Alpha Binaural Beats and Brain Entrainment: A Machine Learning-Based Assessment of Short-Term Effects

In recent years, binaural beats (BBs) have attracted growing interest in the scientific community for their potential to non-invasively modulate brain activity, based on the brainwave entrainment hypothesis. BBs are an auditory phenomenon that occurs when two tones of slightly different frequencies are presented separately to each ear, and the brain perceives a third tone corresponding to the frequency difference. Numerous studies suggest that such stimulation can influence neurophysiological parameters and mental states related to relaxation, attention, and emotional regulation. However, current literature is highly heterogeneous in terms of protocols, frequencies, duration, and results, with most studies focusing on prolonged stimulation (≥10 minutes). This study aims to explore the efficacy of a brief (1-minute) alpha-band (7–13 Hz) stimulation to evaluate the emergence of short-term neurophysiological effects and gain insights into rapid activation mechanisms. The experiment involved 14 subjects (mean age: 23.8 ± 2.5 years) who participated in two sessions separated by a 5-minute rest period, with simultaneous EEG and ECG recordings during exposure to BBs and a sham condition, using a personalized protocol based on each participant’s Individual Alpha Frequency (IAF). Each 10-minute session alternated BB and sham segments, delivered via JBL headphones in a quiet room, with participants’ eyes closed and monitored to avoid drowsiness. EEG signals were acquired using the Enobio 8 system and ECG via a Polar H10 chest strap. Data were preprocessed through filtering, artifact rejection, and segmentation. EEG features related to spectral activity, functional connectivity, and signal complexity were extracted, along with HRV-related ECG parameters. The experimental setup was supported by NIC2 for data acquisition and MATLAB for audio generation and signal analysis. After preprocessing, several feature selection techniques were explored, and Recursive Feature Elimination (RFE) was selected to retain the most relevant features for each subject, reducing the total to 10%. Among all participants, the most significant features were related to signal complexity (Higuchi Fractal Dimension) and brain synchronization (Phase Locking Value). Subsequently, a subject-specific Support Vector Machine (SVM) classifier was implemented to track prediction trends over time during both sessions. The analysis revealed two diverging trends: an increasing trend for BB recognition and a decreasing one for sham. However, the classification accuracy did not show significant deterioration, maintaining good overall recognition rates (Sham = 70.54% ± 8.00%, BBs = 68.75% ± 7.54%). Wilcoxon signed-rank tests showed no significant differences between the two sessions (p ≥ 0.05). In conclusion, the study demonstrated the presence of short-term effects induced by BB stimulation, although it did not clearly distinguish between conditions across all sessions.Future studies could aim to determine the minimum duration of BB stimulation required to produce a consistent and significant effect by modifying the experimental protocol. Additionally, increasing the sample size and employing a high-density EEG system would be necessary to reduce inter-subject variability and improve the overall quality of the analysis.