NeuroPulse: A Browser-Native, Privacy-First Architecture for Real-Time Multimodal Cognitive Fatigue Monitoring and Psychoacoustic Remediation

Abstract

Abstract Digital knowledge workers face continuous display screen exposure, leading to cognitive fatigue and acoustic stress that impair executive function, visual reaction time, and working memory. Existing biophysiological monitoring solutions rely heavily on invasive hardware wearables or cloud-based neural processing pipelines that introduce latency jitter, operational bandwidth overhead, and severe data privacy risks. This paper presents NeuroPulse, a zero-server, fully browser-native architecture designed for real-time multimodal cognitive fatigue monitoring and adaptive sensory intervention. NeuroPulse integrates two client-side processing engines: a SaccadeGaze Engine utilizing WebAssembly-accelerated MediaPipe for 3D facial landmark tracking, micro-saccadic latency measurement using high-resolution browser timers (performance.now()), Eye Aspect Ratio (EAR) blink profiling, and 3D head-pose displacement vectors; and a Psychoacoustic Engine leveraging the Web Audio API for fast Fourier transform (FFT) spectral energy analysis and dynamic biquad notch filtering. These streams feed into a unified mathematical framework that normalizes raw telemetry via baseline Z-scores and computes a dynamic Mental Fatigue Index (MFI) smoothed by an Exponential Moving Average (EMA). Experimental benchmarks demonstrate sub-16 ms processing frame times (>60 FPS) and a zero-network telemetry footprint, establishing a private, highly scalable edge-computing paradigm for workplace biometrics. Keywords: Cognitive Fatigue Monitoring, Edge Computing, Computer Vision, Web Audio API, Psychoacoustics, Saccadic Latency, Mental Fatigue Index, Privacy-Preserving AI, Digital Ergonomics.