Six frequency bands → Z6 quantized → projection into Z6 lattice. Structured audio (speech) falls into consistent traps; noise spans many. The friction calibrator tells you which regime hides or reveals the keyword.
The 6 frequency bands are the input to the Z6 projection. Each band's energy maps to a Z6 level. The projection friction determines how many levels are distinguishable per regime. Bars = current Z6 level of each band.
Step 1: Microphone audio is FFT'd into 6 frequency bands (86–8000 Hz). Each band's energy is quantized to the nearest Z6 level [-3..3] (no zero state).
Step 2: The 6 Z6 levels are projected through three regimes (6→1, 6→3, 6→6). Each regime is a different lattice trap — a partition of the input space into Z6N output cells.
Step 3: When you speak a keyword, your voice creates a specific sequence of Z6 projected values. Noise creates random jumps between traps. The trap consistency (low variance = consistent = speech) distinguishes them.
Step 4 (Record): Say your keyword — the Z6 projection pattern is stored. During matching, the Hamming distance in Z6 space between live and template is computed. A low distance = keyword match.
DSP baseline: Simple energy threshold on all 6 bands. Less robust in noise but included for comparison.