Present hearing aids are characterised by impressive algorithms, which can improve the sound perception of the listener. These algorithms consist among others of beamformers to obtain signals from a specific direction, streaming of auditory signals from remote microphones, and advanced auditory scene analysis to segregate multiple speakers.
However, there is one important limitation in utilising these advanced algorithms to improve user benefit for the hearing aid users: The hearing aids do not know how to steer the algorithms in situations with multiple sound sources.
Identifying the attended speaker using EEG
Such steering signals cannot be obtained from the environment but must be extracted from the person with hearing impairment to reflect the auditory attention and intention in various situations. One possible way to identify selective auditory attention is by recording electroencephalography (EEG), which reflects the neural responses in the brain with high temporal resolution.
By formulating an individual transfer function (decoding algorithm), it is possible to identify the attended speaker by correlation of the decoded EEG signals with the audio streams. This phenomenon has been demonstrated in several research laboratories all over the world, and has high scientific value, but little value to the hearing aid industry as the EEG signals are recorded from a grid of electrodes covering the entire scalp.
Ear-EEg may be the part of future hearing aids
However, to overcome this issue, we have shown that the brain signals can be retrieved by electrodes positioned in the ear canal. This is called Ear-EEG. Such electrodes may be embedded in the ear moulds of the hearing aids, and hence provide a non-invasive and feasible solution for everyday use.
We have worked intensively in this research area for several years and been an active partner in the EU Horizon 2020 project “Cognitive control of a hearing aid” (COCOHA), which was successfully completed by the end of 2018. To read more about the COCOHA project click here.