# Truth Detector

> Truth Detector ("Are You A Liar?") is a real-time web-based voice stress and pitch analysis tool designed to detect vocal micro-tremors and fundamental frequency shifts. Running entirely client-side in the browser, it evaluates vocal indicators to calculate a Truth or Lie score while guaranteeing 100% audio privacy.

Truth Detector processes live microphone audio streams using the HTML5 Web Audio API and custom digital signal processing algorithms in `audioAnalyzer.js`. An `AudioContext` and `AnalyserNode` (FFT size 2048) capture raw time-domain PCM samples at 60 frames per second. An autocorrelation algorithm extracts the fundamental frequency (pitch) and root-mean-square (RMS) energy level for each frame.

The application operates in two phases: baseline calibration and test evaluation. During calibration, the user speaks a known truthful statement to establish baseline vocal metrics including pitch, jitter, and shimmer. During the test phase, the user's responses are evaluated against this baseline. The system tracks pitch deviation percentage, micro-tremor jitter, amplitude shimmer, energy variance, and speech hesitations to compute a composite stress score from 0% to 100%. All processing occurs strictly in client memory, with zero audio transmission or cloud storage.

## Overview

- [Truth Detector Overview](/index.md): Complete guide to application architecture, acoustic analysis parameters, usage workflow, and technical stack.

## Optional

- [Acoustic Analysis & Algorithms](/analysis.md): Technical breakdown of autocorrelation, parabolic interpolation, jitter/shimmer equations, and stress scoring heuristics implemented in `audioAnalyzer.js`.