Temel Kavramlar
Adding noise-tags to speech signals enhances decoding performance for auditory attention.
Özet
This pilot study explores auditory attention decoding (AAD) using noise-tagging in a sequential paradigm. It compares modulation depths and decoding approaches, highlighting the potential of integrating noise-codes in speech for enhanced speaker identification. The study reveals insights into fundamental protocol design decisions and the application of noise-tags in auditory attention decoding.
Abstract:
AAD aims to extract the attended speaker from brain activity amidst candidate speakers.
This pilot study uses a noise-tagging stimulus protocol to enhance auditory speaker detection.
Comparisons between unmodulated audio and various modulation depths were conducted.
Introduction:
Hearing aids struggle in scenarios with multiple speakers, leading to the need for AAD.
AAD aims to decode the attended speaker from neural activity by synchronizing brain signals with the attended speech envelope.
Different approaches like EEG-based stimulus reconstruction are used for AAD.
Materials and Methods:
Five participants took part in the experiment using EEG data and Dutch short stories modulated with binary pseudo-random noise-codes.
Envelope CCA (eCCA) and reconvolution CCA (rCCA) methods were employed for classification accuracy evaluation.
Results:
Modulation conditions of 100, 90, and 70 performed better than unmodulated audio for eCCA method.
rCCA method preferred a 70 percent modulation intensity for peak performance.
Discussion:
Adding noise-tags enhances decoding performance compared to unmodulated speech signals.
Frequency range expansion due to modulation may contribute to improved performance.
Conclusion:
Integrating noise-tags in speech signals can enhance auditory attention decoding performance, paving the way for future applications in this domain.
İstatistikler
Participants: Five participants aged 19–31 years participated in the experiment.
EEG Data: Recorded at a sample rate of 500 Hz with 64 active electrodes placed according to the 10-10 system.
Alıntılar
"Adding noise-tags to a speech signal can enhance decoding performance compared to unmodulated signals."
"Frequency range expansion due to modulation may contribute to improved performance."