neural activity. In a previous study, we found similar results since several subjects showed a decrease in envelope entrainment with increasing SNR at high intelligibility levels. We explored this effect by measuring listening effort using a dual task paradigm. The primary task involves a speech-in-noise test and the secondary task a memory test in which response time is measured. By comparing the response times between different SNRs, we obtain a measure for the listening effort needed to understand speech. The results will be presented.