Mel-frequency cepstral coefficients and spectral features directly drive micro-level autonomic reflexes during brief respiration practices

While digital health platforms increasingly deliver guided mindful breathing interventions via audio, the specific acoustic properties of the guidance are fundamentally ignored as potential active ingredients. This within-subject study implemented an unprecedented high-density experimental design to investigate how distinct, computationally engineered acoustic features (e.g., pitch, zero-crossing rate, natural background sounds) modulate acute cardiovascular, respiratory, and autonomic responses. To rigorously eliminate sex-related physiological variance, 21 meditation-naïve female university students completed 13 randomized audio sessions of a standardized 4-minute-and-5-second guided breathing practice (yielding 273 high-density recording epochs). Multiparametric biosignals, including pulse rate variability (PRV), respiratory dynamics, and electrodermal activity (EDA), were continuously recorded. Results demonstrated a robust reduction in blood pressure independent of audio conditions, confirming the fundamental efficacy of a single bout of spontaneous breath awareness. However, high-resolution analyses revealed a highly modular architecture where peripheralautonomic indices (EDA and PRV) were acutely sensitive to fine-grained acoustic features, specifically Mel-frequency cepstral coefficients (MFCCs). Furthermore, physiological coupling patterns reorganized dynamically depending strictly on the acoustic context. These indings fundamentally challenge the assumption that meditation audio is a neutral delivery medium, establishing acoustic parameters as potent, independent modulators of human autonomic reflexes for future digital therapeutics.