Multimodal EEG–fNIRS Brain–Computer Interface Framework for Computational Analysis of Neurocognitive Responses During Vishnu Sahasranama Mantra Chanting: Neural Oscillatory Patterns, Cortical Hemodynamics, and Multimodal Biomarker Characterization
P Sandhya Jeevan K P
Journal of Intelligent Decision Making and Information Science August 21, 2026 DOI: 10.59543/jidmis.v3.2241 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Within-subjects experimental study Peer reviewed |
|---|---|
| Sample size | 90 |
| Population | Human participants (not further specified) |
| Intervention | vocal chanting |
| Duration | 270-second protocol with 30-second resting baseline |
| Key findings | Vocal chanting uniquely increased beta power (+0.886 × 10⁻¹² V²/Hz above listening), while theta and gamma bands showed increasing trends with internalization depth. Prefrontal HbO increased across conditions (listening: +0.0249 mM; vocal: +0.0558 mM; silent: +0.0358 mM). The framework differentiated the three mantra modes. |
Abstract
Background and Motivation: Mantra recitation is a structured phonetic-cognitive practice, within the context of the Vedic tradition, which involves activity in the auditory, phonological, attentional, and motor cortical networks, but this has not yet been well characterized using multimodal neuroimaging analysis.
Methods: The framework of simultaneous EEG–fNIRS Brain–Computer Interface (BCI) is presented, and validated on a 90-subject physiologically parameterized benchmark data set designed after published neurophysiology during meditation. A within-subjects design with a 270-second protocol was used and examined three conditions: passive listening (Shravana), vocal chanting (Vachika Japa), and silent mental recitation (Manasa Japa) with a 30 second resting baseline condition. The EEG signals were preprocessed using fourth-order Butterworth filtering (0.5–45 Hz) and 50 Hz IIR notch filter for notch filtering, average referencing, and FastICA for artifact rejection (24-channel SmartBCI MitsarEEG at 500 Hz). The processing of simultaneous fNIRS (NIRSIT NS1-H20B, 204 channels, 8.138 Hz, dual wavelengths: 780/850 nm) included Modified Beer-Lambert Law (mBLL) optical-density conversion and hemodynamic bandpass filtering (0.01–0.5 Hz). Feature extraction produced a 503 dimensional multimodal vector for each condition epoch: 95 dimensional welch PSD band-power features (19 EEG channels × 5 frequency bands) and 408 dimensional hemodynamic features (204 × ΔHbO + ΔHbR).
Results: One-way ANOVA confirmed significant condition effects for frontal theta (F(2,267) = 15.29, p < 0.001, η² = 0.103), alpha (F = 22.21, p < 0.001, η² = 0.143), beta (F = 153.32, p < 0.001, η² = 0.535), and gamma (F = 15.10, p < 0.001, η² = 0.102). Beta power was the only band to significantly increase with Vocal Chanting (+0.886 × 10⁻¹² V²/Hz above Listening; p < 0.001), whereas the theta and gamma bands showed an increasing trend with the depth of internalization. Prefrontal HbO increased monotonically across conditions (Listening: +0.0249 mM; Vocal: +0.0558 mM; Silent: +0.0358 mM; all F(2,267) = 255.76, p < 0.001, η² = 0.657).
Conclusions: The framework is able to differentiate three engagement modes of the mantra in the respective neural signatures of the three modes, which provides a replicable computational basis for future cognitive state classification and large-scale human validation.