Using ERPs for assessing the (sub) conscious perception of noise

Area of Science:

• Neuroscience
• Auditory Perception
• Signal Processing

Background:

• Assessing audio signal quality is crucial in various applications.
• Disturbances in audio signals can impact user experience and performance.
• Objective measures for audio quality, especially for subtle disturbances, are needed.

Purpose of the Study:

• To investigate the efficacy of event-related potentials (ERPs) as a quantitative measure for assessing the quality of disturbed audio signals.
• To determine if neural responses reflect audio degradation below the threshold of conscious perception.
• To explore the P300 component and machine learning classifiers for objective audio quality assessment.

Main Methods:

• An electroencephalography (EEG) study was conducted with 11 participants.
• An oddball paradigm was used, presenting the phoneme /a/ with varying levels of signal-correlated noise.
• Analysis focused on the amplitude and latency of the P300 component and employed a shrinkage Linear Discriminant Analysis (LDA) classifier.

Main Results:

• The amplitude and latency of the P300 component effectively quantified disturbance in consciously perceived degraded audio.
• For stimuli not consciously recognized as disturbed, a shrinkage LDA classifier successfully identified signals with subconsciously processed noise in two subjects.
• Neural responses correlate with the degree of audio signal degradation.

Conclusions:

• Event-related potentials (ERPs), particularly the P300 component, offer a viable quantitative measure for audio quality assessment.
• Neural activity can detect and quantify audio disturbances even when they are below conscious awareness.
• Machine learning approaches show promise in identifying subconsciously processed audio degradation.