Objective Determination of Speech Perception in Noise
Final Report Abstract
The most frequently cited problem for people with hearing loss is limited speech intelligibility in situations with background noise. Speech intelligibility requires audibility and correct processing in the auditory system, up to cortical pattern recognition. In recent years, speech tests have been developed to quantify speech intelligibility, which enable a precise measurement of speech intelligibility in background noise. However, a disadvantage of all speech audiometric methods is that they require a high degree of cooperation from the test subjects and are therefore not objective. The long-term goal of the proposed project was to derive information about individual speech intelligibility by measuring brain waves (electroencephalography, EEG) in certain acoustic scenarios without any cooperation from the test subjects. In the project, speech intelligibility measurements were carried out on normal hearing adults and EEG multichannel recordings were carried out simultaneously. For this purpose, sentences from the Oldenburg sentence test were presented in silence and in different noise conditions. Using modern signal processing methods, a measure of speech intelligibility was derived from the EEG measurements. In order to take into account the interindividual differences in speech intelligibility, the individual speech intelligibility thresholds (SRT 50) were determined. The signal-to-noise ratio (SNR) in the EEG measurements were related to this individual SRT50. This ensured that comparable intelligibility rates were available. For the analysis, a linear approach was first investigated, in which a temporal response function (TRF) was determined, which allows an estimation of the multichannel EEG from the envelope of the speech signal. Systematic changes in the TRF were demonstrated as a result of SNR variation. Subsequently, a method was developed that allows the EEG measurements to be assigned to the comprehensibility. It was possible to predict intelligibility up to 2 dB. In summary, the developed methods allow a rough prediction of speech intelligibility from EEG measurements. A division into "incomprehensible", "difficult to understand" and "easy to understand" would thus be possible. In this way, the subjective measurements are objectified. The interindividual fluctuations of the EEG caused by variations in vigilance are too large for a precise determination of speech intelligibility thresholds, even in young, healthy people with normal hearing.
Publications
-
Akustisch evozierte Potentiale ausgelöst durch Oldenburger Sätze (2020) Jahrestagung der Deutschen Gesellschaft für Audiologie e.V., Köln,
Muncke, J., Schneider, V. & Hoppe, U.
-
An LMMSE-based Estimation of Temporal Response Function in Auditory Attention Decoding. 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2837-2840. IEEE.
Kuruvila, Ivine; Fischer, Eghart & Hoppe, Ulrich
-
Extracting the Auditory Attention in a Dual-Speaker Scenario From EEG Using a Joint CNN-LSTM Model. Frontiers in Physiology, 12.
Kuruvila, Ivine; Muncke, Jan; Fischer, Eghart & Hoppe, Ulrich
-
Inference of the Selective Auditory Attention Using Sequential LMMSE Estimation. IEEE Transactions on Biomedical Engineering, 68(12), 3501-3512.
Kuruvila, Ivine; Can, Demir Kubilay; Fischer, Eghart & Hoppe, Ulrich
-
Prediction of Individual Intelligibility for Speech in Noise (2021) NBT Berlin – International Forum on neural Engineering & Brain Technologies, Berlin, 17.-18. Mai 2021
Muncke, J.
-
Prediction of Speech Intelligibility by Means of EEG Responses to Sentences in Noise. Frontiers in Neuroscience, 16.
Muncke, Jan; Kuruvila, Ivine & Hoppe, Ulrich
-
Prediction of the EEG for a Speech in Noise Test Using an Auditory Model (2022) 35th World Congress of Audiology, Warsaw, 2022-April
Muncke, J. & , Hoppe, U.
-
Neural Correlates of Auditory Attention in a dual-speaker scenario. Dissertation, Friedrich-Alexander-Universität Erlangen-Nürnberg, 2023
Kuruvila, I.
