Wake-Word Detector: Duty Cycle & False-Accept Simulator

Every "Hey Siri" or "OK Google" moment starts with a tiny, always-on detector running a matched-filter correlation against a stored keyword template on a handful of milliwatts, long before any cloud or on-device speech recognizer wakes up. This simulator streams a live 3D ribbon of analysis frames, scores each one with normalized cross-correlation, and lets you tune the detection threshold, background noise floor, how often the keyword is actually spoken, and the microphone's power-saving duty cycle. Watch true accepts, phantom false-accept wake-ups, and missed false rejects accumulate in real time, alongside an estimated average power draw that reflects the real two-state design — cheap always-on listening punctuated by expensive full-recognition bursts — used in real mobile voice assistants.