Phase-shift keying of EEG signals: Application to detect attention in multitalker scenarios

• We propose an M-PSK model for the detection of auditory attention in a multitalker scenario.
• We generate an M-PSK constellation from EEG signals.
• For a robust detection, we embed attention in M-PSK symbols.
• Our approach extends dichotic listening task to 4 and 6 talkers.
• Methodology could be applied in neuromarketing, attention assessment and for BCIs.

Numerous algorithms have tried the blind separation of multiple auditory sources in the framework of the cocktail party phenomenon with relative success. However, this is a straightforward goal for humans, who use selective attention to filter one target out of many irrelevant sources. In this paper we present a paradigm capable of detecting attention in multitalker scenarios. The paradigm inserts in messages barely-audible tags whose evoked brain response constitutes a M-ary Phase-shift keying (M-PSK) constellation. Then, detection is performed based on the M-PSK receiver. In two experimental sessions, we inserted tags in four and six auditory messages to elicit 4-PSK and 6-PSK constellations, respectively. We instructed thirteen participants to attend only one message and ignore the rest (forced-attention paradigm). The results presented significant detection of the attended messages above the chance level with mean accuracies of 0.47 and 0.32 for 4-PSK and 6-PSK, respectively. The results support that our paradigm detected attention to one target speech in multitalker scenarios. This outcome could be applied for the assessment of auditory attention, as assistive technology for attention impairment, in neuro-marketing or Brain–computer interfaces.